• Suppr超能文献
  • 文献检索
  • 文档翻译
  • 深度研究
  • 学术资讯
  • Suppr Zotero 插件Zotero 插件
定价套餐&价格
应用&插件
Suppr Zotero 插件Zotero 插件浏览器插件Mac 客户端Win 客户端微信小程序
定价
会员套餐积分包API 积分包
服务
文献检索文档翻译深度研究API 文档MCP 服务
关于我们
关于 Suppr公司介绍联系我们用户协议隐私条款
关注我们

Suppr 超能文献

核心技术专利:CN118964589B侵权必究
粤ICP备2023148730 号-1Suppr @ 2026
  1. 首页
  2. 分享广场
  3. 生成式AI治理、算法偏见与法律科技交叉领域可落地研究选题集

生成式AI治理、算法偏见与法律科技交叉领域可落地研究选题集

深度研究匿名用户发表于 2026年05月06日 15:438阅读
发起深度研究
发起深度研究

1. 选题背景与设计规范

1.1 交叉领域研究现实需求与缺口分析

当前,以ChatGPT为代表的生成式人工智能(Generative AI)技术正以前所未有的速度融入社会各领域,展现出巨大的生产力提升潜力12345。然而,在生成式AI规模化落地的同时,其固有的算法偏见风险日益凸显,对社会公平、法律公正乃至个体权益构成潜在威胁267891011。这些偏见可能源于训练数据、模型设计、算法限制或产品策略等多个环节,导致模型输出结果存在歧视性、不准确或不公平的现象67811。特别是在医疗、教育、金融、公共服务等关键领域,算法偏见可能带来严重的伦理、社会和法律后果8101112。例如,在医疗诊断中,偏见可能导致对特定人群的误诊;在教育评估中,可能加剧不平等;在司法审判中,则可能影响公正性。

与此同时,针对生成式AI的治理规则尚处于探索阶段,现有法律法规体系与技术发展的速度之间存在显著鸿沟1012。各国政府和国际组织正积极研究和制定相关政策,但如何有效规制生成式AI的算法偏见,平衡技术创新与风险控制,仍是全球性难题。尤其是在快速变化的法律科技领域,如何将前沿AI技术与法律实务深度融合,开发出能够有效识别、评估和缓解算法偏见,并辅助法律专业人士进行决策的工具,是当前亟待解决的实际问题。现有的法律科技工具在处理生成式AI带来的新型法律问题(如知识产权归属、侵权判定、合规管理等)方面存在供给不足,难以满足司法实践和企业合规的迫切需求。

鉴于上述现实需求与缺口,本研究选题集旨在聚焦生成式AI治理、算法偏见与法律科技的交叉领域,以“可落地”为核心导向。每个选题都力求兼顾学术研究的理论深度与实践应用的价值,致力于探索具有创新性、前瞻性和操作性的解决方案。通过系统性地梳理研究问题、构建理论框架、明确数据来源、设计研究方法,并参考代表性文献,本选题集旨在为相关领域的学术申报、政策制定和产业落地提供有力的支撑。

1.2 选题核心维度设置说明

为确保本研究选题集能够有效服务于学术申报、政策制定和产业落地等多元需求,所有选题均统一覆盖以下六个核心维度:研究问题、理论框架、数据来源、方法设计、代表性文献和创新点。这些维度不仅构成了一个完整的学术研究范式,也为实际应用提供了清晰的指引。

首先,研究问题是每个选题的出发点,它明确了研究的核心目标和待解决的实际难题。一个清晰、具体的研究问题是成功研究的基础,能够指导后续理论构建和方法选择。在学术申报中,研究问题是评估项目创新性和可行性的关键要素;在政策制定中,它直接指向了政策需要解决的社会痛点;在产业落地层面,它反映了企业或行业面临的实际挑战。

其次,理论框架为研究问题提供了概念基础和分析工具。通过整合现有理论(如算法问责理论、多中心治理理论、知识产权法基本原理等),研究能够站在巨人的肩膀上,避免重复劳动,并为研究结论提供坚实的学理支撑。对于学术申报而言,理论框架的深度和广度是衡量研究水平的重要指标;对于政策制定者,理论框架有助于理解问题的本质,从而制定出更具前瞻性和系统性的政策;对于产业界,理论框架则能帮助企业理解生成式AI应用中可能存在的风险与机遇,指导其制定战略。

第三,数据来源明确了研究所需的具体信息和资料。数据是支撑研究结论的基石,其可靠性和丰富性直接影响研究成果的质量。本选题集强调多源异构数据的获取,包括运行日志、投诉案例、政策文本、裁判文书、调研问卷等,旨在确保研究的实证性和说服力。在学术申报中,明确的数据来源方案展现了研究的可操作性;在政策制定中,基于充分数据的分析能够提高政策的科学性;在产业落地中,真实业务数据的支撑是产品开发和优化的前提。

第四,方法设计详细阐述了研究问题解决的具体路径和技术手段。这包括定性分析(如多案例比较分析)、定量分析(如结构方程模型、受控实验),以及利用联合显示等工具整合定性与定量数据以获得新见解的混合研究方法13。严谨的方法设计保证了研究结果的科学性和可信度。学术申报高度重视研究方法的新颖性和适用性;政策制定需要依赖科学方法评估政策效果;产业落地则离不开有效的方法论来指导技术开发和产品测试。

第五,代表性文献不仅展示了研究者对现有知识体系的掌握程度,也为进一步深入研究提供了入口。通过梳理相关领域的国内外核心研究成果(例如,涉及IoT、系统综述方法或结构方程模型的文献1415161718),本部分旨在确保研究的前沿性和学术规范性。在学术申报中,对代表性文献的引用体现了申请者的研究基础;在政策制定和产业落地中,参考前人研究可以帮助避免弯路,吸取经验。

最后,创新点是每个选题的价值核心,它指出了本研究相较于现有工作的独特贡献。创新点可能体现在理论突破、方法创新、实践应用模式或技术工具的开发等方面。无论是学术界、政策界还是产业界,都高度关注创新。清晰的创新点是学术项目获得资助、政策获得采纳、产品获得市场认可的关键所在。

通过统一设置这六个核心维度,本研究选题集旨在构建一个系统、全面且具有高度实践指导意义的框架,以期在生成式AI治理、算法偏见与法律科技的交叉领域,推动形成兼具学术深度与落地价值的研究成果。

2. 公共服务场景生成式AI算法偏见的法律规制路径研究

2.1 核心研究问题

本研究旨在深入探讨在政务服务、民生保障等公共领域中,生成式AI应用所产生的算法偏见如何导致公民权益受损,并在此基础上,系统性地构建一套可操作的法律规制框架和多主体协同治理机制。具体而言,核心研究问题包括:

  1. 公共服务场景下生成式AI算法偏见的具体表现及其对公民权益的影响机制是什么?
    • 此问题将聚焦于识别在政府决策辅助、公共资源分配、社会福利评估、教育入学审查、医疗健康管理(如医疗AI在诊断中的应用可能出现的偏见19202122)等公共服务环节中,生成式AI可能出现的偏见类型(如歧视性、不公平性、不透明性),并分析这些偏见如何通过技术逻辑或应用流程,直接或间接损害特定群体(例如,基于性别、年龄、种族、社会经济地位等)的平等机会、隐私权(生成式AI存在数据隐私风险20)、知情权以及获得基本公共服务的权利。研究将特别关注“黑箱”模型的决策过程缺乏可解释性(尤其是在临床实践中对可解释性的需求2223)和数据偏差(训练数据的非代表性20)如何加剧权益侵害。
  2. 现有法律法规和伦理规范在规制公共服务场景生成式AI算法偏见方面的适用性与局限性何在?
    • 此问题将评估当前《民法典》、《个人信息保护法》、《数据安全法》以及各行业伦理准则等法律法规,在应对生成式AI算法偏见所引发的新型法律问题(如责任划分不清20、AI法律地位未明确24)时的有效性和不足之处。研究还将分析国内外在AI治理方面的最新进展(例如中国正在积极构建医疗领域生成式AI的伦理和法律治理框架,但监管体系仍不完善24),并识别现有法律框架在追责、补救、监督等方面的空白。
  3. 如何设计一套涵盖算法全生命周期管理的法律规制体系,以有效预防和纠正公共服务场景中的生成式AI算法偏见?
    • 此问题旨在探索从算法设计、数据采集、模型训练、部署到持续监测和迭代的全生命周期中,法律应如何介入并发挥作用。研究将考虑引入强制性算法影响评估、数据审计、透明度要求、可解释性标准以及偏见缓解技术等法律工具,并探讨如何通过法律手段确保这些要求得以落实。
  4. 在公共服务场景下,如何构建政府、企业、公民社会和技术专家等多主体协同参与的治理机制,共同应对生成式AI算法偏见问题?
    • 此问题将着眼于建立一种多方协作的治理模式,明确不同主体在算法偏见治理中的角色和责任。例如,政府作为监管者和规则制定者,应如何发挥主导作用;技术公司作为算法开发者和提供者,应承担何种社会责任和法律义务;公民社会组织和技术专家如何参与到算法监督、评估和建议中,以形成有效的制衡和监督力量,确保AI在医疗领域的应用中能最大程度地保障患者权益19。研究将探索在法律框架下,如何通过激励机制、信息共享、公众参与等方式,促进各主体之间的有效互动与合作。

本研究的核心在于通过上述问题的回答,为公共服务领域生成式AI的健康发展提供坚实的法律和治理基础,最大程度地规避算法偏见带来的风险,保障公民的合法权益和社会的公平正义。

2.2 理论框架

本研究将融合算法问责理论(Algorithm Accountability Theory)、行政法比例原则(Proportionality Principle in Administrative Law)以及多中心治理理论(Polycentric Governance Theory)来构建分析框架,以全面、深入地解析公共服务场景下生成式AI算法偏见的法律规制路径。

2.2.1 算法问责理论(Algorithm Accountability Theory)

算法问责理论是分析人工智能系统责任归属和治理机制的核心理论之一。它强调在算法决策过程中,需要明确谁对算法的设计、开发、部署和运行结果负责,以及如何追究这些责任。在生成式AI语境下,算法问责理论提供了一个审视其偏见问题的基本视角。此理论认为,算法的“黑箱”特性不应成为逃避责任的理由,而应通过引入透明度(Transparency)、可解释性(Explainability)、公平性(Fairness)和审计性(Auditability)等原则,确保算法决策过程的公正性和可审查性 2526。

本研究将利用算法问责理论,深入探讨以下方面:

  • 责任主体识别: 区分生成式AI在公共服务中产生偏见时,开发者、部署者(政府部门)、运营者以及数据提供者各自应承担的法律责任边界。
  • 问责机制设计: 考察如何建立有效的问责路径,包括事前影响评估、事中监测和事后补救措施,确保受偏见影响的公民能够获得有效救济。
  • 透明度与可解释性要求: 借鉴问责理论对算法透明度和可解释性的强调,探讨在公共服务场景下,生成式AI系统需达到何种程度的“算法公开”,以及如何向非专业人士解释其决策逻辑,以增强公众信任。

2.2.2 行政法比例原则(Proportionality Principle in Administrative Law)

行政法比例原则是现代行政法中的一项核心原则,要求行政主体在行使公权力时,所采取的措施应与所追求的目的之间保持适当的平衡,不得过度或不足。具体而言,它包括适当性(Suitability)、必要性(Necessity)和狭义比例性(Proportionality in the strict sense)三个子原则。在公共服务中使用生成式AI进行决策,本质上是行政机关利用技术手段行使公权力,因此必须受到比例原则的约束。

本研究将运用比例原则分析以下问题:

  • 必要性审查: 评估在特定公共服务场景中引入生成式AI决策系统的必要性,是否存在其他对公民权益侵犯更小的替代方案。例如,在福利分配中,AI的使用是否真正提高了效率而非仅仅扩大了歧视。
  • 适当性审查: 考量生成式AI系统是否能够有效实现公共管理目标,同时避免或最小化对公民基本权利(如隐私权、平等权)的潜在侵害。这包括对训练数据和算法模型本身是否“适合”其公共服务应用目的的评估。
  • 狭义比例性权衡: 在生成式AI带来的公共利益(如效率提升、资源优化)与可能导致的公民权利损害(如算法偏见造成的歧视、隐私泄露)之间进行利益衡量,确保公共利益的实现不以牺牲或过度侵犯个体权利为代价。

2.2.3 多中心治理理论(Polycentric Governance Theory)

多中心治理理论认为,治理不应仅仅依赖于单一的中央权威,而应由多个相互关联但相对独立的治理中心共同参与。这些中心可以是政府机构、私营企业、非政府组织、行业协会、技术社区乃至公民个体。面对生成式AI治理的复杂性、跨域性和快速演变性,单一主体的治理模式难以奏效,多中心治理提供了更具弹性和适应性的框架 2728。

本研究将从多中心治理的视角,构建公共服务场景下生成式AI算法偏见的协同治理机制:

  • 主体多元化: 识别并明确政府部门、AI技术公司、法律科技企业、社会组织、专家学者及受影响公众等各方在算法偏见治理中的角色定位、权责范围和相互关系。
  • 规则来源多样性: 探讨如何整合国家法律法规、行业自律规范、技术标准、伦理指南以及国际协议等不同层次和来源的规则,形成一套兼容并包的规范体系。
  • 互动与协调机制: 分析如何在不同治理主体之间建立有效的沟通、协调和冲突解决机制,促进信息共享、经验交流和共同决策,以应对算法偏见带来的复杂挑战。

通过整合这三大理论,本研究将能够从责任归属、公权力行使限制以及治理模式构建等多个维度,为公共服务场景下生成式AI算法偏见的法律规制路径提供一套全面而富有操作性的分析框架。

2.3 数据来源

为确保研究的实证性、客观性和全面性,本研究将广泛收集和利用多源异构数据,以支撑对公共服务场景生成式AI算法偏见及其法律规制路径的深入分析。这些数据将涵盖生成式AI应用的运行记录、用户反馈、政策法规文本以及公众认知等多个层面。

具体的数据来源包括:

  1. 各地公共服务生成式AI应用的运行日志:
    • 将通过与政府部门、公共服务提供商等机构合作,获取其在政务咨询、智能客服(例如政务领域的聊天机器人29)、自动化审批、民生政策推荐等场景中部署的生成式AI系统的匿名化运行日志。这些日志数据将包含AI模型的输入、输出、决策过程(如适用)、用户交互记录、以及可能的用户满意度或投诉标记。通过对这些海量日志的分析,可以识别出特定模式下的异常输出、服务偏差、以及可能指向算法偏见的性能指标,例如,对特定人群请求的处理时间显著延长,或对某些关键词的回答存在刻板印象等。虽然直接获取核心算法参数或训练数据可能面临技术和合规挑战,但通过运行日志可以间接推断算法行为和潜在偏见模式。
  2. 公开的算法偏见投诉案例:
    • 将收集国内外已公开的、与公共服务领域生成式AI应用相关的算法偏见投诉案例,包括媒体报道、学术研究中引用的案例、以及公民个人或社会组织提交给监管机构的投诉记录。这些案例将作为定性研究的重要素材,通过内容分析法深入剖析偏见的具体表现形式、受影响群体的特征、造成的损害类型以及现有的处理机制。例如,某些智能推荐系统可能由于训练数据的偏颇,导致对特定地域或教育背景的公民在信息获取上处于劣势,这些都将是重要的分析对象。
  3. 相关政策文本:
    • 本研究将系统性地收集和整理与生成式AI治理、算法偏见规制、数据保护、行政行为规范等相关的法律法规、部门规章、政策文件、技术标准和伦理指南。这包括但不限于国家层面的《个人信息保护法》、《数据安全法》、生成式AI服务管理办法等,以及地方政府出台的AI应用管理规定,国际组织或国家颁布的AI伦理准则(例如,关于AI在医疗领域应用伦理原则的文件26)。通过对这些政策文本的比较分析,可以评估现有法律框架的适用性与局限性,识别监管空白与冲突,并为构建完善的法律规制体系提供依据。对政策文本的分析还将结合对AI政策制定框架的理解,例如香港高校在AI教育政策制定方面的经验30,来评估政策的全面性。
  4. 公众调研问卷数据:
    • 将设计并实施针对公共服务领域生成式AI用户(即普通公民)的问卷调查,以获取公众对AI应用的认知程度、信任度、对算法偏见的担忧、以及对法律规制和多主体治理机制的期望和建议。问卷将包含多维度问题,如公民在使用公共服务AI工具时是否遇到过不公平对待、对AI决策透明度的需求、对个人数据保护的重视程度等。例如,在城市规划中,公众对AI工具的接受度、偏好互动方式以及对隐私和偏见的担忧是重要的考量因素31。通过分析这些数据,可以量化公众感知与接受度,并检验治理要素之间的潜在关联,为政策制定提供民意基础。对学生群体的调查显示,他们普遍对生成式AI持积极态度,但同时也有关于准确性、隐私、伦理和个人发展影响的担忧32。

这些多样化的数据来源将为本研究提供坚实的证据基础,有助于从技术、法律、社会和伦理等多个维度全面剖析公共服务场景生成式AI算法偏见问题,并在此基础上提出具有针对性和可操作性的法律规制与治理策略。

2.4 方法设计

本研究将采用一种混合研究方法(Mixed Research Methods),结合定性分析与定量分析的优势,以期全面、深入地探讨公共服务场景中生成式AI算法偏见的法律规制路径。具体方法包括多案例比较分析、结构方程模型(Structural Equation Modeling, SEM)检验治理要素相关性以及政策仿真模拟规制效果。

2.4.1 多案例比较分析 (Multiple Case Study Comparative Analysis)

目的与应用: 多案例比较分析旨在通过对不同国家或地区在公共服务领域生成式AI算法偏见规制实践的深入考察,识别成功的经验、面临的挑战以及潜在的普遍规律。本研究将选取3-5个具有代表性的案例,这些案例应涵盖不同法律文化背景、不同技术发展水平和不同规制模式的司法辖区(例如,欧盟、美国、中国以及某些在AI治理方面具有创新实践的城市或国家)。每个案例将围绕其公共服务AI应用的算法偏见表现、现有法律法规的应对、政府监管实践、企业自律情况以及公民社会参与程度等方面展开详细分析。

具体步骤:

  1. 案例选择与界定: 根据预设的筛选标准(如公共服务领域AI应用类型、算法偏见发生频率及影响程度、规制框架成熟度等)选取典型案例。
  2. 数据收集: 主要通过文献综述(法律法规、政策文件、学术报告)、公开投诉案例分析、相关新闻报道和专家访谈(如可能)等方式,对每个案例进行全面、细致的数据收集。
  3. 案例内分析: 对每个选定案例的内部机制和特点进行深入剖析,理解其算法偏见的具体形式、成因、以及现有规制措施的运作方式和效果。
  4. 案例间比较: 在案例内分析的基础上,运用横向比较的方法,识别不同案例在规制理念、法律工具、治理模式、主体责任划分等方面的异同。例如,比较不同法域在算法透明度、可解释性要求上的具体差异,以及这些差异对偏见治理效果的影响。
  5. 模式识别与理论构建: 从比较分析中抽象出具有普遍意义的规制模式、成功要素和失败教训,为构建适用于我国公共服务场景的法律规制与多主体协同机制提供实证基础和理论支撑。

预期贡献: 揭示不同规制策略的优劣,为我国借鉴国际经验、形成符合国情的生成式AI算法偏见规制路径提供依据。

2.4.2 结构方程模型 (Structural Equation Modeling, SEM) 检验治理要素相关性

目的与应用: 结构方程模型是一种强大的多元统计分析工具,适用于检验理论模型中各潜在变量之间的复杂关系33。本研究将利用SEM,量化分析算法问责、比例原则、多中心治理等理论框架中的核心要素(如算法透明度、可解释性、公众参与、政府监管强度、企业合规投入等)与生成式AI算法偏见的发生频率、影响程度以及规制效果之间的因果或相关关系。通过SEM,可以检验理论框架的假设,并识别出在公共服务场景下,哪些治理要素对算法偏见的有效防控起关键作用。

具体步骤:

  1. 概念模型构建: 基于理论框架和案例分析的初步发现,构建一个包含潜在变量(如“算法透明度”、“多主体协同”、“规制有效性”)和观测变量(如公众对AI决策的理解度、投诉处理效率、政策满意度等)的概念模型。
  2. 问卷设计与数据收集: 设计结构化的调查问卷,面向公共服务AI系统的使用者(公民)、管理者(政府部门人员)和开发者(技术企业员工)进行大规模抽样调查,收集观测变量数据。问卷将包含李克特量表等形式,量化受访者对各项治理要素的感知和评价。
  3. 模型识别与参数估计: 利用专业的统计软件(如AMOS, R, Mplus等)对收集到的数据进行处理,并对结构方程模型进行识别和参数估计。
  4. 模型拟合度检验与修正: 评估模型的拟合度指标,若拟合不佳,根据修正指数对模型进行合理修正,直至达到良好拟合。
  5. 结果解释与假设检验: 根据模型估计的路径系数,检验各潜在变量之间的关系是否显著,并对理论假设进行验证。

预期贡献: 量化揭示不同治理要素在规制算法偏见中的作用机制和相对重要性,为政策制定提供数据支持和优先顺序建议。

2.4.3 政策仿真模拟规制效果 (Policy Simulation for Regulatory Effectiveness)

目的与应用: 政策仿真模拟是一种通过建立数学模型或计算模型来模拟不同政策干预措施下系统行为和结果的方法。本研究将基于多案例分析和SEM的结果,构建一个简化但能反映核心逻辑的生成式AI算法偏见规制效果仿真模型。该模型将纳入关键的政策变量(如引入强制性算法审计、提高数据质量标准、加强第三方监督、设立专门申诉渠道等),模拟在不同规制强度和组合下,算法偏见发生率、公民权益受损程度以及公共服务效率等指标的变化。

具体步骤:

  1. 模型构建与参数设定: 基于前述研究结果,识别影响算法偏见规制效果的关键变量,并确定其相互作用机制。设定不同政策干预措施的参数值。
  2. 场景设计: 设计多种政策组合场景,例如“高透明度+强监管”、“低透明度+弱监管”、“强化公众参与”等。
  3. 运行模拟与结果分析: 运行仿真模型,观察在不同政策场景下,算法偏见规制效果的动态变化。分析关键指标(如偏见发现率、偏见纠正周期、公众满意度)在不同情境下的表现。
  4. 敏感性分析: 对模型中的关键参数进行敏感性分析,评估参数变化对仿真结果的影响,以增强模型的鲁棒性。

预期贡献: 为政策制定者提供“假设情景”下的决策支持,评估不同规制策略的潜在影响,优化政策选择,避免实践中的“试错成本”,最终形成更具前瞻性和有效性的法律规制方案。

通过上述三种方法的有机结合,本研究旨在从理论、实证和政策实践层面,全面、深入地为公共服务场景生成式AI算法偏见的法律规制路径提供可落地的研究成果。

2.5 代表性文献

本节将梳理公共部门AI治理、算法偏见法律规制以及生成式AI伦理规范领域的核心中英文研究成果,以确立本研究的学术基础,并识别现有研究的缺口。

2.5.1 公共部门AI治理

公共部门人工智能治理的研究日益受到关注,这反映了AI技术在政府服务中日益增长的应用以及随之而来的复杂治理挑战。早期研究主要集中于识别AI在公共部门应用的机遇与风险。Zuiderwijk等人(2021)通过系统文献综述,揭示了AI在公共治理中应用的探索性、概念性、定性和实践驱动研究的现状,并提出了一个涵盖流程和内容两方面的研究议程,强调未来研究需更加关注公共部门、实证性、多学科交叉和特定AI形式 34。他们指出,AI在公共部门的治理模式、绩效评估和扩展影响等方面仍需深入研究。Wirtz等人(2022)提出了一个基于风险和指南的整合框架来治理人工智能,为公共部门AI治理提供了系统性视角 35。

一些学者进一步关注了AI应用对公共行政传统问责机制的冲击。Busuioc(2021)探讨了AI算法系统在公共领域(如招聘、教育、执法和司法裁决)中带来的问责挑战,并强调了在算法决策中保障问责的重要性 36。Bracci(2022)则指出,AI算法在公共服务中的引入改变了责任链,提出了“智能”问责的研究议程,认为需要通过问责治理和技术解决方案来弥合问责差距 37。这些研究均强调了在AI决策过程中,确保透明度、可解释性和可追责性的关键性。

此外,AI在公共部门的应用还涉及到公共价值的实现。Chen等人(2023)通过系统文献综述和美国政府雇员的调查,探讨了公共部门AI对公共价值的影响以及治理挑战和解决方案,强调了包容性公共价值、透明度以及利益相关者参与和协作的重要性 38。Valle-Cruz等人(2019)从公共政策视角审视了AI在政府中的趋势及潜力,但也警示了“算法偏见”等负面结果,并指出了AI在公共卫生、气候变化政策、公共管理等领域的潜在益处 39。

2.5.2 算法偏见法律规制

算法偏见作为AI系统普遍存在的问题,其法律规制已成为一个全球性的热点。Lendvai和Gosztonyi(2023)将算法偏见视为人工智能时代的核心法律困境,通过比较法律分析,审视了美国和欧盟在解决系统性偏见风险方面的监管方法,并强调了现有监管在检测和补救偏见方面的持续执行差距,特别是对于不透明的“黑箱”算法设计 40。他们呼吁全球合作,以制定统一的国际标准来治理跨国AI系统,防止算法偏见加剧现有不平等。

针对特定领域的算法偏见规制也引起了广泛关注。例如,在医疗健康领域,Wang等人(2022)比较了中国、美国和欧盟在使用AI处理医疗数据方面的隐私保护规则,指出了概念边界、知情同意模式和数据跨境流动规则中的关键问题,并建议中国应建立专门的医疗信息保护法规,明确分类并实施更严格的问责机制 41。Elendu等人(2023)和Pham(2024)的综述均强调了医疗AI中隐私、数据安全、算法偏见、透明度、可解释性及责任归属等伦理和法律问题,并呼吁多学科合作制定适应性强的全球框架,以确保AI安全、公平地应用于医疗领域 4243。

O’Connor和Liu(2023)则聚焦于AI技术中性别偏见的延续与缓解,构建了一个分析框架,通过案例研究揭示了AI如何放大现有人类偏见,同时也承认AI在减少偏见方面的作用,呼吁技术、性别研究和公共政策学者间的进一步合作,以探索算法问责制 44。Wach等人(2023)对生成式AI(以ChatGPT为例)的争议和风险进行了批判性分析,将识别出的威胁归类为无监管、质量差、算法偏见、数据侵犯、社会操纵、社会经济不平等加剧和AI技术压力等七大类,强调了AI市场监管、伦理实践和偏见缓解技术的重要性 45。

2.5.3 生成式AI伦理规范

生成式AI的飞速发展带来了前所未有的伦理挑战,促使学术界和政策制定者积极探讨其伦理规范。Liebrenz等人(2023)探讨了ChatGPT对医学出版伦理的实质性影响,提出了版权、归属、抄袭和作者身份等方面的伦理考量,并强调了制定AI作者指南的紧迫性 46。他们指出,AI生成的“看似合理但错误或无意义的答案”可能导致学术错误信息传播的社会危害。

Dwivedi等人(2023)从多学科角度探讨了生成式会话AI(如ChatGPT)带来的机遇、挑战和影响,包括其提升生产力的潜力、对实践的颠覆、对隐私和安全的威胁以及偏见的后果,并呼吁在知识、透明度和伦理、组织和社会数字化转型以及教学、学习和学术研究等三个主题领域进行深入研究,以解决AI偏见、评估准确性以及伦理和法律问题 2。Aljuaid(2023)则系统回顾了AI工具对高等教育学术写作教学的影响,指出AI工具可以辅助写作,但无法替代教授批判性思维、研究和伦理的传统课程,强调了在保持教学质量和学术诚信标准的前提下,整合AI支持的平衡方法 47。

在具体规范方面,Abdulai和Hung(2023)深入分析了ChatGPT可能对护理教育、研究和实践中的伦理价值观产生的影响,特别关注了隐私、数据保密性、缺乏人类情感(如同理心和同情心)、决策过程中的智慧缺失以及学术诚信等问题,并提出了确保负责任和合乎道德使用AI工具的策略 10。Tuzov和Lin(2024)通过比较中德AI治理,探讨了平衡技术与伦理的两条路径,为国际AI伦理规范的制定提供了不同视角的经验 48。此外,Mensah等人(2024)研究了加纳的《药学法(1994)》在AI药房系统质量控制和过失责任方面的适用性,揭示了现有法规在应对AI挑战方面的不足,并提出了改革建议,强调了早期政府改革以适应AI发展现实的重要性 49。

总而言之,现有研究已初步描绘了公共部门AI治理的图景,并识别了算法偏见带来的法律伦理挑战。然而,针对生成式AI在公共服务场景中的具体算法偏见表现、系统性的法律规制路径以及多主体协同治理机制的深入实证研究仍相对缺乏,特别是在如何将理论框架与可落地、可操作的法律科技辅助工具相结合方面存在显著研究缺口。本研究将致力于弥合这些空白。

2.6 研究创新点

本研究的核心创新点在于:

  1. 提出适配公共服务场景的“算法全生命周期偏见防控+法律追责”协同机制。 传统算法偏见治理往往侧重于事后补救或单一环节的控制。本研究将突破这一局限,构建一个涵盖生成式AI算法从设计、数据准备, 模型训练、部署运行到持续监测和迭代的全生命周期偏见防控体系。在此基础上,创新性地融入法律追责机制,形成事前预防、事中干预、事后问责的闭环管理。这意味着不仅要在技术层面设计偏见检测和缓解工具,更要通过法律工具(如强制性偏见影响评估、透明度报告要求、可解释性标准)确保这些措施的强制性与执行力,并在偏见导致损害时,提供明确的法律救济和责任追究路径。例如,在医疗领域,生成式AI辅助诊断系统需在全生命周期内接受严格的偏见评估,确保其诊断结果不会因患者的种族、性别或社会经济地位而产生歧视,一旦出现偏见导致误诊,则明确其法律责任主体及追责方式。AI在医疗领域的应用潜力和伦理考量日益受到关注,需要对AI系统的公平性、透明性和问责制进行深入研究以确保其在医疗中的安全有效应用50。

  2. 明确不同参与主体的责任划分边界与协同模式。 在多中心治理理论指导下,本研究将超越泛泛而谈的多主体参与,具体界定在公共服务场景下,政府(作为监管者和AI使用者)、AI技术提供商(开发者)、数据提供方(如公共数据机构)、第三方审计机构、以及公民社会组织等各方在算法偏见防控与法律追责中的具体角色、权利、义务和责任边界。例如:

    • 政府:不仅承担AI应用的最终决策和管理责任,还应设立专门的算法伦理审查委员会,并可能对算法偏见造成的损害承担部分行政赔偿责任。
    • AI技术提供商:需在合同中明确偏见预防的义务,提供可验证的偏见评估报告,并可能因算法设计缺陷或未能充分披露风险而承担产品责任或违约责任。
    • 数据提供方:对所提供数据的质量和代表性负责,避免引入原始数据偏见,并可能因数据不当或存在偏见而承担数据合规责任。
    • 第三方审计机构:提供独立的算法偏见审计服务,其审计报告可作为法律追责的证据,并对其审计结果的公正性和专业性承担责任。
    • 公民社会组织:作为公众利益的代表,参与算法监督、提出投诉和提供法律援助,推动公众对偏见问题的认知和参与。

    这种明确的责任划分与协同模式,旨在克服当前生成式AI领域“责任真空”和“治理碎片化”的挑战,为构建一个公平、公正、透明的公共服务AI生态系统提供坚实的制度基础。通过法律手段明确各方权责,将有效激励各方主动履行偏见防控义务,从而提升整体治理效能,减少因算法偏见引发的社会风险和法律纠纷。当前,AI技术在公共管理中的应用日益增多,但同时也伴随着公平性、透明度和问责制等伦理问题,需要加强对AI采用和推广的背景和过程的理解51。此外,AI的广泛应用也带来了一系列挑战和机遇,包括潜在的偏见、误用和错误信息,这些都需要通过有效的政策和监管框架来解决2。

3. 面向生成式AI知识产权纠纷的法律科技辅助裁判工具研究

3.1 核心研究问题

生成式AI技术的飞速发展及其在内容创作领域的广泛应用,在极大丰富人类精神产品供给的同时,也以前所未有的速度和复杂性冲击着现有的知识产权法律体系,尤其是在知识产权归属认定和侵权判定方面带来了严峻挑战。传统的知识产权法以人类创作为核心,强调独创性和思想表达的保护,但生成式AI创作过程的自动化、半自动化特性以及其对海量现有作品的学习模仿机制,使得“谁是作者”、“何为独创性”、“是否存在实质性相似”等核心问题变得模糊不清。

当前司法实践在处理生成式AI相关的知识产权纠纷时,面临多重困境。例如,对于AI生成内容的著作权归属,各国法律和司法判例尚无统一的明确规定,是归属于AI开发者、使用者,还是不具备著作权主体资格而直接进入公共领域,争议巨大5253545556。在侵权判定上,生成式AI在学习阶段是否构成著作权侵权,以及其生成内容与现有作品的相似度达到何种程度构成侵权,尤其是在“转换性使用”(transformative use)的语境下,如何平衡创新与保护,缺乏清晰的判断标准和可操作的量化工具。这导致司法审判周期长、成本高,且判决结果存在不确定性,严重影响了知识产权的有效保护和AI技术的健康发展。

本研究的核心问题正是要解决这些司法实践中的难点问题,通过法律科技的手段,开发一套可直接嵌入司法办案体系的辅助裁判工具架构。具体而言,本研究将聚焦以下几个关键问题:

  1. 如何界定生成式AI产出内容的知识产权归属?
    • 此问题旨在深入探讨在不同生成式AI应用场景下(例如,完全自主生成、人机协作生成等),作品的独创性贡献如何认定,并在此基础上,探索符合现有知识产权法律精神且具有前瞻性的著作权主体认定规则。研究将考虑用户指令(prompt)的独创性、AI模型设计的独创性、以及AI训练数据的合法性与贡献度等因素,以期提出一套多维度、可操作的归属判定模型。
  2. 如何构建生成式AI侵权判定的量化评估模型与法律适用规则?
    • 本问题将侧重于侵权判定的技术与法律融合。在技术层面,研究将探索先进的文本、图像、音视频内容相似度算法,开发能够客观量化AI生成内容与现有作品之间相似程度的工具,并能够识别出“表达”而非“思想”的实质性相似。在法律层面,研究将结合法教义学,分析现有侵权判断标准(如“接触+实质性相似”原则、避风港原则、合理使用/公平交易原则)在生成式AI语境下的适用性与挑战,并根据量化评估结果,提出一套适应AI创作特点的侵权判定流程和法律适用规则。
  3. 如何设计一套可直接嵌入司法办案流程的法律科技辅助裁判工具架构?
    • 此问题是本研究的实践落脚点。研究目标是设计一个集知识产权归属智能分析、侵权风险评估、类案推送、法律法规检索、裁判文书智能生成等功能于一体的辅助裁判工具原型。该工具不仅要提高审判效率,更要确保裁判结果的公平性、一致性和权威性。架构设计将充分考虑司法人员的实际操作需求、数据安全与隐私保护,以及与现有司法信息系统的兼容性。研究还将特别关注如何将AI的可解释性(Explainable AI, XAI)技术融入工具,使得工具的分析结果和判定逻辑能够被司法人员理解和采纳。

通过对上述核心问题的深入研究和解决,本选题旨在为生成式AI时代的知识产权司法实践提供理论支持、技术工具和解决方案,有效应对技术创新带来的法律挑战,促进数字经济的健康发展。

3.2 理论框架

本研究将融合知识产权法基本原理(Fundamental Principles of Intellectual Property Law)、技术中立原则(Principle of Technological Neutrality)和计算法学理论(Computational Law Theory)来构建分析框架,以全面应对生成式AI知识产权纠纷带来的复杂挑战,并指导法律科技辅助裁判工具的开发。

3.2.1 知识产权法基本原理

知识产权法基本原理构成了本研究的基石,特别是在著作权领域,主要关注独创性(Originality)、作者资格(Authorship)和权利归属(Ownership)等核心概念。生成式AI的介入,直接冲击了这些传统原理的适用性525355。

  • 独创性原则: 传统著作权法要求作品具有“独创性”,即由作者独立创作并表现出最低程度的创造性。然而,生成式AI通过学习海量现有数据来生成内容,其产出是否具备“独创性”以及独创性来源于何处(是AI模型本身,还是用户输入的提示词Prompt,或是开发者对模型的训练),是需要深入分析的关键。本研究将基于独创性原则,探讨如何在AI生成内容的语境下重新理解和界定独创性的标准,以及如何通过技术手段(如溯源分析)评估AI生成内容中可能包含的人类创作元素。
  • 作者资格与权利归属: 著作权通常赋予作品的创作者。但AI作为非人类主体,是否能成为法律意义上的“作者”,以及其生成作品的权利应归属于模型开发者、用户,还是不应享有著作权直接进入公共领域,是亟待解决的问题5456。本研究将审视不同法域(如欧盟、美国、中国)在AI生成作品作者资格认定上的异同54,并探索在法律科技辅助下,如何根据AI生成过程中的人类干预程度、投入成本、风险承担等因素,合理分配权利归属,从而激励创新并保障公平。
  • 思想与表达二分法: 著作权仅保护作品的“表达”,而不保护其“思想”。在生成式AI侵权判定中,区分AI生成内容与现有作品之间是思想的相似还是表达的相似至关重要。本研究将利用这一原理指导侵权判定模型的构建,确保辅助裁判工具能够识别出构成侵权的“实质性相似的表达”,而非仅仅是相似的“思想”或“概念”。

3.2.2 技术中立原则(Principle of Technological Neutrality)

技术中立原则主张法律规制应着眼于行为的实质和效果,而非区分其所采用的技术手段,即“同样的法律适用于同样的行为,无论其通过何种技术完成”。将此原则应用于生成式AI知识产权纠纷,意味着不能因为内容是由AI生成就给予特殊豁免或额外限制,而应在现有法律框架下,考察AI创作行为是否符合法律规定的侵权构成要件。

本研究将运用技术中立原则来:

  • 平衡创新与保护: 避免因技术发展而过度限制AI的合理使用,同时也要防止AI成为大规模侵权的工具。辅助裁判工具的设计应体现对AI技术的理解和包容,但绝不纵容侵权。
  • 统一法律适用: 确保AI生成内容在知识产权法框架下的评判标准与人类创作内容保持一致,避免出现“双重标准”,从而维护法律的稳定性和权威性。例如,在判断AI学习阶段是否构成版权侵权时,需要考量其是否符合合理使用/公平交易的原则,而不是仅仅因为其是AI的行为就作出不同判断。
  • 指导技术工具开发: 辅助裁判工具应致力于将技术分析结果“翻译”成法律语言,使其能够与现有知识产权法中的概念和标准对接,帮助法官和律师基于相同法律原则进行判断。

3.2.3 计算法学理论(Computational Law Theory)

计算法学理论主张将法律规则和原则转化为可计算的形式,通过计算机程序来辅助法律推理和决策。它将法律视为一种可计算的逻辑系统,旨在提高法律适用的一致性、效率和可预测性。在生成式AI知识产权纠纷的背景下,计算法学为开发辅助裁判工具提供了强大的方法论支持。

本研究将基于计算法学理论:

  • 法律规则形式化: 将知识产权法中关于独创性、作者资格、侵权构成要件、合理使用等复杂规则进行结构化、逻辑化处理,转化为计算机可识别和处理的算法或知识图谱。
  • 智能辅助决策: 设计推理引擎,通过分析输入证据(如AI生成内容、原作品、用户Prompt等),结合形式化的法律规则,智能地提供关于知识产权归属、侵权可能性的初步判断和法律分析建议。
  • 自动化类案推送: 运用自然语言处理(NLP)技术对海量裁判文书进行分析,提取关键法律要素和事实特征,实现与当前案件高度相关的类案智能推送,为司法人员提供参考。
  • 可解释性AI融入: 强调辅助裁判工具的决策过程应具备可解释性,即使底层是复杂的算法,其输出结果也应能清晰地展示推理链条和法律依据,增强司法人员的信任和采纳度。

通过整合知识产权法基本原理、技术中立原则和计算法学理论,本研究将能够为构建面向生成式AI知识产权纠纷的法律科技辅助裁判工具提供坚实的理论基础,确保所开发的工具既符合法律的实质正义,又具备技术上的先进性和实用性。

3.3 数据来源

为开发面向生成式AI知识产权纠纷的法律科技辅助裁判工具,并确保其具有充分的理论依据和实践指导价值,本研究将收集和利用多类型、多模态的数据。这些数据不仅是构建模型和验证算法的基础,也为深入理解司法实践中的痛点和需求提供了实证支撑。

具体数据来源包括:

  1. 已公开的生成式AI知识产权纠纷裁判文书:

    • 本研究将从中国裁判文书网、各高级人民法院的公开判例库、以及国际上(如美国、欧盟等)已公开的AI相关知识产权诉讼案例中,收集涉及生成式AI内容创作、著作权归属、侵权判定等争议的裁判文书。这些文书将作为重要的实证数据,通过自然语言处理(NLP)技术进行结构化提取和分析。关注的要素包括:案由、原被告主体、争议焦点、AI生成内容的类型(文本、图像、音频、代码等)、创作过程描述(如prompt输入方式)、侵权比对方法、法院的认定理由、判决结果及法律依据。对这些裁判文书的分析有助于识别当前司法实践中对AI生成内容独创性的判断标准、侵权认定的关键证据以及法律适用的主要困境。此外,本研究还将关注生成式AI在法律文档中可能引发的抄袭和版权侵权问题,以确保数据处理的严谨性 57。
  2. AI生成内容特征数据集:

    • 为支撑侵权判定模型和归属分析模型的开发,需要构建一个包含大量AI生成内容及其对应原作品(如有)、Prompt信息、生成参数等元数据的数据集。
    • AI生成文本数据集: 收集通过不同生成式AI模型(如GPT系列、文心一言等)生成的新闻报道、小说片段、法律文书草稿、诗歌等文本内容,并附带生成时使用的Prompt指令、模型版本等信息。同时,收集与这些生成文本在主题、风格、表达上相似的人类创作文本作为比对参照。
    • AI生成图像数据集: 收集通过Midjourney、Stable Diffusion、DALL-E等图像生成模型生成的各类图像作品,包括其对应的文本Prompt、风格参数、模型版本以及可能的参考图像源。同时,收集对应的人类创作图像作品。例如,“ArtConstellation”数据集包含了6,000张WikiArt人类艺术作品和3,200张AI生成艺术作品的注释,通过深度卷积神经网络(CNN)训练进行风格分类,发现AI生成作品与现代艺术概念(1800-2000年)相符,并且在某些主题(如风景、几何抽象)上与人类艺术相似,但在变形、扭曲图形方面表现出独特的超出分布的特征 58。这类数据集对于研究AI生成内容的独创性与相似度评估至关重要。
    • AI生成代码/音乐数据集: 收集通过GitHub Copilot、Amper Music等工具生成的代码片段或音乐作品,并包含相应的输入需求、风格要求及模型参数。
    • 这些数据集的构建将有助于我们训练文本相似度算法、图像特征匹配算法等,以客观量化AI生成内容与现有作品的相似度,从而为侵权判定提供技术依据。
  3. 知识产权相关法律法规文本:

    • 涵盖中国现行的《著作权法》、《专利法》、《商标法》、《反不正当竞争法》及其配套法规、司法解释。同时,关注国际条约(如《伯尔尼公约》)、以及美国、欧盟等主要法域的著作权法案及相关判例法。这些文本是构建法律科技辅助裁判工具中法律知识库的核心,通过对法律条文的结构化处理和知识图谱构建,实现法律规则的自动化检索和智能匹配。对AI生成作品的版权保护,需要深入研究《伯尔尼公约》、欧盟版权法和各国法律等现有法律框架 52。研究还将探索原创性、创造性和法律原则(如思想表达二分法)在文本提示和AI输出背景下的应用,借鉴拼贴画保护的哲学基础 53。
  4. 司法人员调研访谈数据:

    • 通过对知识产权法官、检察官、律师、司法辅助人员等进行深度访谈和问卷调查,收集他们在使用生成式AI辅助创作或审理AI相关案件时遇到的实际问题、对辅助裁判工具的需求、对AI生成内容归属和侵权认定的困惑、以及对现有法律科技工具的评价。
    • 访谈内容将包括:当前处理AI知识产权案件的流程、主要挑战(如技术理解、证据收集、事实认定)、对法律科技辅助工具的期望功能(如类案推送、智能比对、法律条文匹配)、对AI技术可信度和可解释性的要求等。这些定性数据对于确保辅助裁判工具的实用性、易用性和符合司法实践需求至关重要。

通过整合上述数据来源,本研究旨在建立一个全面、多维的数据基础,为面向生成式AI知识产权纠纷的法律科技辅助裁判工具的理论研究、模型开发和系统验证提供强有力的数据支撑。

3.4 方法设计

本研究将采用多阶段、融合技术与法学的方法设计,旨在开发一个可落地的法律科技辅助裁判工具原型,该工具能够有效处理生成式AI引发的知识产权纠纷。具体方法包括文本相似度算法训练侵权识别模型、结合法教义学分析权属判定规则,以及原型工具测试优化,以实现法律规则与技术算法的双向适配。

3.4.1 文本相似度算法训练侵权识别模型

目的: 开发一套能够客观量化AI生成内容与现有作品之间相似程度的算法模型,为侵权判定提供技术支撑。

具体步骤:

  1. 数据预处理与特征工程:

    • 清洗与规范化: 对收集到的AI生成文本数据集和人类创作文本数据集(包括裁判文书、原创作品等)进行文本清洗,去除噪声、标点符号、停用词等。对文本进行标准化处理,如统一编码格式、进行词形还原或词干提取。
    • 特征提取:
      • 传统文本特征: 采用TF-IDF(词频-逆文档频率)等方法提取关键词权重,反映词语在文本中的重要性。还可以考虑词长、词性、出现位置和主题相关性等多种词语特征,并调整TF-IDF权重,以提高文本指纹的生成质量 59。
      • 语义特征: 利用预训练的深度学习模型(如BERT、RoBERTa等)生成文本的词嵌入(Word Embeddings)或句嵌入(Sentence Embeddings),捕捉词语和句子之间的语义关系。这有助于识别语义上的相似性,而非仅仅是表层文字的匹配。
      • 结构特征: 针对法律文书或特定文体,提取段落结构、句法结构等特征,例如句子的复杂度、段落的逻辑关系等。
  2. 相似度计算模型构建:

    • SimHash改进: 在传统SimHash算法的基础上进行改进。传统SimHash仅考虑词语的单一特征,而本研究将结合多种词语特征(如词长、词性、出现位置、主题相关性),调整TF-IDF权重来生成文本指纹,从而更准确地反映文本的语义信息。通过选择权重值排名前50%的关键词生成文本指纹,再通过计算汉明距离(Hamming distance)来评估文本相似度,实验证明此方法能显著提高F1均值和整体性能 59。
    • 深度学习相似度模型: 构建基于孪生网络(Siamese Network)或交叉编码器(Cross-Encoder)结构的深度学习模型。该模型将接收两段文本作为输入,通过共享权重或交互层学习文本间的语义匹配关系,并输出一个相似度得分。可采用对比学习(Contrastive Learning)或度量学习(Metric Learning)进行训练,使其能够区分细微的语义差异和潜在的“实质性相似”。
    • 多模态融合: 针对图像、音视频等AI生成内容,利用卷积神经网络(CNN)或Transformer等模型提取视觉、听觉特征,并构建相应的多模态相似度计算模型。例如,使用CNN模型识别商标侵权,通过分析图像的视觉特征(边缘、纹理、颜色、形状),即使是细微的修改也能检测出来 60。
  3. 模型训练与评估:

    • 使用标注好的AI生成内容与现有作品的相似度数据集进行模型训练,数据集应包含明确的“侵权”与“非侵权”标注。
    • 采用准确率(Accuracy)、召回率(Recall)、F1分数(F1-Score)以及AUC(Area Under the Curve)等指标评估模型性能,并进行交叉验证。
    • 重点优化模型在识别“实质性相似”方面的能力,并降低误报率和漏报率,确保其在法律实践中的可靠性。

3.4.2 结合法教义学分析权属判定规则

目的: 将知识产权法基本原理和法教义学分析融入辅助裁判工具,为AI生成内容的著作权归属和侵权判定提供法律依据和解释。

具体步骤:

  1. 法律知识图谱构建:

    • 规则提取与形式化: 对《著作权法》、《专利法》等法律法规文本以及相关司法解释和典型案例进行深度分析,提取出关于“独创性”、“作者资格”、“法人作品”、“职务作品”、“合理使用”、“实质性相似”等核心法律概念及其构成要件。
    • 本体论构建: 建立知识产权领域的本体论,定义各类法律实体(如作品、作者、创作行为、侵权行为)、属性(如独创性程度、人机协作程度)及其相互关系。
    • 逻辑规则编码: 将法律条文和司法判例中蕴含的逻辑规则(如“若满足A条件且不满足B条件,则构成侵权”)编码为可计算的逻辑表达式(如SWRL、Datalog),形成法律推理引擎的知识库。
  2. 权属判定与侵权分析推理引擎开发:

    • 独创性评估模块: 基于法律知识图谱,结合用户输入的AI生成内容描述、Prompt信息、人机交互记录等数据,通过设定的规则和案例推理,辅助判断作品的独创性程度以及人类在创作过程中的贡献比例。
    • 作者识别与归属建议模块: 依据独创性评估结果、用户对AI模型的控制程度、投入资金与风险承担等因素,以及法律关于作者资格和法人作品的规定,推理并给出作品著作权归属的初步建议(如归用户、归开发者、共同共有或不享有著作权)。
    • 侵权判定模块: 结合文本相似度算法的输出结果和法律知识图谱中的侵权构成要件规则,对AI生成内容与现有作品进行比对分析。推理引擎将判断是否满足“接触+实质性相似”原则,并进一步考虑“合理使用”或“技术中立”等抗辩事由的适用性,输出侵权的可能性分析和法律依据。
  3. 结果说理与案例推送:

    • 说理生成: 根据推理引擎的分析结果,自动生成符合法律逻辑和语言习惯的分析报告,解释判定结果的法律依据、事实支撑和推理过程。这有助于提升司法人员对AI辅助决策的信任度和理解。
    • 类案检索与推送: 利用自然语言处理(NLP)技术对已收集的裁判文书进行特征提取和主题建模。当输入新的纠纷案例时,系统能够智能匹配和推送与当前案件法律事实、争议焦点、AI应用场景高度相似的过往判例,为司法人员提供决策参考。

3.4.3 原型工具测试与优化

目的: 验证辅助裁判工具的实用性、准确性和用户友好性,并根据反馈进行迭代优化。

具体步骤:

  1. 功能模块集成: 将文本相似度算法模型、法律知识图谱、推理引擎和结果说理、类案推送模块集成到一个统一的软件平台,形成辅助裁判工具原型。
  2. 模拟环境测试:
    • 专家测试: 邀请知识产权法官、律师等法律专业人士,使用工具原型对一批模拟的AI知识产权纠纷案例进行测试。测试内容包括:AI生成内容的独创性判断、著作权归属建议、侵权可能性分析和类案推送的准确性、合理性。
    • 性能评估: 评估工具的响应速度、处理效率、系统稳定性等技术指标。
  3. 用户反馈收集与分析:
    • 通过问卷调查、深度访谈、可用性测试等方式,收集法律专业人士对工具界面设计、功能操作、结果展示、解释可信度等方面的反馈意见。
    • 重点关注工具在实际司法办案流程中的可嵌入性、对决策效率和准确性的提升作用以及可能存在的不足。
  4. 迭代优化: 根据测试结果和用户反馈,对工具原型进行持续优化。这包括调整算法参数、完善法律知识图谱、改进推理逻辑、优化用户界面和交互体验,确保工具能够更好地满足司法实践需求,并实现法律规则与技术算法的双向适配。

通过上述多维度的方法设计,本研究旨在突破现有法律科技工具在处理生成式AI知识产权纠纷方面的局限,提供一个既有坚实理论支撑又具备高度实用价值的辅助裁判工具解决方案。

3.5 代表性文献

本节将系统梳理生成式AI知识产权争议、计算法学应用以及司法辅助工具开发领域的核心研究成果,为面向生成式AI知识产权纠纷的法律科技辅助裁判工具研究提供坚实的理论和实践基础。

3.5.1 生成式AI知识产权争议

生成式AI,尤其是大语言模型(LLMs)和图像生成模型的发展,对传统的知识产权法,特别是著作权法,构成了前所未有的挑战。核心争议点集中在AI生成内容的著作权归属、AI训练数据的合法性以及AI生成内容可能导致的侵权行为。

在著作权归属方面,一个根本性问题是AI能否成为著作权法意义上的“作者”。传统著作权法要求作品具有人类的创作性投入和独创性,但AI并非人类主体。例如,Jessica Gillotte(2019)在其研究中探讨了AI生成艺术品可能涉及的版权侵权问题,并认为在当前版权法下,工程师可以在不承担侵权责任的情况下使用受版权保护的作品来训练AI程序生成艺术品 61。然而,这也引出了后续作品的版权归属问题。Michael Goodyear(2023)的文章直接探讨了“谁应该为AI版权侵权负责?”这一核心问题,触及了AI生成内容的责任主体认定困境 62。Michael D. Murray(2023)深入分析了生成式AI艺术作品的版权侵权与合理使用问题,他指出当前讨论常跳过侵权分析的实际步骤,直接关注AI是否能生成侵权作品以及是否适用“转换性使用”的合理使用抗辩,并强调需要将焦点从训练数据集的编译者和AI系统设计者转移到最终用户,因为最终用户通过Prompt驱动AI创作并对最终作品负责 63。他详细阐述了版权侵权分析的各个步骤,并修正了对生成式AI流程的普遍误解,这对于本研究在侵权判定模型设计中识别责任主体具有重要参考价值。

在训练数据的使用方面,AI模型对海量现有作品的抓取和学习是否构成侵权是一个争议焦点。许多学者认为,这种训练过程可以被视为一种“合理使用”(Fair Use)或“公平交易”(Fair Dealing),因为它没有直接复制原始作品,而是提取其特征和模式。然而,如果AI生成的内容与受版权保护的原始作品高度相似,则可能构成侵权。Amy Cyphert(2023)的研究关注生成式AI在法律文档中可能引发的抄袭和版权侵权问题,这揭示了生成式AI不仅在艺术创作领域,也在文本生成领域带来了法律风险 57。Eric Sunray(2022)在分析AI音乐生成器输出时,通过“最小使用”(de minimis use)的判例法,解释了AI音乐生成器的输出如何潜在侵犯音乐作品和录音版权所有者的专有复制权,这为其他模态AI生成内容的侵权分析提供了方法论上的启发 64。

此外,还有学者关注如何平衡创新与保护。Krzysztof Wach等人(2023)对生成式AI(以ChatGPT为例)的负面影响进行了批判性分析,其中就包括缺乏监管、质量控制不足、算法偏见和侵犯个人数据等问题,强调了对AI市场进行监管的重要性,以保护知识产权和隐私 45。Philipp Hacker等人(2023)探讨了如何监管ChatGPT等大型生成式AI模型,提出应将监管义务分层,并针对AI价值链中的不同参与者(开发者、部署者、用户)设定责任,这为本研究在责任主体划分和法律规制设计方面提供了宏观指导 65。

3.5.2 计算法学应用

计算法学(Computational Law)旨在利用计算机科学和人工智能技术来辅助法律推理、自动化法律流程和提高法律服务的效率与可及性。它为构建法律科技辅助裁判工具提供了核心理论和技术支撑。

早期研究主要集中于法律知识的表达与推理。Richard Susskind在其著作中多次探讨法律职业的未来,强调技术将如何改变法律服务的提供方式,包括使用AI进行法律研究、合同审查和争议解决等。在法律知识建模方面,计算法学致力于将法律规则转化为计算机可理解和处理的形式,例如通过专家系统、本体论和逻辑编程等。

随着自然语言处理(NLP)和机器学习技术的发展,计算法学的应用场景日益丰富。在法律文本分析方面,AI技术可以用于自动识别法律实体、提取关键信息、进行文本分类和摘要。例如,司法辅助工具可以利用NLP技术对海量裁判文书进行分析,识别案件类型、争议焦点、判决结果和法律依据,从而实现类案的智能推送和法律趋势分析。

在法律推理和决策支持方面,计算法学系统能够根据输入的案件事实,结合预设的法律规则和案例知识,生成初步的法律分析报告或预测判决结果。这尤其对于复杂案件或需要处理大量信息的场景(如知识产权侵权认定)具有显著优势。然而,要实现高水平的法律推理,需要解决AI系统的可解释性问题,确保其决策过程透明且符合法律逻辑。Lorna Christie(2023)的研究探讨了可解释机器学习(Interpretable Machine Learning, IML)的重要性,这对于司法辅助工具来说至关重要,因为法官和律师需要理解AI系统是如何得出结论的,以便做出最终的法律判断和说理 66。

此外,计算法学还涉及到智能合约、区块链等前沿技术在法律领域的应用,这些技术有望在未来进一步提升法律服务的自动化和信任度。计算法学的核心目标是提高法律系统的效率、一致性和可访问性,这与本研究开发辅助裁判工具的目标高度契合。

3.5.3 司法辅助工具开发

司法辅助工具的开发是法律科技领域的一个重要分支,旨在利用信息技术和人工智能为司法工作者提供支持,提高审判效率和司法公正性。

在知识产权领域,司法辅助工具的应用主要体现在以下几个方面:

  • 法律检索与案例分析: 传统的法律检索依赖关键词匹配,而现代司法辅助工具则利用语义检索、案例推荐等AI技术,帮助法官和律师快速找到相关法律条文、司法解释和相似案例。例如,通过对裁判文书的深度挖掘,识别出涉及AI生成内容的判例,并分析其争议焦点和裁判思路。
  • 证据审查与管理: 对于涉及大量数字证据(如源代码、图片、视频、聊天记录等)的知识产权案件,司法辅助工具可以协助证据的收集、筛选、分类和展示。例如,图像识别技术可用于对比AI生成图片与现有作品的相似度。
  • 文书辅助生成: 利用生成式AI技术,辅助生成诉讼文书、法律意见书甚至判决书的初稿,减轻司法人员的文字工作负担,提高效率。
  • 侵权比对与风险评估: 针对音乐、图像、文本等多种形式的作品,开发智能比对工具,能够快速检测出潜在的侵权行为。例如,通过音频指纹识别、图像特征匹配或文本相似度算法,评估作品之间的相似性。

在司法辅助工具开发过程中,研究者和开发者需要关注工具的准确性、可靠性、易用性和安全性。特别是在涉及国家司法主权和公平正义的场景中,工具的决策逻辑必须透明、可解释,且最终的判断权必须保留给人类法官。Andrew D. Selbst等人(2019)提出的关于社会技术系统中公平性和抽象的“陷阱”提醒我们,在设计司法辅助工具时,必须充分考虑其在社会背景下的复杂性,避免技术介入可能带来的不准确或误导性结果,并强调要将社会行动者纳入抽象边界的考量中,而非仅仅关注技术层面 67。

总而言之,现有研究为本课题提供了坚实的理论和技术基础,但也揭示了在生成式AI知识产权治理和司法实践中存在的巨大挑战和研究空白。本研究将在此基础上,致力于将生成式AI知识产权争议的法律原理与计算法学的技术能力相结合,开发出具有创新性和实用性的司法辅助工具,以期有效弥补当前法律科技在应对生成式AI挑战方面的不足。

3.6 研究创新点

本研究的核心创新点在于:

  1. 实现法律规则与技术算法的双向适配,而非单向赋能。 现有法律科技工具在处理知识产权纠纷时,往往是技术单向赋能法律,即将法律规则作为输入,技术进行自动化处理。本研究则强调“双向适配”,即不仅要将法律规则形式化并嵌入技术工具,更要反过来,利用技术算法的客观分析能力,对现有的、模糊的知识产权法律概念(如“实质性相似”、“独创性程度”)提供可量化的、具象化的评估标准。例如,文本相似度算法(如改进的SimHash算法)不仅用于检测相似性,其输出的量化指标和特征差异分析将为法官理解“实质性相似”的边界提供新的视角,从而辅助司法实践在生成式AI语境下,对传统法律概念进行更加精准的解释和适用。这有助于克服传统司法实践中因技术理解不足导致判断困境,提升裁判的科学性和一致性。

  2. 形成可落地的类案推送、侵权判定、结果说理一体化辅助裁判工具方案。 现有司法辅助工具通常功能单一,如仅提供法律检索或文书生成。本研究旨在构建一个集成化的辅助裁判工具,覆盖从案件受理到判决形成的关键环节:

    • 类案推送: 通过深度挖掘已公开的生成式AI知识产权纠纷裁判文书,利用自然语言处理和图谱技术,智能识别案件特征,并精准推送高度相关的法律法规、司法解释以及具有参考价值的类案判例,帮助司法人员快速掌握同类案件的裁判思路和法律适用。
    • 侵权判定: 融合基于多模态特征(文本、图像等)的相似度算法模型(如改进SimHash、孪生网络)与法律知识图谱,对AI生成内容与既存作品进行客观、量化的相似性比对,并结合知识产权法基本原理,对侵权构成要件进行智能分析,给出侵权风险评估和初步判定建议。
    • 结果说理: 在侵权判定和归属分析的基础上,工具能够根据法律知识图谱和推理引擎的逻辑链条,自动生成结构清晰、逻辑严谨、符合法律规范的分析报告,解释判定结果的法律依据、事实支撑和技术分析过程。这不仅能辅助司法人员撰写判决书,也能增强判决结果的说服力和透明度,提升公众对司法公正的信任。
  3. 深度融合法教义学与计算方法,构建面向AI生成内容的著作权归属智能分析模型。 针对生成式AI内容著作权归属这一核心难题,本研究将超越简单的“谁输入谁拥有”或“谁开发谁拥有”的思路。它将把法教义学中关于独创性、作者资格、法人作品、职务作品等复杂概念,结合计算法学的方法,转化为多维度、可量化的评估指标体系。例如,通过分析用户Prompt的复杂程度、创作意图、修改和筛选AI输出的程度、AI模型在生成过程中自主性与创造性的体现(如通过对模型训练数据的分析和模型参数的解读)、以及投入成本与风险承担等要素,构建一个综合性的智能分析模型。该模型将能够根据具体案情,在人机协作的谱系上,给出更精细、更具说服力的著作权归属建议,填补当前法律在AI生成内容归属认定上的空白,为司法实践提供前瞻性指引。

4. 生成式AI内容治理场景的算法偏见审计体系研究

4.1 核心研究问题

当前,生成式AI技术被广泛应用于各类在线平台和内容服务提供商的内容治理环节,旨在应对海量的有害内容,如仇恨言论、虚假信息、以及不适宜图片等 68。然而,过度依赖AI进行内容审核和过滤,在提高效率、减轻人工审核压力的同时,也带来了新的算法偏见问题,严重影响了内容治理的公平性、客观性,并可能侵犯用户的言论自由和信息获取权。Gillespie (2020) 指出,尽管AI似乎是应对社交媒体内容审核规模挑战的完美方案,但即便能有效自动化内容审核,其合理性仍值得商榷 69。Ma和Kou (2021) 的研究揭示了YouTube的算法内容审核如何通过去货币化等惩罚机制,对YouTuber的劳动力条件造成影响,凸显了算法不透明性和不稳定性带来的社会经济后果 70。Zhuk (2023) 在元宇宙背景下分析AI的伦理影响时,也强调了内容审核的挑战,认为需要人机混合方法来平衡创作自由和用户安全 71。

本研究的核心问题正是要针对当前生成式AI内容治理中存在的歧视性审核、偏见性过滤等突出问题,构建一套标准化、可落地的算法偏见第三方审计体系。具体而言,本研究将聚焦以下几个关键问题:

  1. 生成式AI在内容治理中产生算法偏见的具体机制与表现形式是什么?
    • 此问题将深入分析生成式AI内容审核系统(如文本审查、图像识别、语音识别等)如何由于训练数据偏差、模型设计缺陷或特定产品策略,导致对特定群体(如基于种族、性别、宗教信仰、政治观点等)生成或发布的内容进行歧视性识别、过滤或删除 72。例如,系统可能对某些方言、少数族裔文化表达或新兴亚文化群体的内容存在过度敏感或误判,导致“劣币驱逐良币”效应,抑制多元化表达。研究还将考察AI系统中的“过度审查”(over-censorship)问题,即为了确保平台安全而过度限制合法内容的发布,以及“审查不足”(under-censorship)问题,即未能有效识别和处理某些隐蔽的有害信息,从而影响用户体验和平台生态健康。
  2. 现有内容治理规则与算法偏见检测方法在应对生成式AI内容治理挑战方面的有效性与局限性何在?
    • 本问题将评估现有的内容治理政策、平台社区准则以及AI算法偏见检测技术在生成式AI语境下的适用性。研究将分析现有规则是否能够充分覆盖生成式AI带来的新型偏见(如深度伪造内容审查中的偏见),以及当前的偏见检测工具和方法(如传统的数据偏见检测、模型输出评估)是否足以应对其复杂性和动态性。同时,研究还将探讨当前内容治理中“黑箱”决策导致的用户申诉无门、透明度不足、问责困难等问题。Banchio (2023) 讨论了AI驱动内容审核的法律、伦理和实践挑战,突出了透明度、偏见缓解和问责机制的重要性 73。
  3. 如何构建一套标准化、可落地的第三方算法偏见审计指标体系和评估流程,以有效识别和量化生成式AI内容治理中的偏见?
    • 此问题旨在设计一个具体可行的审计框架。研究将聚焦于开发一套多维度、可量化的审计指标体系,涵盖数据输入偏见、模型决策偏见、输出结果偏见、以及用户体验偏见等层面。同时,将设计一套标准化的审计流程,包括审计范围界定、数据收集(如受控实验数据、用户投诉数据)、评估方法(如公平性指标计算、定性分析)、报告生成及建议。Raji等人 (2022) 和Costanza-Chock等人 (2022) 的研究强调了算法审计在促进问责方面的作用,并提出了建立第三方审计生态系统的建议,这为本研究提供了重要参考 7475。
  4. 如何设计第三方审计机构的资质要求、运行规则和法律责任,确保审计结果的独立性、权威性与有效性?
    • 本问题将探讨建立健全第三方审计生态系统的制度保障。研究将关注如何规范第三方审计机构的准入门槛、专业能力要求、审计方法标准、信息披露义务以及对审计结果承担的法律责任。同时,还将探索审计结果如何与监管机构的监督、平台公司的整改以及用户投诉处理机制有效衔接,从而形成一个从发现问题、评估偏见到推动改进的闭环,最终促进生成式AI内容治理的公平性和透明度。例如,对审计师进行评估和认证,并考虑将受AI系统影响的利益相关者纳入审计过程,这些都是确保审计有效性的关键因素 74。

通过回答上述核心问题,本研究旨在为生成式AI内容治理领域提供一个科学、公正、可信赖的第三方审计体系方案,以期在保障内容安全的同时,最大程度地减少算法偏见对用户权益的损害,促进在线言论空间的健康发展。

4.2 理论框架

本研究将整合算法审计理论(Algorithm Auditing Theory)、传播法相关规则(Communication Law Regulations)以及信任治理理论(Trust Governance Theory)来构建分析框架,以全面深入地分析生成式AI内容治理场景中的算法偏见审计体系。这三大理论从不同维度为理解和构建有效的算法偏见审计机制提供了坚实的学理基础。

4.2.1 算法审计理论(Algorithm Auditing Theory)

算法审计理论关注对算法系统,特别是AI系统进行系统性、独立性评估,以识别、测量和评估其行为、性能、公平性、透明度和合规性。其核心目的是验证算法是否按照预期运行,以及是否存在潜在的偏见或歧视,尤其是在“黑箱”算法决策难以直接解释的情况下。算法审计可以包括代码审计、数据审计、模型行为审计和影响审计等多个层面。在生成式AI内容治理场景中,算法审计理论提供了识别和量化算法偏见的方法论和工具。

本研究将利用算法审计理论,重点关注以下方面:

  • 审计范围与类型: 界定生成式AI内容治理中算法偏见的具体审计对象,包括训练数据的偏见、模型架构导致的偏见、内容生成或审核规则中的偏见以及最终输出结果(如内容过滤、推荐结果)的偏见。将涵盖事前评估(如偏见影响评估)、事中监控(如实时性能监测)和事后审计(如对已发生偏见事件的溯源分析)等多种审计类型。
  • 偏见指标量化: 借鉴算法公平性研究中的各类公平性指标(如差异化待遇、差异化影响、群体公平性、个体公平性等),以及特定内容治理场景下的偏见量化方法,开发一套适用于生成式AI内容治理的算法偏见度量体系。例如,通过受控实验测试,量化不同人口群体生成内容被误判、删除或推荐的差异。
  • 审计方法与工具: 探索将受控实验、反事实分析、因果推理、解释性AI(XAI)技术等融入算法审计流程 23。通过XAI技术,尝试打开生成式AI内容治理模型的“黑箱”,理解其决策逻辑,识别潜在的偏见来源。例如,使用局部可解释模型无关解释(LIME)或SHAP值来解释特定内容被审核的理由,从而判断是否存在偏见。

4.2.2 传播法相关规则(Communication Law Regulations)

传播法是规制信息传播活动及其相关社会关系的法律规范的总和,其核心在于平衡言论自由、信息流通与公共秩序、个人权利保护之间的关系。在生成式AI参与内容治理的背景下,传播法提供了评估算法偏见合法性与合理性的法律准绳,特别是言论自由保护、信息公开透明以及平台责任等方面的规则。

本研究将运用传播法相关规则,分析以下问题:

  • 言论自由与算法审查的边界: 评估生成式AI内容治理系统在过滤或删除用户内容时,是否过度限制了公民的言论自由权利。传播法要求内容限制应具备合法性、必要性和比例性。研究将分析现有平台政策和算法审查标准是否符合这些原则,以及算法偏见如何导致对合法言论的不当压制。
  • 信息公开与算法透明度: 传播法通常要求信息传播者对信息来源和传播过程保持透明。在生成式AI内容治理中,这意味着平台应在合理范围内公开其内容审核标准、算法工作原理以及偏见检测和纠正机制。研究将探讨如何通过法律强制力,要求平台提升算法透明度,确保用户能够理解其内容被审核或处理的原因。
  • 平台责任与第三方审计: 传播法及相关部门规章(如《互联网信息服务管理办法》)对网络平台的内容管理责任提出了明确要求。本研究将探讨在生成式AI内容治理中,平台如何履行其算法偏见防控责任,以及第三方审计机构在协助平台履行责任、提升合规性方面的法律地位和作用。这包括平台是否应强制接受第三方审计,以及审计结果如何影响平台的法律责任。

4.2.3 信任治理理论(Trust Governance Theory)

信任治理理论强调在复杂、不确定的社会系统中,信任是维系社会关系、促进协作和实现有效治理的关键因素 76。在生成式AI内容治理场景中,用户、监管机构和社会公众对AI系统的信任程度,直接影响着其接受度、合法性和治理效果。算法偏见的存在会严重侵蚀这种信任。因此,通过信任治理理论,可以探讨如何通过审计机制来重建和维护对生成式AI内容治理系统的信任 77。

本研究将从信任治理理论的视角,着重分析:

  • 信任的构成要素: 识别影响用户对生成式AI内容治理系统信任的核心要素,包括算法的公平性、透明度、可解释性、可靠性以及纠错机制的有效性等 7879。算法偏见的审计正是旨在提升这些信任要素。
  • 审计对信任的影响路径: 独立的第三方算法偏见审计可以通过提供公正的评估、揭示潜在风险并推动改进,从而提升用户、监管机构和公众对生成式AI内容治理系统的信心。审计报告的公开性和可验证性是重建信任的重要途径。
  • 多主体参与构建信任: 探讨如何通过构建政府、企业、独立审计机构、用户代表等多元主体共同参与的治理框架,形成一种多层次、互动式的信任机制。例如,通过用户反馈和申诉机制收集偏见案例,经审计机构独立评估后,督促平台改进,并向公众披露改进情况,从而形成信任的良性循环。

通过整合这三大理论,本研究旨在构建一个全面且富有操作性的分析框架,不仅能深入剖析生成式AI内容治理中算法偏见的成因和表现,还能为设计标准化、可落地的第三方算法偏见审计体系提供理论指导,最终促进生成式AI技术在内容治理领域的健康、负责任发展。

4.3 数据来源

为构建一套标准化、可落地的生成式AI内容治理场景算法偏见第三方审计体系,本研究需要广泛收集和分析多源异构数据。这些数据将为识别偏见机制、量化偏见程度、设计审计指标以及验证审计体系提供实证支撑。

具体的数据来源包括:

  1. 主流生成式AI产品的内容审核结果样本:

    • 本研究将与主流的生成式AI内容服务平台(如提供AI写作、AI绘画、AI视频生成等服务的平台)或使用生成式AI进行内容审核的社交媒体平台合作,获取其内容审核系统的匿名化运行数据和结果样本。这些样本应包含:
      • 原始用户生成/AI生成内容: 文本、图片、音视频等多种模态的内容。
      • AI审核系统的处理结果: 是否通过审核、是否被标记为违规、被删除、被限流等处理方式。
      • AI系统标记的违规类型: 仇恨言论、虚假信息、色情、暴力、歧视等具体分类。
      • 人工复审结果(如有): 平台人工审核员对AI处理结果的复核意见,以及与AI判断的差异。
      • 用户画像信息(匿名化): 用户的地理位置、语言、性别、注册时间等可能与偏见相关的非敏感特征,用于分析不同群体内容被处理的差异性。
    • 通过分析这些样本数据,研究可以发现生成式AI内容审核系统在不同内容类型、不同用户群体上的表现差异,例如,识别出对特定语言、文化表达或特定群体言论的过度敏感或不足,从而揭示算法偏见的具体表现形式和潜在影响机制。
  2. 用户投诉的偏见审核案例:

    • 本研究将从媒体报道、消费者维权平台、法律援助机构、非政府组织(NGO)以及平台自身的公开透明度报告中,收集用户投诉的、涉及生成式AI内容审核偏见的具体案例。这些案例应详细描述:
      • 被审核的内容及其上下文: 明确内容的原貌,被判定违规或被删除的具体环节。
      • 用户的申诉理由: 用户认为内容被误判或遭受不公平对待的原因。
      • 平台的回应和处理结果: 平台是否撤销了审核决定,给出的解释和依据。
      • 受影响用户群体的特征: 如果是群体性投诉,则分析受影响群体的共性特征。
    • 这些案例将作为定性分析的重要依据,通过对具体案例的深入剖析,可以补充定量数据难以捕捉的细致偏见现象,验证算法审计理论在实际问题中的适用性,并为构建审计指标提供真实世界的需求导向。例如,社交媒体内容审核中AI系统对某些社会群体的歧视性审核就可能导致用户投诉,而这些投诉案例将直接揭示偏见的存在和影响80。
  3. 国内外算法审计相关标准文本:

    • 本研究将系统收集和整理国际组织(如联合国教科文组织、OECD)、国家机构(如欧盟AI法案、美国NIST AI风险管理框架)、行业协会以及学术界发布的关于算法审计、AI伦理、数据公平性、AI透明度等方面的标准、指南和最佳实践文件。例如:
      • 欧盟《人工智能法案》中关于高风险AI系统的合规评估和上市后监督要求65。
      • 美国国家标准与技术研究院(NIST)的AI风险管理框架中关于评估和管理AI系统固有风险(包括偏见)的方法。
      • IEEE、ISO等组织发布的AI伦理或AI系统评估标准。
    • 通过对这些标准文本的比较分析,可以借鉴成熟的审计框架和方法,识别通用的审计原则和核心指标,为本研究构建的审计体系提供权威性和普适性基础。Jacqui Ayling和Adriane Chapman (2020) 的研究指出,AI伦理工具在审计和风险评估方面存在差距,强调借鉴其他行业(如技术、环境、隐私、金融)的最佳实践来填补这些空白81。
  4. 审计机构调研数据:

    • 本研究将通过对已从事或有意向从事算法审计的第三方机构(包括专业审计公司、学术研究机构、非营利组织等)进行深度访谈和问卷调查,收集以下信息:
      • 现有审计实践: 审计机构当前采用的审计方法、工具、流程、评估指标。
      • 面临的挑战: 在进行算法审计过程中遇到的技术、法律、数据获取、合作意愿等方面的困难。
      • 能力建设需求: 审计机构在专业人才、技术工具、标准规范等方面的需求。
      • 对标准化体系的期望: 对标准化审计流程、资质要求和运行规则的建议。
    • 这些调研数据将为设计第三方审计机构的资质要求、运行规则和法律责任提供实践依据,确保所提出的审计体系具有可操作性和可持续性。对审计师评估和认证,以及将受AI系统影响的利益相关者纳入审计过程,都是确保审计有效性的关键因素。

通过整合上述多维度数据,本研究将能够全面揭示生成式AI内容治理中算法偏见的复杂性,并在此基础上,设计出兼具理论严谨性与实践可行性的第三方算法偏见审计体系,有效推动生成式AI在内容治理领域的负责任应用。

4.4 方法设计

本研究将采用一种混合研究方法,结合受控实验、层次分析法(Analytic Hierarchy Process, AHP)以及试点验证,旨在全面构建和验证生成式AI内容治理场景下的算法偏见第三方审计体系。

4.4.1 受控实验测试不同场景下的算法偏见表现

目的: 通过设计和执行受控实验,系统地测试生成式AI内容治理系统在不同场景下的算法偏见表现,量化偏见程度,并揭示偏见的具体形式和影响机制。

具体步骤:

  1. 实验设计:

    • 选择测试对象: 选取主流的、具有代表性的生成式AI内容治理系统或其核心模块(如文本审核模型、图像识别模型)作为实验对象。
    • 构建测试数据集: 基于真实世界数据特点,人工精心构建包含特定偏见特征的测试数据集。
      • 偏见维度确定: 依据前期理论分析和用户投诉案例,确定可能导致偏见的维度,例如:
        • 人口统计学偏见: 针对不同性别、种族、年龄、地域的用户生成内容(如使用特定方言、提及特定文化符号)。
        • 政治/意识形态偏见: 针对不同政治立场、敏感话题的内容。
        • 社会经济偏见: 针对不同社会经济地位群体相关内容(如贫困、弱势群体的话题)。
        • 内容类型偏见: 针对不同表达形式(如讽刺、幽默、比喻)或不同主题(如科学普及、艺术评论)的内容。
      • 对照组与实验组构建: 为每个偏见维度设计一系列对照内容(无偏见风险)和实验内容(包含偏见风险特征但本质上合规的内容),确保数据集的多样性和代表性。例如,针对性别偏见,可以生成同等程度但涉及不同性别的冒犯性语句,或者在相同主题下使用不同性别代词的文本。
    • 定义偏见衡量指标: 采用量化指标来衡量算法偏见,例如:
      • 误拒率(False Rejection Rate, FRR): 合规内容被错误地标记为违规的比例。
      • 误纳率(False Acceptance Rate, FAR): 违规内容被错误地标记为合规的比例。
      • 差异化处理率(Disparate Treatment Rate): 不同群体内容被系统处理方式(删除、限流、推荐)的显著差异。
      • 用户满意度下降: 通过模拟用户反馈或实际小规模用户测试,评估偏见对用户体验的影响。
  2. 实验执行:

    • 将构建的测试数据集输入到选定的生成式AI内容治理系统,记录系统的处理结果(如审核通过/拒绝、违规类型标记、置信度分数等)。
    • 对比AI系统的处理结果与人类专家的判断结果(真值标签),计算各项偏见衡量指标。
    • 针对同一系统,通过调整不同的参数或规则(如阈值设置),观察其对偏见表现的影响。
  3. 数据分析与偏见识别:

    • 对实验结果进行统计学分析,识别出在哪些场景、哪些群体上,生成式AI内容治理系统表现出显著的算法偏见。
    • 分析偏见发生的具体原因,例如是由于训练数据中特定群体的代表性不足,还是模型在理解特定文化语境或复杂表达时存在缺陷。
    • 撰写详细的实验报告,量化呈现各类偏见表现,为后续审计指标的构建和审计体系的设计提供实证依据。

4.4.2 层次分析法 (AHP) 构建审计指标体系

目的: 基于受控实验的结果、理论框架和专家意见,利用层次分析法构建一套结构化、多维度、可量化的生成式AI内容治理算法偏见审计指标体系。AHP是一种多准则决策方法,能够将复杂问题分解为多个层次,并通过两两比较确定各因素的相对权重,从而实现量化评估。82

具体步骤:

  1. 构建层次结构模型:

    • 目标层: 生成式AI内容治理算法偏见第三方审计体系的有效性。
    • 准则层: 基于算法审计理论、传播法和信任治理理论,确定审计体系的一级准则,例如:
      • 公平性(Fairness): 算法输出是否对不同群体一视同仁。
      • 透明度与可解释性(Transparency & Explainability): 算法决策过程是否可理解、可追溯。
      • 鲁棒性与安全性(Robustness & Security): 算法对对抗性攻击或数据扰动的抵抗能力。
      • 合规性(Compliance): 算法是否符合现有法律法规和伦理规范。
      • 社会影响与责任(Societal Impact & Accountability): 算法对社会和用户造成的实际影响及其问责机制。
    • 指标层: 在每个准则层下,细化具体的、可量化的审计指标。例如,在“公平性”准则下,可以包括:
      • 数据偏见: 训练数据中各群体代表性、标注一致性。
      • 模型偏见: 误拒率、误纳率、差异化处理率。
      • 结果偏见: 内容推荐/过滤结果的群体差异。
      • 申诉处理效率和公平性。
  2. 构建判断矩阵:

    • 邀请相关领域的专家(如AI伦理专家、法律专家、内容治理专家、数据科学家、社会学研究者等),对层次结构模型中同一层次的指标进行两两比较,评估其相对重要性。比较采用1-9标度法,生成判断矩阵。例如,在公平性与透明度之间,哪个因素在算法偏见审计中更为重要。
  3. 计算权重与一致性检验:

    • 利用数学方法(如特征向量法)计算各层次指标的相对权重。
    • 对判断矩阵进行一致性检验,确保专家判断的逻辑一致性。若一致性比率(CR值)超过阈值(通常为0.1),则需要调整判断矩阵,重新进行比较。
  4. 形成审计指标体系:

    • 根据计算出的权重,形成最终的算法偏见审计指标体系,明确各指标的定义、衡量方法和权重,为第三方审计提供统一的评估标准。

4.4.3 试点验证体系可行性

目的: 将构建的算法偏见审计指标体系和流程应用于真实或准真实的生成式AI内容治理场景,验证其可行性、有效性、操作性,并根据试点反馈进行优化。

具体步骤:

  1. 选择试点对象:

    • 与一个或多个有意愿合作的生成式AI内容服务平台或使用AI进行内容治理的机构合作,选取其某个内容治理模块或整个系统进行试点审计。
  2. 组织审计团队:

    • 组建由AI伦理专家、数据科学家、法律专业人士和审计专业人员构成的独立审计团队,参照构建的审计体系和流程进行审计。
  3. 执行试点审计:

    • 审计准备: 审计团队与试点机构沟通,明确审计范围、获取必要数据(匿名化后的内容审核日志、用户投诉数据等)和系统访问权限。
    • 现场审计: 依据审计指标体系,通过数据分析、模型测试、系统审查、访谈相关人员等方式,评估试点对象的算法偏见表现。
    • 撰写审计报告: 审计团队根据审计结果,撰写详细的审计报告,指出发现的算法偏见、偏见程度、潜在原因,并提出改进建议。报告应包含偏见的量化评估结果、定性分析和理论解释。
  4. 评估与反馈:

    • 有效性评估: 评估审计体系在识别和量化算法偏见方面的准确性和深度。
    • 操作性评估: 评估审计流程的顺畅性、审计工具的易用性、所需资源的可获得性。
    • 影响评估: 评估审计结果对试点机构改进算法偏见、提升用户信任、满足合规性要求的实际效果。
    • 收集多方反馈: 收集试点机构、用户代表、监管机构等各方对审计体系的反馈意见和建议。
  5. 迭代优化:

    • 根据试点验证中发现的问题和各方反馈,对审计指标体系、审计流程、审计工具以及第三方审计机构的资质要求和运行规则进行调整和完善,确保最终形成的审计体系具备高度的可行性和实用性。

通过以上混合研究方法,本研究将能够系统地从实验数据、专家经验和实践验证三个层面,全面构建和优化生成式AI内容治理场景下的算法偏见第三方审计体系,为行业健康发展和监管政策制定提供科学支撑。

4.5 代表性文献

本节将系统梳理算法偏见审计、内容治理规制以及生成式AI伦理评估领域的核心研究成果,为本研究构建生成式AI内容治理场景的算法偏见审计体系提供理论依据和实践借鉴。

4.5.1 算法偏见审计

算法偏见审计是确保人工智能系统公平、透明和负责任运行的关键机制,尤其是在决策对个人和社会产生重要影响的场景中。Inioluwa Deborah Raji等学者在《Science》杂志上发表的文章指出,审计对于识别和减轻算法偏见至关重要,特别是那些在医疗健康领域可能导致种族歧视的算法 83。该研究揭示了一个广泛使用的健康算法存在种族偏见,导致分配给黑人患者的额外护理资源减少了一半以上,原因是算法错误地将医疗成本作为健康需求的代理变量。

Raji及其合作者在另一项研究中,进一步探讨了人工智能治理中第三方审计生态系统的设计问题,强调了外部监督的重要性。他们通过对金融、环境和健康等领域的审计系统进行调研,总结了构建有效外部监督系统的经验教训,认为仅关注算法审计本身不足以实现算法问责制,还需要持续关注制度设计 75。这为本研究在设计第三方审计机构的资质要求和运行规则方面提供了重要启发。

此外,针对特定领域的算法偏见审计也受到了广泛关注。例如,在临床AI领域,Daneshjou等人(2021)发现用于训练皮肤疾病诊断AI算法的数据集存在描述不足和潜在偏见,呼吁在AI临床转化前解决数据透明度、非标准化疾病标签和患者多样性评估不足等问题 84。这表明数据审计在算法偏见审计中的基础性作用。

4.5.2 内容治理规制

生成式AI在内容生成和审核方面的应用,使得内容治理面临前所未有的挑战。当前的规制研究主要集中于如何平衡言论自由、内容安全与算法责任。

Xukang Wang和Ying Cheng Wu(2024)对生成式AI带来的颠覆性影响、AI生成内容的法律风险以及如何在创新与监管之间取得平衡的治理策略进行了全面分析,强调了构建统一法律框架、加强国际合作和设立专门监管机构的必要性 85。他们通过文献综述、法律分析和案例研究,深入探讨了生成式AI的伦理和社会经济影响。

针对具体的内容治理工具——ChatGPT,Krzysztof Wach等人(2023)对其争议和风险进行了批判性分析,识别出七大类威胁,包括缺乏监管、内容质量差、算法偏见、数据侵犯、社会操纵等,并强调了加强市场监管、推动伦理实践和缓解偏见技术的重要性。他们的研究强调了生成式AI内容治理中算法偏见的多维度和复杂性,呼吁多方协作解决这些问题。Bernd Carsten Stahl和Damian Eke(2023)也从伦理角度探索了ChatGPT引发的广泛议题,包括社会正义、个人自主、文化认同和环境影响,呼吁多方利益相关者参与,共同应对挑战 86。

此外,中国在生成式AI内容治理方面也出台了相关政策。Sara Migliorini(2024)研究了中国的生成式AI《暂行办法》的出台背景、内容和意义,这为理解特定国家的内容治理规制提供了重要参考 87。

4.5.3 生成式AI伦理评估

生成式AI的快速发展不仅带来了技术上的革新,也引发了广泛的伦理讨论,尤其是在其应用场景日益广泛的当下。

Stefan Harrer(2023)强调了在医疗健康领域负责任地使用大型语言模型(LLMs)的复杂性。他指出,LLMs在没有人类监督、指导和负责任设计的情况下,可能产生和传播错误或有害内容,但如果能负责任地定位和开发,它们可以成为高效、值得信赖的辅助工具 88。这凸显了对生成式AI进行伦理评估的必要性,以确保其在关键领域的应用是安全和有益的。

在教育领域,Mike Perkins等人(2023)提出了“AI评估量表”(AIAS),旨在为教育评估中生成式AI的使用提供一个实用、全面的框架,以平衡其教学机遇与伦理和学术挑战 89。他们强调了透明度、公平性和教师在选择AI使用水平上的自主权。类似地,Tareq Rasul等人(2023)探讨了ChatGPT在高等教育中的潜在益处与挑战,包括学术诚信、可靠性、偏见和虚假信息等问题,并提出了负责任和道德使用的建议 90。Tom Farrelly和Nick Baker(2023)进一步探讨了生成式AI对高等教育的深远影响,特别关注了对国际学生的影响以及AI模型中的偏见,呼吁提升AI素养和伦理考量 91。Christy Boscardin等人(2023)则从医学教育的角度,分析了ChatGPT等生成式AI工具的潜在影响和机遇,并提出了AI素养框架,强调教育者需要增加对AI的理解,以负责任地整合AI 92。

综合来看,现有研究已初步奠定了算法偏见审计、内容治理规制和生成式AI伦理评估的理论基础,并识别了关键的挑战和方向。然而,针对生成式AI内容治理中算法偏见的第三方审计体系的标准化、可落地研究仍存在空白,尤其是在如何将这些理论框架具体转化为可操作的审计指标、流程和机构运行规则方面,需要更深入的探索。本研究将致力于填补这些空白,提供一个全面、实用的解决方案。

4.6 研究创新点

本研究的核心创新点在于:

  1. 提出覆盖事前评估、事中监测、事后追责的全流程算法偏见审计标准。 现有算法审计研究多侧重于事后评估或单一环节的检测。本研究将创新性地构建一套贯穿生成式AI内容治理系统整个生命周期的算法偏见审计标准。

    • 事前评估(Pre-assessment): 强调在生成式AI系统设计和部署前,进行强制性的“算法偏见影响评估”(Algorithm Bias Impact Assessment, ABIA),明确潜在偏见风险、设计缓解策略和透明度要求。这包括对训练数据偏见的审查、模型架构公平性评估以及预期社会影响分析。
    • 事中监测(In-process Monitoring): 建立持续动态监测机制,利用自动化工具和用户反馈渠道,实时追踪生成式AI内容治理系统的偏见表现。例如,通过受控实验(如A/B测试)和用户投诉分析,定期评估不同群体内容被处理的公平性指标,及时发现并预警偏见漂移(bias drift)现象。
    • 事后追责(Post-hoc Accountability): 明确在偏见导致用户权益受损或违反法律法规时,追究相关主体(平台、AI开发者)责任的机制和流程。这包括审计报告的法律效力、用户申诉和法律救济途径的完善,以及对违规行为的惩戒措施。
      这种全流程审计标准旨在将算法偏见防控从被动的响应转变为主动的预防和持续的优化,形成一个闭环管理的治理范式,确保生成式AI内容治理的公平性和负责任性。
  2. 明确第三方审计机构的资质要求与运行规则。 算法审计的独立性和专业性是其有效性的关键。本研究将深入探讨并提出一套明确的第三方算法审计机构资质要求和运行规则,以确保审计结果的客观、公正和权威。

    • 资质要求: 包括机构的独立性(避免与被审计平台存在利益冲突)、专业能力(具备AI伦理、数据科学、法律、社会学等多学科背景的专业团队)、技术工具(具备先进的偏见检测、溯源分析工具和平台)以及行业经验。研究将借鉴会计审计、环境审计等成熟领域的经验,为算法审计机构设置明确的准入门槛和认证标准。
    • 运行规则: 规范审计流程(如审计委托、审计计划、数据获取与保密、发现报告、改进建议、跟踪审计等)、审计方法(如受控实验、人工复核、专家评估、用户访谈等)、信息披露义务(如审计报告的公开程度、对公众的解释方式)、争议解决机制以及法律责任承担(如因审计失职造成的损害赔偿责任)。
      通过建立一套健全的制度体系,本研究旨在提升第三方算法审计的公信力,使其成为生成式AI内容治理中不可或缺的外部监督力量,有效制衡平台内部审核机制的潜在偏见,并为监管机构提供独立的评估依据。这对于应对生成式AI带来的伦理挑战和监管难题至关重要,特别是需要将AI技术与社会背景结合,以确保其负责任地应用。

5. 中小微企业生成式AI应用合规的法律科技赋能工具研究

5.1 核心研究问题

生成式AI技术的快速普及与商业化应用,正深刻改变着各行各业的生产经营模式。对于资源相对有限的中小微企业(SMEs)而言,生成式AI不仅能显著提升运营效率、辅助决策,甚至在财务管理方面也展现出巨大潜力,例如通过AI生成的财务诊断来增强其管理能力,规避商业和金融风险9394。然而,SMEs在拥抱生成式AI带来的机遇时,也面临着严峻的合规挑战。

当前,中小微企业在使用生成式AI过程中普遍存在合规能力不足和风险防控成本过高的问题。这些挑战主要体现在以下几个方面:

  1. 法律法规和伦理规范理解障碍: 生成式AI涉及数据隐私、知识产权、算法偏见、数据安全等多个复杂法律领域。SMEs往往缺乏专业的法律或合规团队,难以及时、准确地理解和消化不断更新的国内外AI治理政策、行业标准和伦理指南,导致在AI应用中埋下合规隐患。例如,SMEs可能在未经授权的情况下使用受版权保护的数据训练模型,或在产品中集成带有偏见的AI算法而不自知。研究表明,尽管AI技术对企业数字化转型至关重要,但中小企业在AI实施上面临着复杂的挑战,包括对技术、文化和伦理方面的顾虑,以及技能短缺和数据质量等问题9596。

  2. 技术风险识别与评估困难: 生成式AI模型的“黑箱”特性、输出内容的不确定性以及潜在的幻觉(hallucination)问题,使得SMEs难以有效评估其应用可能带来的技术风险。例如,AI生成内容的准确性、真实性、是否包含敏感信息或不当言论,都可能为企业带来声誉损失、法律诉讼甚至监管处罚。这种风险识别能力的欠缺,增加了SMEs在实际应用中的不确定性。

  3. 合规工具和解决方案供给不足: 市场上针对大型企业设计的AI合规解决方案往往成本高昂、复杂度大,不适用于中小微企业的资源和规模。SMEs需要的是轻量化、易于操作、成本可控且能够快速部署的合规工具,以帮助它们在日常运营中进行自我审查、风险预警和合规管理。尽管数据合作社等模式能够提供可访问的技术解决方案,并帮助SMEs实现数字化转型97,但专门为SMEs量身定制的AI合规法律科技工具仍然匮乏,这与SMEs在AI采用上面临的挑战(如资源限制和基础设施不足)形成对照96。

  4. 数据治理与安全挑战: SMEs在AI应用中涉及的数据收集、存储、处理和传输,面临着数据泄露、滥用、跨境传输等风险。合规能力的不足使得企业难以建立健全的数据治理体系,确保个人信息保护和数据安全,进而增加了触犯法规的风险。

本研究的核心问题正是要解决上述中小微企业在使用生成式AI时所面临的合规痛点,致力于设计一款轻量化、可落地、具备普惠性的合规法律科技赋能工具。具体而言,本研究将聚焦以下几个关键问题:

  1. 如何系统性地识别和提炼生成式AI应用中中小微企业面临的核心合规风险点?

    • 此问题将通过深度调研和案例分析,梳理出中小微企业在引入、使用和管理生成式AI过程中,最常遇到、影响最大且最难自行解决的法律合规风险,包括数据合规、知识产权、算法公平、消费者权益保护等方面。例如,中小企业在财务管理中应用生成式AI时,可能遇到的风险,如AI模型输出错误导致的财务风险,以及可能影响其财务管理能力的问题9394。
  2. 如何将复杂的法律法规和伦理规范转化为中小微企业易于理解和操作的合规规则,并实现规则的自动化匹配与风险预警?

    • 此问题旨在解决法律文本的可读性和可操作性障碍。研究将探索自然语言处理(NLP)技术,将法律条文结构化,并结合专家知识,构建适用于中小微企业的合规知识图谱。目标是开发一个能够自动识别企业生成式AI应用场景中的合规风险,并给出明确、简洁的合规指引和风险预警的系统。
  3. 如何设计一套低成本、高效率、用户友好的法律科技工具架构,以赋能中小微企业进行生成式AI应用的合规自查、风险评估和日常管理?

    • 此问题侧重于工具的实用性和可落地性。研究将探索基于模块化、可配置的工具设计理念,使SMEs能够根据自身业务特点和需求,灵活选择合规功能。工具可能包括:AI应用场景合规性评估模板、Prompt指令风险检测、AI生成内容合规性初步审查、数据使用合规指南、以及相关法律法规的智能问答系统。
  4. 如何验证所设计的合规法律科技工具在中小微企业的实际应用场景中的有效性、可用性和经济性?

    • 此问题将通过试点应用和用户反馈,评估工具能否有效降低SMEs的合规成本、提升合规能力,以及是否符合其对轻量化和易用性的期待。例如,对建筑行业SMEs的调查显示,尽管AI能提高效率和协作,但其高昂的实施成本、复杂性和对技能人才的需求是主要障碍9899。因此,工具的经济性和易用性至关重要。

通过回答上述核心问题,本研究旨在填补中小微企业在生成式AI应用合规服务领域的供给缺口,为它们提供切实可行的法律科技解决方案,助力其在数字化转型浪潮中实现健康、可持续发展。

5.2 理论框架

本研究将融合合规管理理论(Compliance Management Theory)、成本收益分析理论(Cost-Benefit Analysis Theory)以及技术赋能理论(Technology Empowerment Theory),构建一个多维度、系统性的分析框架,以指导中小微企业生成式AI应用合规法律科技赋能工具的设计与开发。这一综合框架旨在确保所开发的工具既能有效提升中小微企业的合规水平,又能兼顾其资源限制和经济效益。

5.2.1 合规管理理论(Compliance Management Theory)

合规管理理论关注组织如何建立、实施、监控和改进内部控制系统,以确保其运营活动符合外部法律法规、内部政策以及行业标准。在生成式AI应用日益普及的背景下,SMEs面临着数据隐私、知识产权、算法偏见、数据安全等复杂多变的合规要求。传统的合规管理往往由大型企业投入大量资源构建专业团队来完成,但SMEs不具备此条件。因此,本研究将基于合规管理理论,探索如何将复杂的合规要求模块化、流程化,并融入法律科技工具,使其能够适应SMEs的特点。

本研究将利用合规管理理论,重点关注以下方面:

  • 合规风险识别与评估: 参照合规管理理论中的风险导向原则,构建SMEs在生成式AI应用中常见合规风险的识别模型。这包括对生成式AI生命周期中可能出现的法律(如GDPR、CCPA)、行业标准(如ISO 27001)和伦理风险的分类和评估。工具应能帮助SMEs自动化识别其AI应用场景可能涉及的合规风险点。
  • 合规控制与内部机制: 探索如何将合规要求转化为SMEs可执行的内部控制措施。例如,通过工具引导SMEs建立数据使用协议审查流程、AI生成内容审查机制、用户隐私政策制定等。这旨在帮助SMEs在缺乏专业合规人员的情况下,也能建立起基本的合规管理体系。
  • 合规监测与报告: 借鉴合规管理中的持续监测和报告机制,设计工具中的预警功能和合规状态报告模块。工具应能够对SMEs的AI应用行为进行持续监控,识别潜在违规行为,并生成简洁明了的合规报告,帮助企业管理层了解其合规状况。

5.2.2 成本收益分析理论(Cost-Benefit Analysis Theory)

成本收益分析理论是一种经济学分析工具,用于评估决策或项目的预期成本和收益。对于资源有限的中小微企业而言,任何技术或管理工具的引入,都必须经过严格的成本效益考量。如果合规工具的成本过高,SMEs将缺乏采纳的动力,即使其功能再强大也难以落地。因此,本研究在设计法律科技赋能工具时,将以成本收益分析为指导,力求实现“低成本、高效益”的目标。

本研究将运用成本收益分析理论,着重分析:

  • 合规成本量化: 评估中小微企业在未引入法律科技工具前,进行生成式AI应用合规所付出的直接成本(如聘请律师、购买合规服务)和间接成本(如因不合规导致的罚款、声誉损失、业务中断等)。同时,量化引入法律科技工具的成本(如购买/订阅费用、培训成本)。
  • 合规收益量化: 评估法律科技工具能够为中小微企业带来的收益,包括降低法律风险、避免罚款、提升企业声誉、增强客户信任、优化内部管理效率等。
  • 工具经济性优化: 指导工具设计者在功能模块、部署方式(如SaaS模式)、定价策略等方面进行优化,以最大限度地降低中小微企业的采纳和使用成本,同时最大化其合规效益。例如,通过利用微服务架构和API驱动的互操作性,可以实现与现有企业资源规划(ERP)、治理、风险与合规(GRC)平台以及安全信息和事件管理(SIEM)系统的无缝集成,从而提高运营效率并优化成本 100。

5.2.3 技术赋能理论(Technology Empowerment Theory)

技术赋能理论认为,通过引入适当的技术工具,可以增强个体或组织的能力,使其能够更好地实现目标。在中小微企业生成式AI合规的语境下,法律科技工具的核心作用便是赋能。SMEs缺乏专业的合规知识和技术能力,通过法律科技工具,可以弥补这些不足,提升其自主合规的能力。这种赋能并非替代人工,而是通过技术手段,将复杂的法律知识和繁琐的合规流程变得简单化、自动化,降低SMEs遵守法律法规的门槛。

本研究将从技术赋能理论的视角,着重分析:

  • 赋能路径设计: 探索法律科技工具如何通过提供易于理解的合规指引、自动化风险检测、智能化的法律咨询(如智能问答系统)、合规流程标准化等功能,提升中小微企业员工的合规意识和操作能力。
  • 人机协作模式: 强调法律科技工具应作为SMEs合规人员的辅助,而非完全替代。工具应在提供自动化服务的同时,保留人工干预和决策的空间,确保合规决策的灵活性和符合企业实际情况。
  • 用户体验与易用性: 赋能效果与工具的用户体验和易用性密切相关。工具的设计应充分考虑中小微企业用户的技术背景和使用习惯,采用直观的界面设计、简单的操作流程,降低学习曲线,从而提高工具的采纳率和使用效率。例如,香港高校在AI教育政策方面也强调了AI素养的提升,这与工具的易用性是相辅相成的。

通过整合这三大理论,本研究将能够为中小微企业生成式AI应用合规法律科技赋能工具的研究提供一个全面且实用的框架。合规管理理论确保工具内容的专业性和有效性;成本收益分析理论指导工具设计的经济性和可负担性;技术赋能理论则保证工具的用户友好性和能力提升效果,最终开发出真正能够帮助中小微企业应对生成式AI合规挑战的法律科技解决方案。

5.3 数据来源

为开发和验证面向中小微企业生成式AI应用合规的法律科技赋能工具,本研究将收集和利用多维度、多模态的数据。这些数据不仅是识别中小微企业合规痛点、构建合规知识图谱、设计工具功能模块的基础,也将用于评估工具的有效性和实用性。

具体数据来源包括:

  1. 中小微企业生成式AI应用调研数据:

    • 本研究将通过问卷调查和深度访谈的方式,面向不同行业(如制造业、零售业、服务业、科技创新型企业等)、不同规模(员工人数、营收规模)的中小微企业管理者、IT负责人、法务或合规人员(如有)以及实际使用生成式AI工具的员工进行数据采集。
    • 问卷调查内容: 旨在广泛了解中小微企业当前生成式AI的应用现状(如应用场景、使用频率、依赖程度)、对生成式AI合规风险的认知程度、现有合规管理实践、对合规服务的需求、以及对法律科技工具的期望功能和可承受成本等。
    • 深度访谈内容: 将聚焦于深入剖析企业在生成式AI应用中遇到的具体合规挑战(如数据安全与隐私保护、知识产权争议、算法偏见、内容合规等)、现有解决方案的不足、以及对理想合规工具的细节要求。例如,一些企业可能因技术复杂性或资金有限而难以应对AI伦理问题,或者在实施AI系统时面临技能短缺和基础设施不足等障碍。
    • 通过这些调研数据,本研究将能够精准把握中小微企业的真实需求和痛点,确保所开发的工具具有高度的市场契合度和实用性。
  2. 生成式AI相关合规规则文本:

    • 本研究将系统性地收集、整理和分析国内外与生成式AI应用密切相关的法律法规、部门规章、行业标准、伦理指南和监管政策文件。这包括但不限于:
      • 数据合规类: 如《个人信息保护法》、《数据安全法》、GDPR(通用数据保护条例)、CCPA(加州消费者隐私法案)等,特别关注其中关于AI数据处理、跨境传输、个人信息匿名化/假名化的规定。
      • 知识产权类: 如《著作权法》、《专利法》、以及针对AI生成内容知识产权归属和侵权判定的最新指导意见或判例。
      • 算法治理类: 如国家网信办等部门发布的《生成式人工智能服务管理暂行办法》、《互联网信息服务算法推荐管理规定》、欧盟《人工智能法案》(AI Act)等,关注其中关于算法透明度、可解释性、公平性、安全性、问责制等要求。
      • 特定行业规制: 针对中小微企业可能涉及的特定行业(如金融、医疗、教育)中生成式AI应用的合规要求。
    • 这些文本将通过自然语言处理(NLP)技术进行结构化提取和知识图谱构建,形成法律科技工具的核心合规知识库,为自动化风险识别和合规指引提供坚实基础。对这些法律法规的深入理解对于构建一个能够适应快速变化的监管环境的工具至关重要。
  3. 企业合规风险典型案例:

    • 本研究将收集和分析国内外已公开的、涉及生成式AI应用的各类合规风险案例,包括但不限于:
      • 数据泄露/滥用案例: 因生成式AI处理不当导致用户个人信息泄露或被滥用的事件。
      • 知识产权侵权案例: AI生成内容涉及抄袭、侵犯著作权或商业秘密的案例。
      • 算法偏见案例: AI输出内容或决策中存在歧视性、不公平性而引发争议的案例。
      • 虚假信息/伦理争议案例: AI生成虚假新闻、误导性内容或涉及伦理道德问题的案例。
      • 监管处罚案例: 企业因生成式AI应用不合规而遭受监管机构罚款或处罚的案例。
    • 这些案例将作为重要的实证数据,用于验证法律科技工具的风险识别能力,并作为工具中“风险示例”和“预警提示”的素材,帮助中小微企业直观理解合规风险的后果。
  4. 工具试用反馈数据:

    • 在法律科技赋能工具原型开发完成后,将邀请一批代表性中小微企业作为试点用户进行实际试用。
    • 试用跟踪与记录: 记录用户在使用工具过程中的操作日志、功能使用频率、遇到的问题、系统响应时间等客观数据。
    • 用户反馈收集: 通过问卷调查(关于用户满意度、易用性、功能实用性、对合规能力提升的感知等)和深度访谈(收集具体改进建议)等方式,获取用户对工具的主观评价和反馈。
    • 这些试用反馈数据对于工具的迭代优化至关重要,它将帮助研究团队发现工具的不足之处,调整功能设计和交互逻辑,确保最终的工具产品能够真正满足中小微企业的需求,实现其轻量化、易用性和高效率的目标。例如,对建筑行业SMEs的调查显示,用户友好性、可信度和培训支持是影响AI工具采纳的关键因素。

通过上述多方数据来源的综合运用,本研究将能够确保所开发的法律科技赋能工具既有坚实的理论基础和法律依据,又能充分契合中小微企业的实际需求和操作习惯,最终为中小市场主体提供高效、经济的生成式AI合规解决方案。

5.4 方法设计

本研究旨在为中小微企业生成式AI应用合规设计一款轻量化、可落地的法律科技赋能工具。为实现这一目标,将采用多阶段、迭代式的混合研究方法,主要包括问卷调研梳理企业合规痛点、规则结构化搭建合规风险识别模型,以及原型开发与试点优化。

5.4.1 问卷调研梳理企业合规痛点

目的: 全面、系统地了解中小微企业在生成式AI应用过程中面临的合规挑战、需求、认知水平以及对法律科技工具的期望,为工具的功能设计和优先级排序提供数据支持。

具体步骤:

  1. 问卷设计:

    • 目标群体: 针对中小微企业的管理者、IT负责人、合规或法务人员以及实际使用生成式AI工具的员工。
    • 内容维度:
      • 生成式AI应用现状: 采用的AI工具类型、应用场景(如营销文案生成、客服机器人、内部报告撰写、图像设计等)、使用频率和依赖程度。
      • 合规风险认知: 对数据隐私、知识产权、算法偏见、数据安全、消费者保护等法律法规的了解程度;对AI生成内容真实性、准确性、伦理性的担忧。
      • 现有合规实践: 是否有专门的合规团队/人员;如何进行合规审查;是否遇到过合规问题及如何解决。
      • 合规成本感知: 投入合规工作的资金、时间成本;因不合规可能导致的潜在损失评估。
      • 法律科技需求: 对自动化风险识别、合规指引、合同审查、知识产权保护、隐私政策生成等法律科技工具功能的期待;可接受的工具成本和部署模式(如SaaS)。
      • 对AI伦理的考量: 了解企业对AI伦理(如公平性、透明度、可问责性)的重视程度以及在实践中如何平衡经济效益和伦理原则。
    • 题型设置: 采用单选、多选、李克特量表、开放式问题相结合的方式,既能进行量化分析,也能收集定性洞察。
  2. 抽样与实施:

    • 样本选择: 采用分层随机抽样与目的性抽样相结合的方式,确保样本覆盖不同行业、不同规模、不同AI应用程度的中小微企业。
    • 问卷发放: 通过在线问卷平台、行业协会合作、企业直投等多种渠道发放问卷,确保问卷回收率和有效性。
  3. 数据分析:

    • 描述性统计: 分析中小微企业生成式AI应用现状、合规风险认知、需求偏好等总体特征。
    • 交叉分析: 探究不同类型企业(如行业、规模)在合规痛点和需求上的差异。
    • 因子分析/聚类分析: 识别中小微企业合规痛点的核心构成要素和典型模式。
    • 定性内容分析: 对开放式问题的回答进行归纳总结,挖掘深层次的痛点和需求。
    • 结果呈现: 形成详细的调研报告,明确中小微企业对生成式AI合规法律科技工具的核心需求和优先级。

5.4.2 规则结构化搭建合规风险识别模型

目的: 基于问卷调研结果和现有法律法规,将复杂的合规规则转化为可计算、可自动识别的风险点,构建一个能够为中小微企业提供定制化合规指引和风险预警的智能模型。

具体步骤:

  1. 合规知识库构建:

    • 法律法规提取与解析: 对《个人信息保护法》、《数据安全法》、《生成式人工智能服务管理暂行办法》等相关法律法规(详见2.3数据来源部分)进行深度解析,提取与生成式AI应用相关的合规要求。这些要求将以结构化的方式存储,例如,将法律条文分解为“主体-行为-客体-条件-后果”的逻辑单元。
    • 行业标准与伦理准则整合: 将特定行业(如医疗、金融、教育)的AI应用合规要求和国内外主流的AI伦理准则(如公平性、透明度、可问责性)纳入知识库,作为风险评估的补充维度。
    • 专家知识编码: 邀请法律专家、合规顾问等将实践经验、典型案例中蕴含的隐性合规规则进行显性化和编码化,补充到知识库中。
  2. 知识图谱与本体建模:

    • 构建本体论: 定义生成式AI应用合规领域的核心概念(如“用户数据”、“训练数据”、“Prompt指令”、“AI生成内容”、“知识产权”、“算法偏见”等)、属性及其相互关系,形成领域本体。
    • 建立知识图谱: 将结构化的法律法规、行业标准、伦理准则、典型案例等信息映射到知识图谱中,通过实体、关系、属性的连接,构建一个可供机器理解和推理的合规知识网络。例如,将“AI生成内容”与“著作权归属”、“侵权风险”等实体通过特定关系连接。
  3. 合规风险识别模型开发:

    • 自然语言处理(NLP)技术应用: 利用NLP技术对中小微企业输入的AI应用场景描述、Prompt指令、AI生成内容片段、数据使用协议等文本信息进行语义分析和实体识别。
    • 规则引擎与推理: 基于知识图谱和预设的合规规则,开发一个规则引擎。当企业用户输入其AI应用的相关信息时,规则引擎将自动匹配知识库中的合规要求,并进行逻辑推理,识别出潜在的合规风险点。例如,如果AI应用涉及处理“个人敏感信息”但“未征得用户明确同意”,则模型将触发“数据隐私违规”的风险预警。
    • 风险等级评估: 根据风险影响程度和发生概率,对识别出的合规风险进行等级评估(如高、中、低风险),并提供详细的风险解释。
  4. 风险预警与合规指引生成:

    • 定制化预警: 当识别到合规风险时,模型能够即时向用户发出预警。
    • 简明合规指引: 针对每个风险点,模型能自动生成简洁、易懂的合规指引和操作建议,例如“请确保您已获得用户关于个人敏感信息处理的明确授权”、“检查AI生成内容是否与现有作品存在实质性相似”。
    • 案例参考: 关联知识库中的典型案例,辅助用户理解合规要求和风险后果。

5.4.3 原型开发与试点优化

目的: 将上述理论成果和模型转化为具象化的法律科技工具原型,并通过实际试点验证其在中小微企业的有效性、可用性和经济性,并根据用户反馈持续迭代优化。

具体步骤:

  1. 工具原型开发:

    • 功能模块设计: 基于前两阶段的成果,设计工具原型的核心功能模块,可能包括:
      • AI应用场景合规评估模块: 用户输入AI应用场景描述,系统自动识别合规风险。
      • Prompt指令风险检测模块: 用户输入Prompt,系统检测是否可能生成不合规内容。
      • AI生成内容初步审查模块: 用户上传AI生成文本/图片等,系统进行合规性初步审查(如侵权风险、偏见风险、伦理风险)。
      • 数据使用合规指南: 提供关于数据收集、存储、处理、共享的合规建议。
      • 法律法规/案例智能问答: 基于知识图谱的法律法规查询与问答系统。
      • 合规报告生成与状态展示。
    • 用户界面(UI)/用户体验(UX)设计: 强调简洁性、直观性、易用性,确保非法律专业背景的中小微企业用户也能轻松操作。
    • 技术架构: 采用模块化、可扩展的云原生架构,支持SaaS部署,降低企业部署和维护成本。
  2. 试点企业招募与实施:

    • 招募: 从前期问卷调研中筛选出有意愿、有代表性的中小微企业作为试点用户。
    • 培训与部署: 为试点企业提供工具使用培训,协助其在实际业务流程中部署和使用工具。
  3. 效果评估与反馈收集:

    • 客观数据收集: 记录工具使用日志(使用频率、功能模块使用情况)、风险识别准确率、预警响应情况等。
    • 主观反馈收集: 通过用户满意度问卷、深度访谈、焦点小组等方式,收集试点企业对工具功能、性能、易用性、对合规能力提升的感知、以及工具经济性的评价。特别关注工具在降低合规成本、提高合规效率方面的实际效果。
    • 合规风险事件跟踪: 跟踪试点企业在试用期间是否发生与生成式AI应用相关的合规风险事件,并评估工具的预防作用。
  4. 迭代优化:

    • 根据试点结果和用户反馈,对工具原型进行多轮迭代优化,包括功能增删、模型参数调整、知识库更新、UI/UX改进等。
    • 最终目标是形成一个功能完善、性能稳定、高度可用、且经济有效的法律科技赋能工具产品。

通过上述方法设计,本研究将能够系统地从需求梳理、模型构建到原型验证全流程,开发出一款真正能够赋能中小微企业应对生成式AI应用合规挑战的法律科技工具。

5.5 代表性文献

本节将系统梳理中小微企业合规管理、生成式AI应用风险以及法律科技赋能领域的核心研究成果,为本研究开发面向中小微企业生成式AI应用合规的法律科技赋能工具提供坚实的理论和实践基础。

5.5.1 中小微企业合规管理

中小微企业(SMEs)在全球经济中扮演着至关重要的角色,但其在合规管理方面往往面临独特的挑战,主要源于资源有限、专业知识缺乏以及对复杂法规的理解障碍。传统的合规管理研究多集中于大型企业,而针对SMEs的合规研究则强调其特殊性。

Oluwatosin Yetunde Abdul-Azeez等人(2023)的研究强调了网络安全治理、风险管理和合规(GRC)框架对于金融服务行业SMEs增强数字可及性和包容性的关键作用101。该文通过分析SMEs面临的网络安全威胁(如数据泄露、网络钓鱼)以及监管合规要求(如GDPR、PCI DSS、ISO/IEC 27001),提出了通过健全的GRC策略来提升SMEs防御能力和信誉的综合方法。这为本研究在AI合规领域构建SMEs的GRC框架提供了有益借鉴,尤其是在数据安全和隐私保护方面101。

Alexander A. Hernandez等人(2023)针对菲律宾SMEs使用人工智能实现可持续发展的证据研究显示,尽管AI在SMEs中日益受到关注,但其在可持续发展方面的应用仍处于早期阶段。研究指出,基础设施不足、数据可用性、客户隐私和安全、法律框架不完善、管理支持不足以及缺乏AI采纳策略是限制AI在SMEs中应用进展的突出问题102。这些发现直接指出了SMEs在AI应用中存在的合规痛点和挑战,强调了完善法律框架和提供AI采纳策略的重要性,与本研究旨在解决的问题高度契合102。

此外,Mousa Al-kfairy(2025)在探讨生成式AI在组织环境中战略集成时,特别区分了SMEs和大型组织的不同需求。该研究发现,对于SMEs而言,重点在于成本效益高且可扩展的解决方案,以优化资源受限的运营103。这与本研究“轻量化、可落地”的设计理念不谋而合,强调了SMEs在采纳新兴技术时对成本和资源效率的优先考量103。

总的来说,现有研究揭示了SMEs在数字化转型和AI采纳过程中,合规管理面临的普遍性挑战,并强调了成本效益、可扩展性和特定法律框架的重要性。

5.5.2 生成式AI应用风险

生成式AI,尤其是ChatGPT等工具的广泛应用,在带来巨大商业机遇的同时,也伴随着显著的应用风险,这些风险构成了SMEs合规管理的重要组成部分。

David W. Townsend(2023)指出,生成式AI工具为初创企业和小型企业优化流程、增强客户参与度和驱动增长提供了机会,但同时也伴随着潜在的应用风险104。生成式AI带来的风险,如果没有充分的识别和管理机制,可能导致法律诉讼、声誉损害和经济损失。

Pawan Budhwar等人(2023)的讨论表明,生成式AI的引入,包括ChatGPT及其变体,可能会在就业、利益相关者关系和商业模式等方面产生重大影响,并带来诸如幸福感、偏见、错误信息、语境不敏感、隐私问题、伦理困境和安全等风险,这些风险对于SMEs而言,如果没有充分的识别和管理机制,可能导致法律诉讼、声誉损害和经济损失。例如,AI生成内容的准确性问题可能导致商业决策失误,数据侵犯则可能触犯个人信息保护法规105。

The Anh Han等人(2023)的研究聚焦于为SMEs评估AI系统操作风险的可解释AI工具。他们强调,随着SMEs对AI的采纳激增,确保AI系统的安全性、公平性和操作保障变得至关重要。研究指出,SMEs由于资源有限,在道德和安全部署AI系统方面面临独特挑战,并提出了一个评估工具的概念,用于评估AI驱动系统的鲁棒性、潜在偏见、软件和硬件漏洞,以及伦理和法律合规性106。

这些文献表明,生成式AI的应用风险是多方面的,且对SMEs构成了特殊挑战。本研究需要将这些风险具体化、可量化,并转化为法律科技工具中的风险识别和预警功能。

5.5.3 法律科技赋能

法律科技(LegalTech)旨在利用技术改进法律服务的提供和法律行业的运作效率。对于SMEs而言,法律科技工具能够有效降低合规成本,弥补其在法律专业知识和资源方面的不足。

Richard Susskind在其关于法律行业未来的著作中反复强调,技术,尤其是人工智能和自动化,将重塑法律服务的提供方式。虽然他的研究不专门针对SMEs,但其核心思想——通过技术提高法律服务的可及性和效率——对于SMEs合规管理具有普适性。法律科技工具可以通过自动化法律研究、合同审查、合规检查等功能,显著减轻SMEs的合规负担。

Junaid Sattar Butt(2024)对全球首部欧盟《人工智能法案》(AI Act)的分析指出,该法案旨在建立AI系统的全面指南和保障措施,并探索其如何解决AI系统中的“不优雅偏见”和可解释性要求107。虽然《人工智能法案》主要针对高风险AI系统,但其确立的AI治理原则和合规要求,对所有AI使用者,包括SMEs,都具有指导意义。法律科技工具可以帮助SMEs理解并遵循这些复杂的法规,将抽象的法律要求转化为具体的合规操作107108。

IEEE Engineering Management Review的Mousa Al-kfairy(2025)研究再次强调,针对SMEs,生成式AI的战略集成需要成本效益高且可扩展的解决方案,这同样适用于法律科技工具的开发。文章建议,通过自动化工具和专业的合规管理,可以有效弥合中小企业在AI应用中面临的挑战,降低其运营风险和合规成本103。

Pawan Budhwar等人(2023)在讨论生成式AI对人力资源管理的影响时,也提到了生成式AI带来的风险,包括偏见、错误信息和隐私问题,并强调需要制定有效的政策和监管框架来应对这些挑战105。法律科技工具正是在这一背景下,通过提供合规指导和风险管理功能,成为中小微企业应对这些挑战的重要手段。

综上所述,现有文献为本研究提供了坚实的理论基础:SMEs在生成式AI应用中面临资源受限、知识不足、成本敏感等多重合规挑战。法律科技工具,尤其是那些成本效益高、可扩展且能提供可解释风险评估的工具,能够有效赋能SMEs,帮助其降低合规成本,提升合规能力。本研究将致力于将这些理论洞察转化为实际可用的法律科技产品。

5.6 研究创新点

本研究的核心创新点在于:

  1. 形成适配中小微企业成本承受能力的轻量化合规工具方案。 现有市场上针对AI合规的解决方案多面向大型企业,其高昂的成本和复杂的部署维护流程,使得资源有限的中小微企业望而却步。本研究突破了这一局限,以“轻量化”为核心设计理念,从以下几个方面创新性地构建合规工具方案:

    • 模块化与按需配置: 工具将采用高度模块化的设计,SMEs可以根据自身业务类型、AI应用场景和实际需求,灵活选择和组合合规功能模块(例如,仅选择数据隐私合规模块,或仅选择知识产权审查模块),避免为不必要的功能付费,从而大幅降低初始投入成本。
    • SaaS化部署与维护: 工具将以软件即服务(SaaS)的形式提供,SMEs无需投入大量资金购置硬件和软件,也无需专业的IT团队进行部署和维护。通过订阅模式,将一次性高投入转化为可控的运营成本,进一步降低S规章SME的采纳门槛。
    • “傻瓜式”操作界面与流程: 充分考虑中小微企业缺乏专业法律和技术人员的现状,工具将提供直观、易懂、图形化的操作界面,将复杂的法律法规和技术概念转化为简单的问答、选择和引导流程,实现“开箱即用”,最大程度降低用户学习成本。
    • 聚焦核心风险与普适性规则: 深入分析中小微企业最常遇到、影响最大的合规风险,工具将优先覆盖数据隐私、知识产权、算法偏见等通用且高风险的领域,并将其中的法律法规提炼为普适性强、易于理解的合规规则,避免引入过度复杂的、SME不必要的合规要求。例如,根据中小企业在AI采用中面临的挑战(如资源限制、基础设施不足和缺乏内部专业知识),提供易于管理和理解的合规解决方案是至关重要的 106109110。
  2. 填补中小市场主体生成式AI合规服务的供给缺口。 随着生成式AI的普及,中小市场主体对合规服务的需求日益增长,但现有法律服务市场和法律科技产品尚未能有效响应。本研究通过提供定制化、经济高效的法律科技工具,精准填补了这一市场空白,具体体现在:

    • 普惠性服务: 通过降低合规工具的使用门槛和成本,使得更多的中小微企业能够负担得起并有效利用法律科技,从而将以往只属于大型企业的专业合规服务普惠化。
    • 需求匹配度高: 深入调研中小微企业的合规痛点和需求,确保所开发工具的功能设计能够精准解决SME的实际问题,例如针对其在数据治理、知识产权保护、算法偏见识别等方面的薄弱环节提供针对性解决方案 106。这将有助于SME应对快速变化的AI监管环境,如欧盟的《AI法案》对SME构成的合规挑战 109110111。
    • 提升自主合规能力: 工具不仅仅是提供风险警示,更重要的是通过嵌入法律知识、提供合规指引、简化合规流程,赋能中小微企业自主提升合规意识和管理能力,从“被动合规”转向“主动合规”,增强其应对AI时代法律风险的韧性。
    • 促进健康发展: 通过有效的合规管理,帮助中小微企业规避因不合规导致的法律诉讼、监管罚款、声誉受损等风险,保障其在生成式AI浪潮中的健康、可持续发展,进而促进整个数字经济生态的公平与繁荣。SME在确保AI系统安全、公平和操作保障方面面临独特的挑战,而本研究提供的工具能够有效应对这些挑战,帮助其建立信任和问责制的基础 106112。
内容由 AI 生成,仅供参考,请仔细甄别

参考文献

1On the use of AI-based tools like ChatGPT to support management researchOpenAlex

Bastian Burger, Dominik K. Kanbach, Sascha Kraus, et al.
Purpose The article discusses the current relevance of artificial intelligence (AI) in research and how AI improves various research methods. This article focuses on the practical case study of systematic literature reviews (SLRs) to provide a guideline for employing AI in the process. Design/methodology/approach Researchers no longer require technical skills to use AI in their research. The recent discussion about using Chat Generative Pre-trained Transformer (GPT), a chatbot by OpenAI, has reached the academic world and fueled heated debates about the future of academic research. Nevertheless, as the saying goes, AI will not replace our job; a human being using AI will. This editorial aims to provide an overview of the current state of using AI in research, highlighting recent trends and developments in the field. Findings The main result is guidelines for the use of AI in the scientific research process. The guidelines were developed for the literature review case but the authors believe the instructions provided can be adjusted to many fields of research, including but not limited to quantitative research, data qualification, research on unstructured data, qualitative data and even on many support functions and repetitive tasks. Originality/value AI already has the potential to make researchers’ work faster, more reliable and more convenient. The authors highlight the advantages and limitations of AI in the current time, which should be present in any research utilizing AI. Advantages include objectivity and repeatability in research processes that currently are subject to human error. The most substantial disadvantages lie in the architecture of current general-purpose models, which understanding is essential for using them in research. The authors will describe the most critical shortcomings without going into technical detail and suggest how to work with the shortcomings daily.

2Opinion Paper: “So what if ChatGPT wrote it?” Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policyOpenAlex

Yogesh K. Dwivedi, Nir Kshetri, Laurie Hughes, et al.
Transformative artificially intelligent tools, such as ChatGPT, designed to generate sophisticated text indistinguishable from that produced by a human, are applicable across a wide range of contexts. The technology presents opportunities as well as, often ethical and legal, challenges, and has the potential for both positive and negative impacts for organisations, society, and individuals. Offering multi-disciplinary insight into some of these, this article brings together 43 contributions from experts in fields such as computer science, marketing, information systems, education, policy, hospitality and tourism, management, publishing, and nursing. The contributors acknowledge ChatGPT’s capabilities to enhance productivity and suggest that it is likely to offer significant gains in the banking, hospitality and tourism, and information technology industries, and enhance business activities, such as management and marketing. Nevertheless, they also consider its limitations, disruptions to practices, threats to privacy and security, and consequences of biases, misuse, and misinformation. However, opinion is split on whether ChatGPT’s use should be restricted or legislated. Drawing on these contributions, the article identifies questions requiring further research across three thematic areas: knowledge, transparency, and ethics; digital transformation of organisations and societies; and teaching, learning, and scholarly research. The avenues for further research include: identifying skills, resources, and capabilities needed to handle generative AI; examining biases of generative AI attributable to training datasets and processes; exploring business and societal contexts best suited for generative AI implementation; determining optimal combinations of human and generative AI for various tasks; identifying ways to assess accuracy of text produced by generative AI; and uncovering the ethical and legal issues in using generative AI across different contexts.

3From fiction to fact: the growing role of generative AI in business and financeOpenAlex

Boyang Chen, Zongxiao Wu, Ruoran Zhao
Generative Artificial Intelligence (AI), such as ChatGPT by OpenAI, has revolutionized the business world, with benefits including improved accessibility, efficiency, and cost reduction. This article reviews recent developments of generative AI in business and finance, summarizes its practical applications, provides examples of the latest generative AI tools, and demonstrates that generative AI can revolutionize data analysis in industry and academia. To test the ability of generative AI to support decision-making in financial markets, we use the ChatGPT to capture corporate sentiments towards environmental policy by inputting text extracted from corporate financial statements. Our results demonstrate that the sentiment scores generated by ChatGPT can predict firms’ risk-management capabilities and stock return performance. This study also highlights the potential challenges and limitations associated with generative AI. Finally, we propose several questions for future research at the intersection of generative AI with business and finance.

4Why and how to embrace AI such as ChatGPT in your academic lifeOpenAlex

Zhicheng Lin
Generative artificial intelligence (AI), including large language models (LLMs), is poised to transform scientific research, enabling researchers to elevate their research productivity. This article presents a how-to guide for employing LLMs in academic settings, focusing on their unique strengths, constraints and implications through the lens of philosophy of science and epistemology. Using ChatGPT as a case study, I identify and elaborate on three attributes contributing to its effectiveness-intelligence, versatility and collaboration-accompanied by tips on crafting effective prompts, practical use cases and a living resource online (https://osf.io/8vpwu/). Next, I evaluate the limitations of generative AI and its implications for ethical use, equality and education. Regarding ethical and responsible use, I argue from technical and epistemic standpoints that there is no need to restrict the scope or nature of AI assistance, provided that its use is transparently disclosed. A pressing challenge, however, lies in detecting fake research, which can be mitigated by embracing open science practices, such as transparent peer review and sharing data, code and materials. Addressing equality, I contend that while generative AI may promote equality for some, it may simultaneously exacerbate disparities for others-an issue with potentially significant yet unclear ramifications as it unfolds. Lastly, I consider the implications for education, advocating for active engagement with LLMs and cultivating students' critical thinking and analytical skills. The how-to guide seeks to empower researchers with the knowledge and resources necessary to effectively harness generative AI while navigating the complex ethical dilemmas intrinsic to its application.

5Google Gemini as a next generation AI educational tool: a review of emerging educational technologyOpenAlex

Muhammad Imran, Norah Almusharraf
Abstract This emerging technology report discusses Google Gemini as a multimodal generative AI tool and presents its revolutionary potential for future educational technology. It introduces Gemini and its features, including versatility in processing data from text, image, audio, and video inputs and generating diverse content types. This study discusses recent empirical studies, technology in practice, and the relationship between Gemini technology and the educational landscape. This report further explores Gemini’s relevance for future educational endeavors and practical applications in emerging technologies. Also, it discusses the significant challenges and ethical considerations that must be addressed to ensure its responsible and effective integration into the educational landscape.

6Should ChatGPT be biased? Challenges and risks of bias in large language modelsOpenAlex

Emilio Ferrara
As generative language models, exemplified by ChatGPT, continue to advance in their capabilities, the spotlight on biases inherent in these models intensifies. This paper delves into the distinctive challenges and risks associated with biases specifically in large-scale language models. We explore the origins of biases, stemming from factors such as training data, model specifications, algorithmic constraints, product design, and policy decisions. Our examination extends to the ethical implications arising from the unintended consequences of biased model outputs. In addition, we analyze the intricacies of mitigating biases, acknowledging the inevitable persistence of some biases, and consider the consequences of deploying these models across diverse applications, including virtual assistants, content generation, and chatbots. Finally, we provide an overview of current approaches for identifying, quantifying, and mitigating biases in language models, underscoring the need for a collaborative, multidisciplinary effort to craft AI systems that embody equity, transparency, and responsibility. This article aims to catalyze a thoughtful discourse within the AI community, prompting researchers and developers to consider the unique role of biases in the domain of generative language models and the ongoing quest for ethical AI.

7Should ChatGPT Be Biased? Challenges and Risks of Bias in Large Language ModelsOpenAlex

Emilio Ferrara

8The ethical implications of using generative chatbots in higher educationOpenAlex

Ryan Williams
Incorporating artificial intelligence (AI) into education, specifically through generative chatbots, can transform teaching and learning for education professionals in both administrative and pedagogical ways. However, the ethical implications of using generative chatbots in education must be carefully considered. Ethical concerns about advanced chatbots have yet to be explored in the education sector. This short article introduces the ethical concerns associated with introducing platforms such as ChatGPT in education. The article outlines how handling sensitive student data by chatbots presents significant privacy challenges, thus requiring adherence to data protection regulations, which may not always be possible. It highlights the risk of algorithmic bias in chatbots, which could perpetuate societal biases, which can be problematic. The article also examines the balance between fostering student autonomy in learning and the potential impact on academic self-efficacy, noting the risk of over-reliance on AI for educational purposes. Plagiarism continues to emerge as a critical ethical concern, with AI-generated content threatening academic integrity. The article advocates for comprehensive measures to address these ethical issues, including clear policies, advanced plagiarism detection techniques, and innovative assessment methods. By addressing these ethical challenges, the article argues that educators, AI developers, policymakers, and students can fully harness the potential of chatbots in education, creating a more inclusive, empowering, and ethically sound educational future.

9Rural development : putting the last firstOpenAlex

Robert Chambers
Misleading findings Useful surveys Total immersion: long and lost?Cost-effectiveness Four ways in and out -Ladejinsky's tourism and the green revolution -Senaratne's windows into regions -Reconnaissance for crop improvement -BRAC and the net Conclusions CHAPTER FOUR Whose knowledge?Knowledge, power and prejudice Outsiders' biases Rural people's knowledge 82 -farming practices -knowledge of the environment -rural people's faculties -rural people's experiments The best of both CHAPTER FIVE Integrated rural poverty Outsiders' views of the poor Clusters of disadvantage The deprivation trap 111 -poverty 112 -physical weakness 112 -isolation 113 -vulnerability 113 -powerlessness 113 Vulnerability and poverty ratchets 114 -social conventions 115 -disasters 116 -physical incapacity 116 -unproductive expenditure

10Will ChatGPT undermine ethical values in nursing education, research, and practice?OpenAlex

Abdul‐Fatawu Abdulai, Lillian Hung
ChatGPT is an artificial intelligence-based chatbot that uses deep learning techniques to generate natural language text (ChatGPT, 2022). While ChatGPT can enhance efficiency, there are growing concerns about its safety and future implications for humanity. On March 29, 2023, industry leaders in artificial intelligence (AI) signed a petition to pause research into AI that is more powerful than the ChatGPT4. According to the petitioners, this action was in response to the uncontrolled “race to develop and deploy ever more powerful digital minds that no one—not even their creators—can understand, predict, or reliably control” (Future of Life Institute, 2023, para. 1). Indeed, AI leaders are concerned about the “significant risk to humanity” that could result from an uncontrolled AI tools that lacks the necessary safety regulations. As nursing researchers in digital health technology and AI, we are concerned not only about the safety implications of ChatGPT, but also about its ability to capture the ethical values, principles, and core tenets that underpin the unique discipline of nursing. Recent editorials and commentaries in some nursing journals have discussed the potential benefits, limitations, and risks of using ChatGPT in nursing education and practice (Archibald & Clark, 2023; Odom-Forren, 2023; Scerri & Morin, 2023). In this commentary, we extend upon these discussions by reflecting on how the use of ChatGPT may undermine the values, principles, and core assumptions that underpin nursing research, education, and practice. We aim to stimulate a dialogue around current and emerging trends in using ChatGPT in nursing. We also want to engage nursing scholars, educators, and practitioners across international communities to discuss nursing assumptions in the context of technology use. Open AI acknowledged that ChatGPT cannot ensure the confidentiality of information entrusted onto it. Based on this, open AI advises against the inclusion of potentially sensitive, confidential, and identifiable information. The limitations of ChatGPT in processing confidential information may undermine the privacy and confidentiality inherent in nursing practice. Nurses are entrusted with clients' personal and health information and they have a legal and ethical responsibility to protect such information at all times. Compromising such information using Open AI tools like ChatGPT might endanger the nurse–client relationship, thus creating distrust in healthcare systems (Scerri & Morin, 2023). Aside from the potential for creating distrust in nurse–patient relationships, we are also concerned that ChatGPT might not capture human emotions, such as empathy and compassion, that are central to the core discipline of nursing. It is important to note that nursing is value-based, emphasizing compassion, empathy, care, respect, and dignity for the patient. On the other hand, Open AI prioritizes the practical usefulness of computing over offering explanatory or theoretical analysis. Without attuning to nursing disciplinary perspectives, ChatGPT may not be able to integrate the unique perspectives of nursing into the ChatGPT algorithms. With ChatGPT becoming common in the healthcare field (Sallam, 2023), it is important to critically reflect upon the potential and problematics of mixing AI data with a humanistic discipline like nursing, especially when they do not fit well. Another point of reflection is the use of ChatGPT in light of the decision-making process in Nursing. ChatGPT uses deep learning algorithms to process sequences of data, learn patterns in the data and generate text based on a query (ChatGPT, 2022). In effect, the output of ChatGPT is based on data available at any point in time. On the other hand, decision-making and clinical judgment in nursing go beyond data and include intuition, tacit knowledge, wisdom, and personal experiences. Even though ChatGPT may be designed to learn, reason, and solve problems that mimic human cognition, it may not be possible to identify or solve nursing problems that demand wisdom, critical thinking, intuition, and personal judgment. To illustrate, the American Nurses Association has included wisdom in the Data-Information-Knowledge framework of computing systems—changing the framework to Data-Information-Knowledge-Wisdom (Nelson, 2020). While AI tools like ChatGPT may be able to represent and generate information relevant to the first three dimensions, wisdom has been a more elusive concept that has not been adequately represented in computerized systems (Matney et al., 2011). This suggests that an overreliance on ChatGPT means a gradual disengagement in critical thinking, wisdom, intuition, tacit knowledge, and ethical judgment, which are considered attributes that define professional nursing practice, education, and research. The complexity and particularity inherent in nursing also constitute ethical values that may not sit well with the use of AI tools like ChatGPT. A recent blog post by a nursing student indicated how ChatGPT can aid nurses' understanding of complex concepts by breaking down complex ideas into simpler pieces (Tran, 2023). It is important to note that nursing adopts a holistic approach to care and clients cannot simply be understood by breaking the whole into parts as this blog post seems to suggest. Nursing considers each individual client as unique and complex. The reductionist approaches of ChatGPT may be overly simplistic and may not be capable of analyzing or understanding clients that present with complex health challenges or tailoring responses to individual cases. Therefore, ChatGPT-generated responses may not fully capture the holistic knowledge needed to address the complex contexts and nuances of individual health challenges. The reductionist approaches of ChatGPT may also conflate the disciplinary boundaries between nursing and medicine. Current AI applications specific to medicine are increasingly applied in disease diagnosis, personalized disease treatment, medical imaging, and genomic sequencing (Amisha et al., 2019). AI may be common in Medicine because AI applications are programmed to function in ways that are similar to the biomedical model of the human body (Kumar et al., 2022). Rather than focusing on individual ailments, nursing disciplinary boundaries transcend the biomedical model of care to focus on the broader picture of the patient, including the psychological, emotional, and social factors that come together to determine one's health. Therefore, Adopting AI tools like ChatGPT in nursing education and practice may be akin to adopting biomedical approaches in situations that demand psycho-sociological approaches to care. Another major concern of ChatGPT in nursing research and education relates to using the tool as an author and how that practice can facilitate academic dishonesty by enabling students to generate written assignments, discussion posts, or research papers without engaging in original thought or research (Choi et al., 2023). ChatGPT was recently listed as a coauthor in published peer-reviewed nursing papers (ChatGPT Generative Pre-Trained Transformer & Zhavoronkov, 2022; O'Connor & ChatGPT, 2023). Such practice raised serious questions about human and AI authors as the latter do not have accountability as a human author (Stokel-Walker, 2023). This practice of generating content using ChatGPT undermines the moral value of integrity in nursing, which demands nurses to be honest with themselves, their patients, and other healthcare teams. Furthermore, nursing decisions involve moral and value judgments and reliance on ChatGPT in nursing education may produce nurses with less adherence to moral values, little or no sense of trustworthiness, and overconfidence in handling complex cases. While industry experts have called for a pause to AI research, the reality is that the development of AI models will continue to advance and these may well find their way into nursing practice, education, and research. Until robust regulatory mechanisms on AI are developed, nursing must adopt some actionable strategies on ChatGPT that will ensure that the use of the AI tool does not distort the ethical values of nursing practice, education, and research. For instance, nursing educators may consider working with students/learners/trainees to establish clear guidelines and ethical standards that govern the use of AI tools like ChatGPT in nursing education, research, and practice. Additionally, promoting a culture of academic integrity and raising awareness of the ethical implications of AI use can help mitigate against academic dishonesty while preserving integrity and honest nursing practice. To counteract disengaged critical thinking and ensure holistic nursing practice that responds to complex situations, educators should prioritize the development of critical thinking and problem-solving skills in nursing curricula. This can be achieved by incorporating activities that encourage students to critically analyze and evaluate information, synthesize knowledge from various sources, and apply their understanding. Nursing academics should also emphasize the importance of contextualized and individualized decision-making in nursing by incorporating case studies, simulations, and real-life scenarios that highlight the importance of patients' unique needs and complex circumstances. Developing clear guidelines and promoting academic integrity, fostering critical thinking skills and independent problem solving, and encouraging contextualized and individualized decision-making can help ensure that AI tools like ChatGPT are used responsibly and ethically, ultimately enhancing the quality of nursing care practice, research, and education. Using ChatGPT in nursing poses ethical challenges that could undermine nursing's core values at the same time that integrating AI in nursing research, education, and practice offers potential advancement. As AI tools continue to grow, it is vital for nursing scholars and educators to engage in critically reflective dialogues and philosophical analyses of the implications, ethics, and potential drawbacks associated with using AI in nursing contexts. By critically questioning the underlying assumptions of AI-driven innovations, we can ensure that the evolving role of AI in nursing aligns with the values, principles, and goals traditionally shaping the profession, promoting holistic, person-centered, and ethical healthcare outcomes. Failing to do so, we risk eroding the foundations that make nursing a unique and essential discipline. No new data were generated for this commentary.

11Adopting and expanding ethical principles for generative artificial intelligence from military to healthcareOpenAlex

David Oniani, Jordan Hilsman, Yifan Peng, et al.
In 2020, the U.S. Department of Defense officially disclosed a set of ethical principles to guide the use of Artificial Intelligence (AI) technologies on future battlefields. Despite stark differences, there are core similarities between the military and medical service. Warriors on battlefields often face life-altering circumstances that require quick decision-making. Medical providers experience similar challenges in a rapidly changing healthcare environment, such as in the emergency department or during surgery treating a life-threatening condition. Generative AI, an emerging technology designed to efficiently generate valuable information, holds great promise. As computing power becomes more accessible and the abundance of health data, such as electronic health records, electrocardiograms, and medical images, increases, it is inevitable that healthcare will be revolutionized by this technology. Recently, generative AI has garnered a lot of attention in the medical research community, leading to debates about its application in the healthcare sector, mainly due to concerns about transparency and related issues. Meanwhile, questions around the potential exacerbation of health disparities due to modeling biases have raised notable ethical concerns regarding the use of this technology in healthcare. However, the ethical principles for generative AI in healthcare have been understudied. As a result, there are no clear solutions to address ethical concerns, and decision-makers often neglect to consider the significance of ethical principles before implementing generative AI in clinical practice. In an attempt to address these issues, we explore ethical principles from the military perspective and propose the "GREAT PLEA" ethical principles, namely Governability, Reliability, Equity, Accountability, Traceability, Privacy, Lawfulness, Empathy, and Eutonomy, for generative AI in healthcare. Furthermore, we introduce a framework for adopting and expanding these ethical principles in a practical way that has been useful in the military and can be applied to healthcare for generative AI, based on contrasting their ethical concerns and risks. Ultimately, we aim to proactively address the ethical dilemmas and challenges posed by the integration of generative AI into healthcare practice.

12The imperative for regulatory oversight of large language models (or generative AI) in healthcareOpenAlex

Bertalan Meskó, Eric J. Topol
The rapid advancements in artificial intelligence (AI) have led to the development of sophisticated large language models (LLMs) such as GPT-4 and Bard. The potential implementation of LLMs in healthcare settings has already garnered considerable attention because of their diverse applications that include facilitating clinical documentation, obtaining insurance pre-authorization, summarizing research papers, or working as a chatbot to answer questions for patients about their specific data and concerns. While offering transformative potential, LLMs warrant a very cautious approach since these models are trained differently from AI-based medical technologies that are regulated already, especially within the critical context of caring for patients. The newest version, GPT-4, that was released in March, 2023, brings the potentials of this technology to support multiple medical tasks; and risks from mishandling results it provides to varying reliability to a new level. Besides being an advanced LLM, it will be able to read texts on images and analyze the context of those images. The regulation of GPT-4 and generative AI in medicine and healthcare without damaging their exciting and transformative potential is a timely and critical challenge to ensure safety, maintain ethical standards, and protect patient privacy. We argue that regulatory oversight should assure medical professionals and patients can use LLMs without causing harm or compromising their data or privacy. This paper summarizes our practical recommendations for what we can expect from regulators to bring this vision to reality.

13Integrating Quantitative and Qualitative Results in Health Science Mixed Methods Research Through Joint DisplaysOpenAlex

Timothy C. Guetterman, Michael D. Fetters, John W. Creswell
PURPOSE: Mixed methods research is becoming an important methodology to investigate complex health-related topics, yet the meaningful integration of qualitative and quantitative data remains elusive and needs further development. A promising innovation to facilitate integration is the use of visual joint displays that bring data together visually to draw out new insights. The purpose of this study was to identify exemplar joint displays by analyzing the various types of joint displays being used in published articles. METHODS: We searched for empirical articles that included joint displays in 3 journals that publish state-of-the-art mixed methods research. We analyzed each of 19 identified joint displays to extract the type of display, mixed methods design, purpose, rationale, qualitative and quantitative data sources, integration approaches, and analytic strategies. Our analysis focused on what each display communicated and its representation of mixed methods analysis. RESULTS: The most prevalent types of joint displays were statistics-by-themes and side-by-side comparisons. Innovative joint displays connected findings to theoretical frameworks or recommendations. Researchers used joint displays for convergent, explanatory sequential, exploratory sequential, and intervention designs. We identified exemplars for each of these designs by analyzing the inferences gained through using the joint display. Exemplars represented mixed methods integration, presented integrated results, and yielded new insights. CONCLUSIONS: Joint displays appear to provide a structure to discuss the integrated analysis and assist both researchers and readers in understanding how mixed methods provides new insights. We encourage researchers to use joint displays to integrate and represent mixed methods analysis and discuss their value.

14Internet of Things (IoT): A vision, architectural elements, and future directionsOpenAlex

Jayavardhana Gubbi, Rajkumar Buyya, Slaven Marusic, et al.

15Internet of Things: A Survey on Enabling Technologies, Protocols, and ApplicationsOpenAlex

Ala Al‐Fuqaha, Mohsen Guizani, Mehdi Mohammadi, et al.
This paper provides an overview of the Internet of Things (IoT) with emphasis on enabling technologies, protocols, and application issues. The IoT is enabled by the latest developments in RFID, smart sensors, communication technologies, and Internet protocols. The basic premise is to have smart sensors collaborate directly without human involvement to deliver a new class of applications. The current revolution in Internet, mobile, and machine-to-machine (M2M) technologies can be seen as the first phase of the IoT. In the coming years, the IoT is expected to bridge diverse technologies to enable new applications by connecting physical objects together in support of intelligent decision making. This paper starts by providing a horizontal overview of the IoT. Then, we give an overview of some technical details that pertain to the IoT enabling technologies, protocols, and applications. Compared to other survey papers in the field, our objective is to provide a more thorough summary of the most relevant protocols and application issues to enable researchers and application developers to get up to speed quickly on how the different protocols fit together to deliver desired functionalities without having to go through RFCs and the standards specifications. We also provide an overview of some of the key IoT challenges presented in the recent literature and provide a summary of related research work. Moreover, we explore the relation between the IoT and other emerging technologies including big data analytics and cloud and fog computing. We also present the need for better horizontal integration among IoT services. Finally, we present detailed service use-cases to illustrate how the different protocols presented in the paper fit together to deliver desired IoT services.

16The PRISMA 2020 statement: an updated guideline for reporting systematic reviewsOpenAlex

Matthew J. Page, Joanne E. McKenzie, Patrick M. Bossuyt, et al.
The Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) statement, published in 2009, was designed to help systematic reviewers transparently report why the review was done, what the authors did, and what they found. Over the past decade, advances in systematic review methodology and terminology have necessitated an update to the guideline. The PRISMA 2020 statement replaces the 2009 statement and includes new reporting guidance that reflects advances in methods to identify, select, appraise, and synthesise studies. The structure and presentation of the items have been modified to facilitate implementation. In this article, we present the PRISMA 2020 27-item checklist, an expanded checklist that details reporting recommendations for each item, the PRISMA 2020 abstract checklist, and the revised flow diagrams for original and updated reviews.

17A new criterion for assessing discriminant validity in variance-based structural equation modelingOpenAlex

Jörg Henseler, Christian M. Ringle, Marko Sarstedt
Discriminant validity assessment has become a generally accepted prerequisite for analyzing relationships between latent variables. For variance-based structural equation modeling, such as partial least squares, the Fornell-Larcker criterion and the examination of cross-loadings are the dominant approaches for evaluating discriminant validity. By means of a simulation study, we show that these approaches do not reliably detect the lack of discriminant validity in common research situations. We therefore propose an alternative approach, based on the multitrait-multimethod matrix, to assess discriminant validity: the heterotrait-monotrait ratio of correlations. We demonstrate its superior performance by means of a Monte Carlo simulation study, in which we compare the new approach to the Fornell-Larcker criterion and the assessment of (partial) cross-loadings. Finally, we provide guidelines on how to handle discriminant validity issues in variance-based structural equation modeling.

18The PRISMA Statement for Reporting Systematic Reviews and Meta-Analyses of Studies That Evaluate Health Care Interventions: Explanation and ElaborationOpenAlex

Alessandro Liberati, Douglas G. Altman, Jennifer Tetzlaff, et al.
Systematic reviews and meta-analyses are essential to summarize evidence relating to efficacy and safety of health care interventions accurately and reliably. The clarity and transparency of these reports, however, is not optimal. Poor reporting of systematic reviews diminishes their value to clinicians, policy makers, and other users.Since the development of the QUOROM (QUality Of Reporting Of Meta-analysis) Statement--a reporting guideline published in 1999--there have been several conceptual, methodological, and practical advances regarding the conduct and reporting of systematic reviews and meta-analyses. Also, reviews of published systematic reviews have found that key information about these studies is often poorly reported. Realizing these issues, an international group that included experienced authors and methodologists developed PRISMA (Preferred Reporting Items for Systematic reviews and Meta-Analyses) as an evolution of the original QUOROM guideline for systematic reviews and meta-analyses of evaluations of health care interventions.The PRISMA Statement consists of a 27-item checklist and a four-phase flow diagram. The checklist includes items deemed essential for transparent reporting of a systematic review. In this Explanation and Elaboration document, we explain the meaning and rationale for each checklist item. For each item, we include an example of good reporting and, where possible, references to relevant empirical studies and methodological literature. The PRISMA Statement, this document, and the associated Web site (http://www.prisma-statement.org/) should be helpful resources to improve reporting of systematic reviews and meta-analyses.

19Artificial Intelligence in Australian Dental and General Healthcare: A Scoping Review.PubMed

Arosha T Weerakoon, Tonia Girdis, Ove Peters
Aust Dent J. 2025 Dec;70(4):209-256. doi: 10.1111/adj.70000. Epub 2025 Aug 10.
This review contextualises the role of generative Artificial Intelligence (AI) in healthcare within an Australian healthcare regulatory and ethical framework. Four online databases (PubMed, Scopus, CINAHL and Web of Science) were searched for peer-reviewed publications that addressed at least two of the three topics: (1) the current applications of AI in dentistry and healthcare; (2) data security and privacy in AI-enhanced healthcare; (3) ethical, legal and clinical implications of machine errors, and the delegation of healthcare responsibilities to large technology companies. A total of 31 articles were retrieved for full-text analysis using traditional and AI-assisted software. All studies showed promising use of AI to enhance clinical decision-making, automate administrative tasks and augment personalised care. However, integrating AI into the Australian healthcare context introduces complex ethical, regulatory and legal considerations that include bias, data privacy and ambiguous chains of responsibility. To maximise the benefits of AI technologies while safeguarding patient rights, practitioners and developers must establish regulatory frameworks, mandate practitioner training, foster multidisciplinary collaboration and maintain continuous rigorous oversight.

20Ethical and practical challenges of generative AI in healthcare and proposed solutions: a survey.PubMed

Tina Tung, Shah Md Nehal Hasnaeen, Xiaopeng Zhao
Front Digit Health. 2025 Nov 17;7:1692517. doi: 10.3389/fdgth.2025.1692517. eCollection 2025.
BACKGROUND: Generative artificial intelligence (AI) is rapidly transforming healthcare, but its adoption introduces significant ethical and practical challenges. Algorithmic bias, ambiguous liability, lack of transparency, and data privacy risks can undermine patient trust and create health disparities, making their resolution critical for responsible AI integration. OBJECTIVES: This systematic review analyzes the generative AI landscape in healthcare. Our objectives were to: (1) identify AI applications and their associated ethical and practical challenges; (2) evaluate current data-centric, model-centric, and regulatory solutions; and (3) propose a framework for responsible AI deployment. METHODS: Following the PRISMA 2020 statement, we conducted a systematic review of PubMed and Google Scholar for articles published between January 2020 and May 2025. A multi-stage screening process yielded 54 articles, which were analyzed using a thematic narrative synthesis. RESULTS: Our review confirmed AI's growing integration into medical training, research, and clinical practice. Key challenges identified include systemic bias from non-representative data, unresolved legal liability, the "black box" nature of complex models, and significant data privacy risks. Proposed solutions are multifaceted, spanning technical (e.g., explainable AI), procedural (e.g., stakeholder oversight), and regulatory strategies. DISCUSSION: Current solutions are fragmented and face significant implementation barriers. Technical fixes are insufficient without robust governance, clear legal guidelines, and comprehensive professional education. Gaps in global regulatory harmonization and frameworks ill-suited for adaptive AI persist. A multi-layered, socio-technical approach is essential to build trust and ensure the safe, equitable, and ethical deployment of generative AI in healthcare. CONCLUSIONS: The review confirmed that generative AI has a growing integration into medical training, research, and clinical practice. Key challenges identified include systemic bias stemming from non-representative data, unresolved legal liability, the "black box" nature of complex models, and significant data privacy risks. These challenges can undermine patient trust and create health disparities. Proposed solutions are multifaceted, spanning technical (such as explainable AI), procedural (like stakeholder oversight), and regulatory strategies.

21What Should ChatGPT Mean for Bioethics?PubMed

I Glenn Cohen
Am J Bioeth. 2023 Oct;23(10):8-16. doi: 10.1080/15265161.2023.2233357. Epub 2023 Jul 13.
In the last several months, several major disciplines have started their initial reckoning with what ChatGPT and other Large Language Models (LLMs) mean for them - law, medicine, business among other professions. With a heavy dose of humility, given how fast the technology is moving and how uncertain its social implications are, this article attempts to give some early tentative thoughts on what ChatGPT might mean for bioethics. I will first argue that many bioethics issues raised by ChatGPT are similar to those raised by current medical AI - built into devices, decision support tools, data analytics, etc. These include issues of data ownership, consent for data use, data representativeness and bias, and privacy. I describe how these familiar issues appear somewhat differently in the ChatGPT context, but much of the existing bioethical thinking on these issues provides a strong starting point. There are, however, a few "new-ish" issues I highlight - by new-ish I mean issues that while perhaps not truly new seem much more important for it than other forms of medical AI. These include issues about informed consent and the right to know we are dealing with an AI, the problem of medical deepfakes, the risk of oligopoly and inequitable access related to foundational models, environmental effects, and on the positive side opportunities for the democratization of knowledge and empowering patients. I also discuss how races towards dominance (between large companies and between the U.S. and geopolitical rivals like China) risk sidelining ethics.

22Applications and implementation of generative artificial intelligence in cardiovascular imaging with a focus on ethical and legal considerations: what cardiovascular imagers need to know!PubMed

Ahmed Marey, Kevin Christopher Serdysnki, Benjamin D Killeen, et al.
BJR Artif Intell. 2024 May 30;1(1):ubae008. doi: 10.1093/bjrai/ubae008. eCollection 2024 Jan.
Machine learning (ML) and deep learning (DL) have potential applications in medicine. This overview explores the applications of AI in cardiovascular imaging, focusing on echocardiography, cardiac MRI (CMR), coronary CT angiography (CCTA), and CT morphology and function. AI, particularly DL approaches like convolutional neural networks, enhances standardization in echocardiography. In CMR, undersampling techniques and DL-based reconstruction methods, such as variational neural networks, improve efficiency and accuracy. ML in CCTA aids in diagnosing coronary artery disease, assessing stenosis severity, and analyzing plaque characteristics. Automatic segmentation of cardiac structures and vessels using AI is discussed, along with its potential in congenital heart disease diagnosis and 3D printing applications. Overall, AI integration in cardiovascular imaging shows promise for enhancing diagnostic accuracy and efficiency across modalities. The growing use of Generative Adversarial Networks in cardiovascular imaging brings substantial advancements but raises ethical concerns. The "black box" problem in DL models poses challenges for interpretability crucial in clinical practice. Evaluation metrics like ROC curves, image quality, clinical relevance, diversity, and quantitative performance assess GAI models. Automation bias highlights the risk of unquestioned reliance on AI outputs, demanding careful implementation and ethical frameworks. Ethical considerations involve transparency, respect for persons, beneficence, and justice, necessitating standardized evaluation protocols. Health disparities emerge if AI training lacks diversity, impacting diagnostic accuracy. AI language models, like GPT-4, face hallucination issues, posing ethical and legal challenges in healthcare. Regulatory frameworks and ethical governance are crucial for fair and accountable AI. Ongoing research and development are vital to evolving AI ethics.

23Explainable Artificial Intelligence (XAI): What we know and what is left to attain Trustworthy Artificial IntelligenceOpenAlex

Sajid Ali, Tamer Abuhmed, Shaker El–Sappagh, et al.
Artificial intelligence (AI) is currently being utilized in a wide range of sophisticated applications, but the outcomes of many AI models are challenging to comprehend and trust due to their black-box nature. Usually, it is essential to understand the reasoning behind an AI model’s decision-making. Thus, the need for eXplainable AI (XAI) methods for improving trust in AI models has arisen. XAI has become a popular research subject within the AI field in recent years. Existing survey papers have tackled the concepts of XAI, its general terms, and post-hoc explainability methods but there have not been any reviews that have looked at the assessment methods, available tools, XAI datasets, and other related aspects. Therefore, in this comprehensive study, we provide readers with an overview of the current research and trends in this rapidly emerging area with a case study example. The study starts by explaining the background of XAI, common definitions, and summarizing recently proposed techniques in XAI for supervised machine learning. The review divides XAI techniques into four axes using a hierarchical categorization system: (i) data explainability, (ii) model explainability, (iii) post-hoc explainability, and (iv) assessment of explanations. We also introduce available evaluation metrics as well as open-source packages and datasets with future research directions. Then, the significance of explainability in terms of legal demands, user viewpoints, and application orientation is outlined, termed as XAI concerns. This paper advocates for tailoring explanation content to specific user types. An examination of XAI techniques and evaluation was conducted by looking at 410 critical articles, published between January 2016 and October 2022, in reputed journals and using a wide range of research databases as a source of information. The article is aimed at XAI researchers who are interested in making their AI models more trustworthy, as well as towards researchers from other disciplines who are looking for effective XAI methods to complete tasks with confidence while communicating meaning from data.

24Ethical and Legal Governance of Generative AI in Chinese Healthcare.PubMed

Jinrun Jia, Shiqiao Zhao
J Multidiscip Healthc. 2025 Sep 1;18:5405-5419. doi: 10.2147/JMDH.S541271. eCollection 2025.
The application of generative artificial intelligence (AI) technology in the healthcare sector can significantly enhance the efficiency of China's healthcare services. However, risks persist in terms of accuracy, transparency, data privacy, ethics, and bias. These risks are manifested in three key areas: first, the potential erosion of human agency; second, issues of fairness and justice; and third, questions of liability and responsibility. This study reviews and analyzes the legal and regulatory frameworks established in China for the application of generative AI in healthcare, as well as relevant academic literature. Our research findings indicate that while China is actively constructing an ethical and legal governance framework in this field, the regulatory system remains inadequate and faces numerous challenges. These challenges include lagging regulatory rules; an unclear legal status of AI in laws such as the Civil Code; immature standards and regulatory schemes for medical AI training data; and the lack of a coordinated regulatory mechanism among different government departments. In response, this study attempts to establish a governance framework for generative AI in the medical field in China from both legal and ethical perspectives, yielding relevant research findings. Given the latest developments in generative AI in China, it is necessary to address the challenges of its application in the medical field from both ethical and legal perspectives. This includes enhancing algorithm transparency, standardizing medical data management, and promoting AI legislation. As AI technology continues to evolve, more diverse technical models will emerge in the future. This study also proposes that to address potential risks associated with medical AI, efforts should be made to establish a global AI ethics review committee to promote the formation of internationally unified ethical and legal review mechanisms.

25Fairness of artificial intelligence in healthcare: review and recommendations.PubMed

Daiju Ueda, Taichi Kakinuma, Shohei Fujita, et al.
Jpn J Radiol. 2024 Jan;42(1):3-15. doi: 10.1007/s11604-023-01474-3. Epub 2023 Aug 4.
In this review, we address the issue of fairness in the clinical integration of artificial intelligence (AI) in the medical field. As the clinical adoption of deep learning algorithms, a subfield of AI, progresses, concerns have arisen regarding the impact of AI biases and discrimination on patient health. This review aims to provide a comprehensive overview of concerns associated with AI fairness; discuss strategies to mitigate AI biases; and emphasize the need for cooperation among physicians, AI researchers, AI developers, policymakers, and patients to ensure equitable AI integration. First, we define and introduce the concept of fairness in AI applications in healthcare and radiology, emphasizing the benefits and challenges of incorporating AI into clinical practice. Next, we delve into concerns regarding fairness in healthcare, addressing the various causes of biases in AI and potential concerns such as misdiagnosis, unequal access to treatment, and ethical considerations. We then outline strategies for addressing fairness, such as the importance of diverse and representative data and algorithm audits. Additionally, we discuss ethical and legal considerations such as data privacy, responsibility, accountability, transparency, and explainability in AI. Finally, we present the Fairness of Artificial Intelligence Recommendations in healthcare (FAIR) statement to offer best practices. Through these efforts, we aim to provide a foundation for discussing the responsible and equitable implementation and deployment of AI in healthcare.

26Defining AMIA's artificial intelligence principles.PubMed

Anthony E Solomonides, Eileen Koski, Shireen M Atabaki, et al.
J Am Med Inform Assoc. 2022 Mar 15;29(4):585-591. doi: 10.1093/jamia/ocac006.
Recent advances in the science and technology of artificial intelligence (AI) and growing numbers of deployed AI systems in healthcare and other services have called attention to the need for ethical principles and governance. We define and provide a rationale for principles that should guide the commission, creation, implementation, maintenance, and retirement of AI systems as a foundation for governance throughout the lifecycle. Some principles are derived from the familiar requirements of practice and research in medicine and healthcare: beneficence, nonmaleficence, autonomy, and justice come first. A set of principles follow from the creation and engineering of AI systems: explainability of the technology in plain terms; interpretability, that is, plausible reasoning for decisions; fairness and absence of bias; dependability, including "safe failure"; provision of an audit trail for decisions; and active management of the knowledge base to remain up to date and sensitive to any changes in the environment. In organizational terms, the principles require benevolence-aiming to do good through the use of AI; transparency, ensuring that all assumptions and potential conflicts of interest are declared; and accountability, including active oversight of AI systems and management of any risks that may arise. Particular attention is drawn to the case of vulnerable populations, where extreme care must be exercised. Finally, the principles emphasize the need for user education at all levels of engagement with AI and for continuing research into AI and its biomedical and healthcare applications.

27Governance of Responsible AI: From Ethical Guidelines to Cooperative PoliciesOpenAlex

Robert Gianni, Santtu Lehtinen, Mika Nieminen
The increasingly pervasive role of Artificial Intelligence (AI) in our societies is radically changing the way that social interaction takes place within all fields of knowledge. The obvious opportunities in terms of accuracy, speed and originality of research are accompanied by questions about the possible risks and the consequent responsibilities involved in such a disruptive technology. In recent years, this twofold aspect has led to an increase in analyses of the ethical and political implications of AI. As a result, there has been a proliferation of documents that seek to define the strategic objectives of AI together with the ethical precautions required for its acceptable development and deployment. Although the number of documents is certainly significant, doubts remain as to whether they can effectively play a role in safeguarding democratic decision-making processes. Indeed, a common feature of the national strategies and ethical guidelines published in recent years is that they only timidly address how to integrate civil society into the selection of AI objectives. Although scholars are increasingly advocating the necessity to include civil society, it remains unclear which modalities should be selected. If both national strategies and ethics guidelines appear to be neglecting the necessary role of a democratic scrutiny for identifying challenges, objectives, strategies and the appropriate regulatory measures that such a disruptive technology should undergo, the question is then, what measures can we advocate that are able to overcome such limitations? Considering the necessity to operate holistically with AI as a social object, what theoretical framework can we adopt in order to implement a model of governance? What conceptual methodology shall we develop that is able to offer fruitful insights to governance of AI? Drawing on the insights of classical pragmatist scholars, we propose a framework of democratic experimentation based on the method of social inquiry. In this article, we first summarize some of the main points of discussion around the potential societal, ethical and political issues of AI systems. We then identify the main answers and solutions by analyzing current national strategies and ethics guidelines. After showing the theoretical and practical limits of these approaches, we outline an alternative proposal that can help strengthening the active role of society in the discussion about the role and extent of AI systems.

28The Global Governance of Artificial Intelligence: Some Normative ConcernsOpenAlex

Eva Erman, Markus Furendal
Abstract The creation of increasingly complex artificial intelligence (AI) systems raises urgent questions about their ethical and social impact on society. Since this impact ultimately depends on political decisions about normative issues, political philosophers can make valuable contributions by addressing such questions. Currently, AI development and application are to a large extent regulated through non-binding ethics guidelines penned by transnational entities. Assuming that the global governance of AI should be at least minimally democratic and fair, this paper sets out three desiderata that an account should satisfy when theorizing about what this means. We argue, first, that an analysis of democratic values, political entities and decision-making should be done in a holistic way; second, that fairness is not only about how AI systems treat individuals, but also about how the benefits and burdens of transformative AI are distributed; and finally, that justice requires that governance mechanisms are not limited to AI technology, but are incorporated into a range of basic institutions. Thus, rather than offering a substantive theory of democratic and fair AI governance, our contribution is metatheoretical: we propose a theoretical framework that sets up certain normative boundary conditions for a satisfactory account.

29A survey on providing customer and public administration based services using AI: chatbotOpenAlex

Krishna Kumar Nirala, Nikhil Kumar Singh, Vinay Shivshanker Purani
A chatbot is emerged as an effective tool to address the user queries in automated, most appropriate and accurate way. Depending upon the complexity of the subject domain, researchers are employing variety of soft-computing techniques to make the chatbot user-friendly. It is observed that chatbots have flooded the globe with wide range of services including ordering foods, suggesting products, advising for insurance policies, providing customer support, giving financial assistance, schedule meetings etc. However, public administration based services wherein chatbot intervention influence the most, is not explored yet. This paper discuses about artificial intelligence based chatbots including their applications, challenges, architecture and models. It also talks about evolution of chatbots starting from Turing Test and Rule-based chatbots to advanced Artificial Intelligence based Chatbots (AI-Chatbots). AI-Chatbots are providing much kind of services, which this paper outlines into two main aspects including customer based services and public administration based services. The purpose of this survey is to understand and explore the possibility of customer & public administration services based chatbot. The survey demonstrates that there exist an immense potential in the AI assisted chatbot system for providing customer services and providing better governance in public administration services.

30A comprehensive AI policy education framework for university teaching and learningOpenAlex

Cecilia Ka Yuk Chan
Abstract This study aims to develop an AI education policy for higher education by examining the perceptions and implications of text generative AI technologies. Data was collected from 457 students and 180 teachers and staff across various disciplines in Hong Kong universities, using both quantitative and qualitative research methods. Based on the findings, the study proposes an AI Ecological Education Policy Framework to address the multifaceted implications of AI integration in university teaching and learning. This framework is organized into three dimensions: Pedagogical, Governance, and Operational. The Pedagogical dimension concentrates on using AI to improve teaching and learning outcomes, while the Governance dimension tackles issues related to privacy, security, and accountability. The Operational dimension addresses matters concerning infrastructure and training. The framework fosters a nuanced understanding of the implications of AI integration in academic settings, ensuring that stakeholders are aware of their responsibilities and can take appropriate actions accordingly.

31Bridging Cities and Citizens with Generative AI: Public Readiness and Trust in Urban PlanningOpenAlex

Adnan Alshahrani
As part of its modernisation and economic diversification policies, Saudi Arabia is building smart, sustainable cities intended to improve quality of life and meet environmental goals. However, involving the public in urban planning remains complex, with traditional methods often proving expensive, time-consuming, and inaccessible to many groups. Integrating artificial intelligence (AI) into public participation may help to address these limitations. This study explores whether Saudi residents are ready to engage with AI-driven tools in urban planning, how they prefer to interact with them, and what ethical concerns may arise. Using a quantitative, survey-based approach, the study collected data from 232 Saudi residents using non-probability stratified sampling. The survey assessed demographic influences on AI readiness, preferred engagement methods, and perceptions of ethical risks. The results showed a strong willingness among participants (200 respondents, 86%)—especially younger and university-educated respondents—to engage through AI platforms. Visual tools such as image and video analysis were the most preferred (96 respondents, 41%), while chatbots were less favoured (16 respondents, 17%). However, concerns were raised about privacy (76 respondents, 33%), bias (52 respondents, 22%), and over-reliance on technology (84 respondents, 36%). By exploring the intersection of generative AI and participatory urban governance, this study contributes directly to the discourse on inclusive smart city development. The research also offers insights into how AI-driven public engagement tools can be integrated into urban planning workflows to enhance the design, governance, and performance of the built environment. The findings suggest that AI has the potential to improve inclusivity and responsiveness in urban planning, but that its success depends on public trust, ethical safeguards, and the thoughtful design of accessible, user-friendly engagement platforms.

32Students’ voices on generative AI: perceptions, benefits, and challenges in higher educationOpenAlex

Cecilia Ka Yuk Chan, Wenjie Hu
Abstract This study explores university students’ perceptions of generative AI (GenAI) technologies, such as ChatGPT, in higher education, focusing on familiarity, their willingness to engage, potential benefits and challenges, and effective integration. A survey of 399 undergraduate and postgraduate students from various disciplines in Hong Kong revealed a generally positive attitude towards GenAI in teaching and learning. Students recognized the potential for personalized learning support, writing and brainstorming assistance, and research and analysis capabilities. However, concerns about accuracy, privacy, ethical issues, and the impact on personal development, career prospects, and societal values were also expressed. According to John Biggs’ 3P model, student perceptions significantly influence learning approaches and outcomes. By understanding students’ perceptions, educators and policymakers can tailor GenAI technologies to address needs and concerns while promoting effective learning outcomes. Insights from this study can inform policy development around the integration of GenAI technologies into higher education. By understanding students’ perceptions and addressing their concerns, policymakers can create well-informed guidelines and strategies for the responsible and effective implementation of GenAI tools, ultimately enhancing teaching and learning experiences in higher education.

33Improving PLS-SEM use for business marketing researchOpenAlex

Peter Guenther, Miriam Guenther, Christian M. Ringle, et al.
A review of studies published in Industrial Marketing Management over the past two decades and more shows that these studies not only used partial least squares structural equation modeling (PLS-SEM) widely to estimate and empirically substantiate theoretically established models with constructs, but did so increasingly. In line with their study goals, researchers provided reasons for using PLS-SEM (e.g., model complexity, limited sample size, and prediction). These reasons are frequently not fully convincing, requiring further clarification. Additionally, our review reveals that researchers' assessment and reporting of their measurement and structural models are insufficient. Certain tests and thresholds that they use are also inappropriate. Finally, researchers seldom apply more advanced PLS-SEM analytic techniques, although these can support the results' robustness and may create new insights. This paper addresses the issues by reviewing business marketing studies to clarify PLS-SEM's appropriate use. Furthermore, the paper provides researchers and practitioners in the business marketing field with a best practice orientation and describes new opportunities for using PLS-SEM. To this end, the paper offers guidelines and checklists to support future PLS-SEM applications.

34Implications of the use of artificial intelligence in public governance: A systematic literature review and a research agendaOpenAlex

Anneke Zuiderwijk, Yu‐Che Chen, Fadi Salem
To lay the foundation for the special issue that this research article introduces, we present 1) a systematic review of existing literature on the implications of the use of Artificial Intelligence (AI) in public governance and 2) develop a research agenda. First, an assessment based on 26 articles on this topic reveals much exploratory, conceptual, qualitative, and practice-driven research in studies reflecting the increasing complexities of using AI in government – and the resulting implications, opportunities, and risks thereof for public governance. Second, based on both the literature review and the analysis of articles included in this special issue, we propose a research agenda comprising eight process-related recommendations and seven content-related recommendations. Process-wise, future research on the implications of the use of AI for public governance should move towards more public sector-focused, empirical, multidisciplinary, and explanatory research while focusing more on specific forms of AI rather than AI in general. Content-wise, our research agenda calls for the development of solid, multidisciplinary, theoretical foundations for the use of AI for public governance, as well as investigations of effective implementation, engagement, and communication plans for government strategies on AI use in the public sector. Finally, the research agenda calls for research into managing the risks of AI use in the public sector, governance modes possible for AI use in the public sector, performance and impact measurement of AI use in government, and impact evaluation of scaling-up AI usage in the public sector.

35Governance of artificial intelligence: A risk and guideline-based integrative frameworkOpenAlex

Bernd W. Wirtz, Jan C. Weyerer, Ines Kehl

36Accountable Artificial Intelligence: Holding Algorithms to AccountOpenAlex

Madalina Busuioc
Artificial intelligence (AI) algorithms govern in subtle yet fundamental ways the way we live and are transforming our societies. The promise of efficient, low-cost, or "neutral" solutions harnessing the potential of big data has led public bodies to adopt algorithmic systems in the provision of public services. As AI algorithms have permeated high-stakes aspects of our public existence-from hiring and education decisions to the governmental use of enforcement powers (policing) or liberty-restricting decisions (bail and sentencing)-this necessarily raises important accountability questions: What accountability challenges do AI algorithmic systems bring with them, and how can we safeguard accountability in algorithmic decision-making? Drawing on a decidedly public administration perspective, and given the current challenges that have thus far become manifest in the field, we critically reflect on and map out in a conceptually guided manner the implications of these systems, and the limitations they pose, for public accountability.

37The loopholes of algorithmic public services: an “intelligent” accountability research agendaOpenAlex

Enrico Bracci
Purpose Governments are increasingly turning to artificial intelligence (AI) algorithmic systems to increase efficiency and effectiveness of public service delivery. While the diffusion of AI offers several desirable benefits, caution and attention should be posed to the accountability of AI algorithm decision-making systems in the public sector. The purpose of this paper is to establish the main challenges that an AI algorithm might bring about to public service accountability. In doing so, the paper also delineates future avenues of investigation for scholars. Design/methodology/approach This paper builds on previous literature and anecdotal cases of AI applications in public services, drawing on streams of literature from accounting, public administration and information technology ethics. Findings Based on previous literature, the paper highlights the accountability gaps that AI can bring about and the possible countermeasures. The introduction of AI algorithms in public services modifies the chain of responsibility. This distributed responsibility requires an accountability governance, together with technical solutions, to meet multiple accountabilities and close the accountability gaps. The paper also delineates a research agenda for accounting scholars to make accountability more “intelligent”. Originality/value The findings of the paper shed new light and perspective on how public service accountability in AI should be considered and addressed. The results developed in this paper will stimulate scholars to explore, also from an interdisciplinary perspective, the issues public service organizations are facing to make AI algorithms accountable.

38Artificial Intelligence and Public Values: Value Impacts and Governance in the Public SectorOpenAlex

Yu‐Che Chen, Michael J. Ahn, Yifan Wang
While there has been growth in the literature exploring the governance of artificial intelligence (AI) and recognition of the critical importance of guiding public values, the literature lacks a systematic study focusing on public values as well as the governance challenges and solutions to advance these values. This article conducts a systematic literature review of the relationships between the public sector AI and public values to identify the impacts on public values and the governance challenges and solutions. It further explores the perspectives of U.S. government employees on AI governance and public values via a national survey. The results suggest the need for a broad inclusion of diverse public values, the salience of transparency regarding several governance challenges, and the importance of stakeholder participation and collaboration as governance solutions. This article also explores and reports the nuances in these results and their practical implications.

39A Review of Artificial Intelligence in Government and its Potential from a Public Policy PerspectiveOpenAlex

David Valle-Cruz, Edgar A. Ruvalcaba-Gómez, Rodrigo Sandoval‐Almazán, et al.
Artificial intelligence (AI) is the latest trend being implemented in the public sector. Recent advances in this field and the AI explosion in the private sector have served to promote a revolution for government, public service management, accountability, and public value. Incipient research to understand, conceptualize and express challenges and limitations is now ongoing. This paper is the first approach in such a direction; our research question is: What are the current AI trends in the public sector? In order to achieve that goal, we collected 78 papers related to this new field in recent years. We also used a public policy framework to identify future areas of implementation for this trend. We found that only normative and exploratory papers have been published so far and there are a lot of public policy challenges facing in this area, and that AI implementation results are unknown and unexpected; since there may be great benefits for governments and society, but, on the other hand, it may have negative results like the so-called ”algorithmic bias” of AI when making important decisions for social development. However, we consider that AI has potential benefits in the public health, public policies on climate change, public management, decision-making, disaster prevention and response, improving government-citizen interaction, personalization of services, interoperability, analyzing large amounts of data, detecting abnormalities and patterns, and discovering new solutions through dynamic models and simulation in real time.

40Algorithmic Bias as a Core Legal Dilemma in the Age of Artificial Intelligence: Conceptual Basis and the Current State of RegulationOpenAlex

Gergely Ferenc Lendvai, Gergely Gosztonyi
This article examines algorithmic bias as a pressing legal challenge, situating the issue within the broader context of artificial intelligence (AI) governance. We employed comparative legal analysis and reviewed pertinent regulatory documents to examine how the fragmented U.S. approaches and the EU’s user-centric legal frameworks, such as the GDPR, DSA, and AI Act, address the systemic risks posed by biased algorithms. The findings underscore persistent enforcement gaps, particularly concerning opaque black-box algorithmic design, which hampers bias detection and remediation. The paper highlights how current regulatory efforts disproportionately affect marginalized communities and fail to provide effective protection across jurisdictions. It also identifies structural imbalances in legal instruments, particularly in relation to risk classification, transparency, and fairness standards. Notably, emerging regulations often lack the technical and ethical capacity for implementation. We argue that global cooperation is not only necessary but inevitable, as regional solutions alone are insufficient to govern transnational AI systems. Without harmonized international standards, algorithmic bias will continue to reproduce existing inequalities under the guise of objectivity. The article advocates for inclusive, cross-sectoral collaboration among governments, developers, and civil society to ensure the responsible development of AI and uphold fundamental rights.

41Privacy Protection in Using Artificial Intelligence for Healthcare: Chinese Regulation in Comparative PerspectiveOpenAlex

Chao Wang, Jieyu Zhang, Nicholas Lassi, et al.
Advanced artificial intelligence (AI) technologies are now widely employed in China's medical and healthcare fields. Enormous amounts of personal data are collected from various sources and inserted into AI algorithms for medical purposes, producing challenges to patient's privacy. This is a comparative study of Chinese, United States, and European Union operational rules for healthcare data that is collected and then used in AI functions, particularly focusing on legal differences and deficiencies. The conceptual boundaries of privacy and personal information, the influence of technological development on the informed consent model, and conflicts between freedom and security in rules of cross-border data flow were found to be key issues requiring consideration when regulating healthcare data used for AI purposes. Furthermore, the results indicate that the appropriate balance between privacy protections and technological development, between individual and group interests, and between corporate profits and the public interest should be identified and observed. In terms of specific rule-making, it was found that China should establish special regulations protecting healthcare information, provide clear definitions and classification schemas for different types of healthcare information, and enact stricter accountability mechanisms. Examining and contrasting operational rules for AI in health care promotes informed privacy governance and improved privacy legislation.

42Ethical implications of AI and robotics in healthcare: A reviewOpenAlex

Chukwuka Elendu, Dependable C. Amaechi, Tochi C. Elendu, et al.
Integrating Artificial Intelligence (AI) and robotics in healthcare heralds a new era of medical innovation, promising enhanced diagnostics, streamlined processes, and improved patient care. However, this technological revolution is accompanied by intricate ethical implications that demand meticulous consideration. This article navigates the complex ethical terrain surrounding AI and robotics in healthcare, delving into specific dimensions and providing strategies and best practices for ethical navigation. Privacy and data security are paramount concerns, necessitating robust encryption and anonymization techniques to safeguard patient data. Responsible data handling practices, including decentralized data sharing, are critical to preserve patient privacy. Algorithmic bias poses a significant challenge, demanding diverse datasets and ongoing monitoring to ensure fairness. Transparency and explainability in AI decision-making processes enhance trust and accountability. Clear responsibility frameworks are essential to address the accountability of manufacturers, healthcare institutions, and professionals. Ethical guidelines, regularly updated and accessible to all stakeholders, guide decision-making in this dynamic landscape. Moreover, the societal implications of AI and robotics extend to accessibility, equity, and societal trust. Strategies to bridge the digital divide and ensure equitable access must be prioritized. Global collaboration is pivotal in developing adaptable regulations and addressing legal challenges like liability and intellectual property. Ethics must remain at the forefront in the ever-evolving realm of healthcare technology. By embracing these strategies and best practices, healthcare systems and professionals can harness the potential of AI and robotics, ensuring responsible and ethical integration that benefits patients while upholding the highest ethical standards.

43Ethical and legal considerations in healthcare AI: innovation and policy for safe and fair useOpenAlex

Tuan D. Pham
Artificial intelligence (AI) is transforming healthcare by enhancing diagnostics, personalizing medicine and improving surgical precision. However, its integration into healthcare systems raises significant ethical and legal challenges. This review explores key ethical principles-autonomy, beneficence, non-maleficence, justice, transparency and accountability-highlighting their relevance in AI-driven decision-making. Legal challenges, including data privacy and security, liability for AI errors, regulatory approval processes, intellectual property and cross-border regulations, are also addressed. As AI systems become increasingly autonomous, questions of responsibility and fairness must be carefully considered, particularly with the potential for biased algorithms to amplify healthcare disparities. This paper underscores the importance of multi-disciplinary collaboration between technologists, healthcare providers, legal experts and policymakers to create adaptive, globally harmonized frameworks. Public engagement is emphasized as essential for fostering trust and ensuring ethical AI adoption. With AI technologies advancing rapidly, a flexible regulatory environment that evolves with innovation is critical. Aligning AI innovation with ethical and legal imperatives will lead to a safer, more equitable healthcare system for all.

44Gender bias perpetuation and mitigation in AI technologies: challenges and opportunitiesOpenAlex

Sinead O’Connor, Helen K. Liu
Abstract Across the world, artificial intelligence (AI) technologies are being more widely employed in public sector decision-making and processes as a supposedly neutral and an efficient method for optimizing delivery of services. However, the deployment of these technologies has also prompted investigation into the potentially unanticipated consequences of their introduction, to both positive and negative ends. This paper chooses to focus specifically on the relationship between gender bias and AI, exploring claims of the neutrality of such technologies and how its understanding of bias could influence policy and outcomes. Building on a rich seam of literature from both technological and sociological fields, this article constructs an original framework through which to analyse both the perpetuation and mitigation of gender biases, choosing to categorize AI technologies based on whether their input is text or images. Through the close analysis and pairing of four case studies, the paper thus unites two often disparate approaches to the investigation of bias in technology, revealing the large and varied potential for AI to echo and even amplify existing human bias, while acknowledging the important role AI itself can play in reducing or reversing these effects. The conclusion calls for further collaboration between scholars from the worlds of technology, gender studies and public policy in fully exploring algorithmic accountability as well as in accurately and transparently exploring the potential consequences of the introduction of AI technologies.

45The dark side of generative artificial intelligence: A critical analysis of controversies and risks of ChatGPTOpenAlex

Krzysztof Wach, Cong Doanh Duong, Joanna Ejdys, et al.
Objective: The objective of the article is to provide a comprehensive identification and understanding of the challenges and opportunities associated with the use of generative artificial intelligence (GAI) in business. This study sought to develop a conceptual framework that gathers the negative aspects of GAI development in management and economics, with a focus on ChatGPT. Research Design & Methods: The study employed a narrative and critical literature review and developed a conceptual framework based on prior literature. We used a line of deductive reasoning in formulating our theoretical framework to make the study's overall structure rational and productive. Therefore, this article should be viewed as a conceptual article that highlights the controversies and threats of GAI in management and economics, with ChatGPT as a case study. Findings: Based on the conducted deep and extensive query of academic literature on the subject as well as professional press and Internet portals, we identified various controversies, threats, defects, and disadvantages of GAI, in particular ChatGPT. Next, we grouped the identified threats into clusters to summarize the seven main threats we see. In our opinion they are as follows: (i) no regulation of the AI market and urgent need for regulation, (ii) poor quality, lack of quality control, disinformation, deepfake content, algorithmic bias, (iii) automation-spurred job losses, (iv) personal data violation, social surveillance, and privacy violation, (v) social manipulation, weakening ethics and goodwill, (vi) widening socio-economic inequalities, and (vii) AI technostress. Implications & Recommendations: It is important to regulate the AI/GAI market. Advocating for the regulation of the AI market is crucial to ensure a level playing field, promote fair competition, protect intellectual property rights and privacy, and prevent potential geopolitical risks. The changing job market requires workers to continuously acquire new (digital) skills through education and retraining. As the training of AI systems becomes a prominent job category, it is important to adapt and take advantage of new opportunities. To mitigate the risks related to personal data violation, social surveillance, and privacy violation, GAI developers must prioritize ethical considerations and work to develop systems that prioritize user privacy and security. To avoid social manipulation and weaken ethics and goodwill, it is important to implement responsible AI practices and ethical guidelines: transparency in data usage, bias mitigation techniques, and monitoring of generated content for harmful or misleading information. Contribution & Value Added: This article may aid in bringing attention to the significance of resolving the ethical and legal considerations that arise from the use of GAI and ChatGPT by drawing attention to the controversies and hazards associated with these technologies.

46Generating scholarly content with ChatGPT: ethical challenges for medical publishingOpenAlex

Michael Liebrenz, Roman Schleifer, Anna Buadze, et al.
The impact of generative artificial intelligence (AI) on medical publishing practices is currently unknown. However, as our experiences underline, generative AI could have substantial ethical implications. ChatGPT (OpenAI, San Francisco, CA, USA) is an AI chatbot released in November, 2022.1Open AI ChatGPT.https://openai.com/blog/chatgpt/Date: 2022Date accessed: December 21, 2022Google Scholar Developed using human feedback and freely accessible, the platform has already attracted millions of interactions.2Grant N Metz C A New chat bot is a ‘code red’ for Google's search business. The New York Times, Dec 21, 2022https://www.nytimes.com/2022/12/21/technology/ai-chatgpt-google-search.htmlDate accessed: December 23, 2022Google Scholar When presented with a query, ChatGPT will automatically generate a response, which is based on thousands of internet sources, often without further input from the user. Resultantly, individuals have reportedly used ChatGPT to formulate university essays and scholarly articles3Bowman E AI bot ChatGPT stuns academics with essay-writing skills and usability. NPR, Dec 19, 2022https://www.npr.org/2022/12/19/1143912956/chatgpt-ai-chatbot-homework-academiaDate accessed: December 21, 2022Google Scholar and, if prompted, the system can deliver accompanying references. Given these accounts and its popular usage, we requested that ChatGPT write a Comment for The Lancet Digital Health about AI and medical publishing ethics. We then asked ChatGPT how the editorial team should address academic content produced by AI. The results make for interesting reading (appendix). The functionality of ChatGPT highlights the growing necessity of implementing robust AI author guidelines in scholarly publishing. Ethical considerations abound concerning copyright, attribution, plagiarism, and authorship when AI produces academic text. These concerns are especially pertinent because whether copy is AI generated is currently imperceptible to human readers and anti-plagiarism software. Studies across various fields have already listed ChatGPT as an author,4Frye B Should using an AI text generator to produce academic writing be plagiarism?.SSRN. 2022; (published online Dec 20.) (preprint).https://ssrn.com/abstract=4292283Google Scholar but whether generative AI fulfils the International Committee of Medical Journal Editors' criteria for authorship is a point of debate: can a chatbot really provide approval for work and be accountable for its contents? The Committee on Publication Ethics has developed AI recommendations for editorial decision making5Committee on Publication EthicsArtificial intelligence (AI) in decision making.https://doi.org/10.24318/9kvAgrnJDate: 2021Date accessed: December 20, 2022Google Scholar and the trade body for scholarly publishers, the International Association of Scientific, Technical, and Medical Publishers, produced a white paper on AI ethics.6International Association of Scientific, Technical, and Medical PublishersAI ethics in scholarly communication—STM best practice principles for ethical, trustworthy and human-centric AI.https://www.stm-assoc.org/2021_05_11_STM_AI_White_Paper_April2021.pdfDate: 2021Date accessed: December 21, 2022Google Scholar As technologies become better tailored to user needs and more commonly adopted, we believe comprehensive discussions about authorship policies are urgent and essential. Elsevier, who publish the Lancet family of journals, alongside other major publishers, have stated that AI cannot be listed as an author and its use must be properly acknowledged.7ElsevierPublishing ethics. Elsevier.https://www.elsevier.com/about/policies/publishing-ethicsDate accessed: February 1, 2023Google Scholar ChatGPT is available to use without cost.1Open AI ChatGPT.https://openai.com/blog/chatgpt/Date: 2022Date accessed: December 21, 2022Google Scholar However, OpenAI's leadership have affirmed that free use is temporary and the product will eventually be monetised.8Karpf D Money will kill ChatGPT's magic. The Atlantic, Dec 21, 2022https://www.theatlantic.com/technology/archive/2022/12/chatgpt-ai-chatbots-openai-cost-regulations/672539/Date accessed: December 23, 2022Google Scholar One commercial option for the platform could conceivably involve some form of paywall, which might entrench existing international inequalities in scholarly publishing. Although institutions in socioeconomically advantaged areas could probably afford access, those in low-income and middle-income countries might not be able to, thus widening existing disparities in knowledge dissemination and scholarly publishing. In our opinion, as the program remains freely available in the short term, ChatGPT's ease of use and accessibility could substantially increase scholarly output. ChatGPT might democratise the dissemination of knowledge since the chatbot can receive and produce copy in multiple languages, circumventing English-language requirements that can be a publishing barrier for speakers of other languages. Nonetheless, the functionality of ChatGPT has the capacity to cause harm by producing misleading or inaccurate content,3Bowman E AI bot ChatGPT stuns academics with essay-writing skills and usability. NPR, Dec 19, 2022https://www.npr.org/2022/12/19/1143912956/chatgpt-ai-chatbot-homework-academiaDate accessed: December 21, 2022Google Scholar thereby eliciting concerns around scholarly misinformation. As the so-called COVID-19 infodemic shows, the potential spread of misinformation in medical publishing can entail significant societal hazards.9The Lancet Infectious DiseasesThe COVID-19 infodemic.Lancet Infect Dis. 2020; 20: 875Summary Full Text Full Text PDF PubMed Scopus (198) Google Scholar Listed by OpenAI as a limitation, “ChatGPT sometimes writes plausible-sounding but incorrect or nonsensical answers”;1Open AI ChatGPT.https://openai.com/blog/chatgpt/Date: 2022Date accessed: December 21, 2022Google Scholar interestingly, the chatbot itself highlighted this possibility when responding to us (appendix). The early rollout of ChatGPT will inevitably spawn competitors, potentially rendering this a far-reaching problem. Accordingly, per ChatGPT's response to our query, The Lancet Digital Health should “carefully consider the ethical implications of publishing articles produced by AI.” We would go further: as pioneers of publishing ethics and academic standards, we call on The Lancet Digital Health and the Lancet family to initiate rigorous exchanges around the implications of AI-generated content within scholarly publishing, with a view to creating comprehensive guidance. ChatGPT's burgeoning popularity and our experiences illustrate that the time for these conversations is now; after all, can you really be sure that what you are currently reading was written by human authors? We declare no competing interests. Download .pdf (.19 MB) Help with pdf files Supplementary appendix

47The Impact of Artificial Intelligence Tools on Academic Writing Instruction in Higher Education: A Systematic ReviewOpenAlex

Hind Aljuaid
With the growth of Artificial Intelligence technologies, there is interest in studying their potential impact on university academic writing courses. This study examined whether AI tools are replacing these courses by exploring how they effectively replace traditional academic writing instruction and this shift’s potential benefits and drawbacks. The researcher reviewed existing literature on integrating AI tools into academic writing instruction. The findings provide insights to educators navigating the integration of Artificial Intelligence tools into writing curricula while maintaining instructional quality and academic integrity standards. By synthesizing the latest research, this study can inform decisions about the appropriate use of Artificial Intelligence in teaching essential writing skills. Increased use of Artificial Intelligence writing tools has sparked debate about their role in academic writing instruction. Universities like Stanford have updated policies around Artificial Intelligence tool usage and academic integrity. The University of California issued guidance acknowledging the prevalence of generative Artificial Intelligence on campuses. Middlebury College banned classroom use of ChatGPT over concerns it could impede critical thinking and writing skill development. Results show that while Artificial Intelligence helps with grammar and style, questions remain about its impact on creativity and critical thinking. However, Artificial Intelligence is not replacing university writing courses. These courses teach critical thinking, research, citation, argumentation, creativity, originality, and ethics, which Artificial Intelligence lacks. Academic writing courses offer a complete learning experience. Artificial Intelligence may improve academic writing but is unlikely to replace traditional courses soon. A balanced approach integrating Artificial Intelligence support while preserving core elements of academic writing education appears most effective for preparing students for diverse writing challenges.

48Two paths of balancing technology and ethics: A comparative study on AI governance in China and GermanyOpenAlex

Viktor Tuzov, Fen Lin

49Analyzing Ghana's Pharmacy Act, 1994 (Act 489) Regarding Quality Control and Negligence Liability Measures for Artificial Intelligence Pharmacy SystemsOpenAlex

George Mensah, Maad M. Mıjwıl, Ioannis Adamopoulos
The objective of this systematic review was to assess the adequacy of current medication management in Ghana considering the risks posed by increased artificial intelligence (AI) automation in pharmacies worldwide A qualitative comparative approach was used despite reviewed the Ghana 1994 Pharmacy Act against recognition of AI challenges and international governance guidelines . The results revealed flaws in terms of quality prerequisites, transparency checklists and liability mechanisms developed for AI systems compared to existing regulations of the manual process. Outdated approaches to patient care that fail to ensure patient safety or address threats to the accuracy of recommendations from data collection biases and technical errors. Proposed changes include a requirement for usability testing before approving AI pharmacy deployments and the creation of a review board to review post-implementation systems for validity. Updating regulations to deal with modern equipment puts innovation and responsible regulation in the fast-paced healthcare industry. This study contributes significantly to preliminary research on AI policy readiness in the Ghanaian legal context, and suggests a feasible methodology for exploring qualitative differences for use in companies and countries competing for technology a disturbing, increasingly beyond the date code. Early government reform helps keep pace with the realities of adoption.

50Competencies for the Use of Artificial Intelligence in Primary Care.PubMed

Winston Liaw, Jacqueline K Kueper, Steven Lin, et al.
Ann Fam Med. 2022 Nov-Dec;20(6):559-563. doi: 10.1370/afm.2887.
The artificial intelligence (AI) revolution has arrived for the health care sector and is finally penetrating the far-reaching but perpetually underfinanced primary care platform. While AI has the potential to facilitate the achievement of the Quintuple Aim (better patient outcomes, population health, and health equity at lower costs while preserving clinician well-being), inattention to primary care training in the use of AI-based tools risks the opposite effects, imposing harm and exacerbating inequalities. The impact of AI-based tools on these aims will depend heavily on the decisions and skills of primary care clinicians; therefore, appropriate medical education and training will be crucial to maximize potential benefits and minimize harms. To facilitate this training, we propose 6 domains of competency for the effective deployment of AI-based tools in primary care: (1) foundational knowledge (what is this tool?), (2) critical appraisal (should I use this tool?), (3) medical decision making (when should I use this tool?), (4) technical use (how do I use this tool?), (5) patient communication (how should I communicate with patients regarding the use of this tool?), and (6) awareness of unintended consequences (what are the "side effects" of this tool?). Integrating these competencies will not be straightforward because of the breadth of knowledge already incorporated into family medicine training and the constantly changing technological landscape. Nonetheless, even incremental increases in AI-relevant training may be beneficial, and the sooner these challenges are tackled, the sooner the primary care workforce and those served by it will begin to reap the benefits.

51AI adoption and diffusion in public administration: A systematic literature review and future research agendaOpenAlex

Rohit Madan, Mona Ashok
Artificial Intelligence (AI) implementation in public administration is gaining momentum heralded by the hope of smart public services that are personalised, lean, and efficient. However, the use of AI in public administration is riddled with ethical tensions of fairness, transparency, privacy, and human rights. We call these AI tensions. The current literature lacks a contextual and processual understanding of AI adoption and diffusion in public administration to be able to explore such tensions. Previous studies have outlined risks, benefits, and challenges with the use of AI in public administration. However, a large gap remains in understanding AI tensions as they relate to public value creation. Through a systematic literature review grounded in public value management and the resource-based view of the firms, we identify technology-organisational-environmental (TOE) contextual variables and absorptive capacity as factors influencing AI adoption as discussed in the literature. To our knowledge, this is the first paper that outlines distinct AI tensions from an AI implementation and diffusion perspective within public administration. We develop a future research agenda for the full AI innovation lifecycle of adoption, implementation, and diffusion.

52Copyright Protection for AI-Generated Works: Exploring Originality and Ownership in a Digital LandscapeOpenAlex

Hafiz GAFFAR, Saleh Hamed Albarashdi
Abstract This research explores AI-generated originality's impact on copyright regulations. It meticulously examines legal frameworks such as the Berne Convention, EU Copyright Law, and national legislation. Rigorously analyzing cases, including Infopaq International A/S v Danske Dagblades Forening and Levola Hengelo BV v Smilde Foods BV, illuminates evolving originality and human involvement in AI creativity. The study also contemplates global perspectives, drawing from esteemed organizations such as the World Intellectual Property Organization and the European Court of Justice and exploring diverse approaches adopted by individual nations. The paper emphasizes the imperative need for legislative updates to address the challenges and opportunities of AI-generated works. It highlights the pivotal role of international collaboration and public awareness in shaping copyright policies for the AI-driven creativity era. It also offers insights and recommendations for policymakers and researchers navigating this complex terrain.

53Authorship in artificial intelligence‐generated works: Exploring originality in text prompts and artificial intelligence outputs through philosophical foundations of copyright and collage protectionOpenAlex

Francesca Mazzi
Abstract The advent of artificial intelligence (AI) and its generative capabilities have propelled innovation across various industries, yet they have also sparked intricate legal debates, particularly in the realm of copyright law. Generative AI systems, capable of producing original content based on user‐provided input or prompts, have introduced novel challenges regarding ownership and authorship of AI‐generated works. One crucial aspect of this discussion revolves around text prompts, which serve as instructions for AI systems to generate specific content types, be it text, images, or music. Despite the transformative potential of AI‐generated works, the legal landscape remains fragmented, with disparate jurisdictional interpretations and a lack of uniform approaches. This disparity has led to legal uncertainty and ambiguity, necessitating a nuanced exploration of originality, creativity, and legal principles in the context of text prompts and resulting outputs. This article seeks to contribute to the ongoing debate by delving into the complexities surrounding AI‐generated works, focusing specifically on the originality of text prompts and their correlation with resulting outputs. While previous literature has extensively examined copyright issues related to AI, the originality of text prompts remains largely unexplored, representing a significant gap in the existing discourse. By analysing the originality of text prompts, this article aims to uncover new insights into the creative process underlying AI‐generated works and its implications for copyright law. Drawing parallels from traditional creative works, such as collages, the article will assess how legal principles apply to AI‐generated content, considering philosophical foundations as well as copyright principles, such as the idea‐expression dichotomy. Furthermore, the article will explore the divergent approaches taken by different jurisdictions, including the United Kingdom, United States, and European Union, in determining originality in the context of copyright law. While refraining from providing definitive answers, the article aims to stimulate critical thinking and dialogue among stakeholders. By offering fresh perspectives and insights, it seeks to enrich the discourse surrounding the copyrightability of AI‐generated works and pave the way for informed policy decisions and legal interpretations. The article aims to contribute valuable perspectives to the ongoing debate on copyright and AI, shaping the future trajectory of intellectual property law in the era of artificial intelligence.

54The Copyright Protection Dilemma of AI-Generated Content: A Comparative Study of Legislative Approaches in the European Union, the United States, and ChinaOpenAlex

Haoran Wang
This study focuses on the legal dilemmas faced by Artificial Intelligence-Generated Content (AIGC) in the field of copyright protection. By comparing the legislative approaches and practical developments in the European Union, the United States, and China, it explores the key issues and underlying causes surrounding AIGC copyright protection. The research finds that AIGC presents multiple challenges, including the identification of authorship, the establishment of originality standards, the quantification of human involvement, and the coordination with traditional copyright systems. The three jurisdictions have adopted divergent legislative and judicial interpretative paths in addressing these challenges, reflecting their respective legal traditions and policy orientations. Through empirical analysis, the study evaluates the applicability and effectiveness of these varied approaches and proposes a legal framework for AIGC copyright protection that is both inclusive and forward-looking. The goal is to strike a balance between safeguarding incentives for innovation and promoting technological advancement. The findings have significant theoretical and practical implications for improving the international copyright legal regime and responding to the intellectual property challenges of the AI era.

55Artificial intelligence as producer and consumer of copyright works: evaluating the consequences of algorithmic creativityOpenAlex

Enrico Bonadio, Luke McDonagh
In copyright theory, property rights are justified in large part by, e.g., the Lockean argument that the human has laboured to create the work; or in Kantian terms, by emphasising that the work arose from the personality of the human author. 19 Such theories do not fit neatly with a non-human author.Thus, as machines have learned to mimic human creativity, the copyright world has accordingly entered into AI-driven uncharted territory.We aim here to navigate through this territory, providing a road-map of what the legal repercussions of AI are, and guidance on what routes the law should take in the future. 20 Given the wide-ranging nature of the legal and policy issues raised in this article, we take into account the laws of several different jurisdictions, especially the US, the UK and the European Union (EU).Is an AI-created work protected by copyright?Should it be?Who would be viewed as the author?Who should own such a work?These questions are the focus of the first part of this article: "AI as producer".In the second part, "AI as consumer", our analysis shifts to the questions of whether and to what extent the use of data fed into the system, for example to train the algorithms, may amount to copyright infringement, or may in certain circumstances be exempted under fair use, fair dealing or similar doctrines.In the third part we consider possible legal regimes for dealing with machine produced works including alternatives to copyright, such as a public domain scenario and a sui generis system.Finally, we provide our conclusions. Artificial intelligence as producerBy any rational measure AI systems can be said to produce works creatively.Moreover, they do so in an accurate, logical and independent way, with the final output often consisting of something unpredictable to the humans who programmed the AI platform in the first place (as well as the user(s) of the AI). 21 As with much content produced by humans, it is often the case that a key element of the creative process in AI-enabled works is based on randomness. 22So, is the final output generated by a machine protected by copyright-and, if so, who would be seen as its author (and owner)? This is not a new questionThe issue of whether computational creativity can be protected by copyright is not actually a new one, at least in the US.As far back as 1965, the US Register of Copyrights reported concerns to Congress about the rise of computer technology and wondered if and where the line between human authorship and computer production should be drawn. 23 More than a decade later, in 1978, the US National Commission on New Technological Uses of Copyrighted Works (CONTU Commission) reported on the issue and concluded that computers used to produce works were just "inert tools of creation".The CONTU Commission remarked: "[t]his discussion may have stemmed from a concern that computers either had or were likely to soon achieve powers that would enable them independently to create works that, although similar to other copyrightable works, would not or should not be copyrightable because they had no human author.The development of this capacity for 'artificial intelligence' has not yet come to pass, and, indeed, it has been suggested that such development is too speculative to consider at this time."(emphasis added) 24What the CONTU Commission considered too speculative in late 1970s became commonplace only a few years later.Advances in computing technology prompted the US Congress Office of Technology Assessment (OTA) to issue a report in 1986 arguing that CONTU's prior approach had been too simplistic and that computer programs were more than mere "inert tools of creation". 25Uncertainty still reigned, however, as in the same report OTA recognised that "we know that these works would be copyrightable if they were done by people, but we don't know what to do with them if they're done by computers". 26The developments in computational-and more recently artificial intelligence and machine learning-technologies in the subsequent decades have made the issues related to copyright in works created by machines more pressing.Academic interest in this topic has soared. 27Many of these academic scholars are interested in the question we now turn to: who can be an author?

56The Protection of Intellectual Property Rights for the Xinqiao Yanglan Bamboo Weaving Craft in Zhaoqing City Empowered by Artificial IntelligenceOpenAlex

Junyan Chen
Artificial intelligence (AI) technology is bringing about a revolutionary paradigm shift in the safeguarding and transmission of intangible cultural heritage (ICH). While it enhances conservation efficiency and expands transmission pathways, it simultaneously poses structural challenges to the intellectual property (IP) system, which is traditionally based on the protection of individual intellectual creations. This paper takes the Xinqiao Yanglan Bamboo Weaving Craft in Gaoyao District, Zhaoqing City, Guangdong Province, as a specific case study. It delves into the frontier legal dilemmas concerning the definition of rights holders, the assessment of copyrightability, the delineation of fair use boundaries, and the determination of infringement liability for digital ICH outcomes in the context of AI empowerment. The research indicates that the inherent tension between public and private interests becomes more complex with AI's involvement. Consequently, this paper proposes clarifying the ownership rules for AI-generated content through legislation or judicial interpretation, constructing a special statutory licensing system tailored for the digital utilization of ICH, enhancing the flexibility of fair use doctrines to accommodate technological advancements, and comprehensively utilizing technologies like blockchain to build a credible system for rights confirmation, utilization, and protection. The aim is to construct an IP legal framework that effectively incentivizes innovation, safeguards the legitimate rights and interests of all parties, and promotes the widespread dissemination and sustainable development of ICH knowledge, providing theoretical support and pathway guidance for the living inheritance of ICH in the digital age.

57Generative AI, Plagiarism, and Copyright Infringement in Legal DocumentsOpenAlex

Amy Cyphert

58AI Art Neural Constellation: Revealing the Collective and Contrastive State of AI-Generated and Human ArtOpenAlex

Faizan Farooq Khan, Diana Kim, Divyansh Jha, et al.
Discovering the creative potentials of a random signal to various artistic expressions in aesthetic and conceptual richness is a ground for the recent success of generative machine learning as a way of art creation. To understand the new artistic medium better, in this work, we comprehensively analyze AI-generated art within the context of human art heritage using our dataset, "ArtConstellation," comprising annotations for 6,000 WikiArt and 3,200 AI-generated artworks. After training various generative models, we compare the produced art samples with WikiArt data using the last hidden layer of a deep-CNN trained for style classification. By interpreting neural representations with important artistic concepts like Wölfflin’s principles, we find that AI-generated artworks align with modern period art concepts (1800 - 2000). Out-Of-Distribution (OOD) and In-Distribution (ID) detection in CLIP space reveal that AI-generated art is ID to human art with landscapes and geometric abstract figures but OOD with deformed and twisted figures, showcasing unique characteristics. A human survey on emotional experience indicates color composition and familiar subjects as key factors in likability and emotions. We introduce our methodologies and dataset, "ArtNeural-Constellation," as a framework for contrasting human and AI-generated art. Code and data are available here.

59An Intellectual Property Protection Platform Based on Blockchain TechnologyOpenAlex

Fei Miao
In order to solve the problems in the traditional field of intellectual property protection, such as difficulty in property rights confirmation, infringement monitoring, and evidence collection, The paper proposes an intellectual property protection platform based on blockchain technology, mainly introducing the overall architecture design, implementation goals, and the main characteristics of the proposed framework; According to practical application requirements, the framework modules are divided into three main modules: user digital content upload, secure distributed storage, and digital content infringement detection. The content involved in each module is explained; Introduced the copyright infringement detection process for digital content and provided a specific description of the steps involved. By building relevant experimental environments and deploying blockchain and IPFS platforms, a prototype system for digital copyright certification and access control was developed. The experimental results indicate that semantic factors were considered and incorporated into the text, with TF-IDF weights adjusted based on multiple features of words (word length, part of speech, occurrence position, and topic relevance), resulting in a 0.145 increase in F1 mean and an overall 0.425 improvement compared to the original SimHash algorithm. This paper improves the traditional SimHash algorithm by only considering the single feature of words in word weight calculation, selecting keywords with the top 50% weight values to generate text fingerprints, and then comparing the similarity between texts through Hamming distance. Additionally, the final experimental results also prove the effectiveness of this method.

60Convolutional Networks for Real-Time Trademark Violation IdentificationOpenAlex

Ashu Nayak, Pooja Sharma
As an important part of the intellectual property rights protection, copyright infringement detection in the information age, is faced with threats monitored target for logos, images, social media platform trademarks. With the increase of user-generated process being located on the web, it is important to develop automatized systems allowing to identify trademarks violations in real time in order to ensure brand identity is preserved and that trademarks are not infringed upon; Traditional systems to detect trademarks are for the most part keyword-based systems or manual search, however this timeconsuming, has high-margin error and does not detect complex or slight modifications of the trademark. We tackle the problem of real-time detection of trademark violations and we propose a new approach based on Convolutional Neural Networks, a powerful class of deep learning algorithms that achieve high accuracy in image recognition problems. These imagegeneration and real-time features extraction step provided a base that guided for the newly proposed method which uses the power of CNNs in analyzing an image at every single timestamp for potentially matching images to learn the products that may have similarities to a flagged image for trademark violation. A deep CNN model was established and trained on a large dataset of branded logos and marks. Since the model has been trained to “see” images merely based on visual features (and not text), it can detect even minor differences in trademarks, whether that be color or distortion or shape. And this is especially important in cases of trademark infringement, where the infringer may try to take every effort to change the original trademark enough to evade getting caught. Our CNN model is built with several layers which learn more and more abstract features from images and is therefore fitting for this task as it is able to extract critical data to guarantee that the system is operating properly in an efficient manner. At the early layers, the model learns to identify simple features, like edges and textures, but eventually learns to correlate with complex features that may characterize the uniqueness of a trademark. One of the key advantages of CNNs in this type of applications is their ability to generalize — enabling the model to potentially detect violations in low-res images or noisy images, as it often the case with images processed in real-time. To accomplish real-time detection, the CNN model is docked to a streaming pipeline for dynamic images, and video processing. This means that even on multiple such as social media, ecommerce websites and more there will be constant vigilance of the content for trademark violations. It will provide brand owners with real-time alerts of in-progress infringements and enable swift enforcement actions such as takedown of infringing content and litigation.

61Copyright Infringement in AI-Generated ArtworksOpenAlex

Jessica Gillotte
This Note examines potential copyright infringement issues arising from AI-generated artwork and argues that, under current copyright law, an engineer may use copyrighted works to train an AI program to generate artwork without incurring infringement liability.

62Who Is Responsible for AI Copyright Infringement?OpenAlex

Michael Goodyear

63Generative AI Art: Copyright Infringement and Fair UseOpenAlex

Michael D. Murray
The discussion of AI copyright infringement or fair use often skips over all the required steps of the infringement analysis in order to focus on the most intriguing question, “Could a visual generative AI generate a work that potentially infringes a preexisting copyrighted work?” and then the discussion skips further ahead to, “Would the AI have a fair use defense, most likely under the transformative test?” These are relevant questions, but without considering the actual steps of the copyright infringement analysis, the discussion is misleading or even irrelevant. This neglecting of topics and stages of the infringement analysis fails to direct our attention to a properly accused party or entity whose actions prompt the question. Making a sudden transition from a question of infringement in the creation of training datasets to the creation of foundation models that draw from the training data to the actual operation of the generative AI system to produce images makes a false equivalency regarding the processes themselves and the persons responsible for them. The questions ought to shift focus from the persons compiling the training dataset used to train the AI system and the designers and creators of the AI system itself to the end users of the AI system who conceive of and cause the creation of images. The analysis of infringement or fair use in the generative AI context has suffered from widespread misunderstanding concerning the generative AI processes and the control and authorship of the end-user. Claimants, commentators, and regulators have made incorrect assumptions and inaccurate simplifications concerning the process, which I refer to as the Magic File Drawer theory, the Magic Copy Machine theory, and the Magic Box Artist theory. These theories, if they were true, would be much easier to envision and understand than the actual science and technology that goes into the creation and operation of a contemporary visual generative AI system. Throughout this Article, I will attempt to clarify and correct the understanding of the science and technology of the generative AI processes and explain the different roles of the training dataset designers, the generative AI system designers, and the end-users in the rendering of visual works by a generative AI system. Part II will discuss the requirements of a claim of copyright infringement including each step from the copyrightability of the claimant’s work, the doctrines that limit copyrightability, the requirement of an act of copying, and the infringement elements. Part III will summarize the copyright fair use test paying particular attention to the purpose and character of the use analysis, 17 U.S.C. § 107(1), and the current interpretation of the “transformative” test after Andy Warhol Foundation v. Goldsmith, particularly in circumstances relating to technology and the use of copyrighted or copyrightable data sources. Part IV will analyze potential infringement or fair use by the creators of generative AI training datasets. Part V will analyze potential infringement or fair use by the creators of visual generative AI systems. Part VI will analyze potential infringement or fair use by the end-users of visual generative AI systems. For all their complexity, visual generative AI systems are tools that depend on an end-user who conceives of and designs the image and provides the system with a prompt to set the generative process in motion. The end-users are responsible for crafting the prompt or series of prompts used, for evaluating the outputs of the generative AI, for adjusting and editing the iterations of images offered by the AI system, and ultimately for selecting and adopting one of the images generated by the AI as the final image. The end-users then make further decisions about the actual use and its function and purpose for the images the end-users selected and adopted from the outputs of the AI. While working with the AI tool to try to produce a certain image, an end-user might steer the system to produce a work that could, under an infringement analysis, be regarded as potentially infringing, which would lead us again to the fair use analysis based on the end-user’s use of the image.

64Sounds of Science: Copyright Infringement in AI Music Generator OutputsOpenAlex

Eric Sunray
The music business is no stranger to disruptive technology. The industry’s apparent comeback from the devastating downturn caused by illegal file sharing seems to have arrived just in time for what may be an even more disruptive technological phenomenon: artificial intelligence (“AI”). Much has been said about the implications of AI-generated music, ranging from issues of ownership, to rights of publicity. However, there has been surprisingly little discussion of infringement in the AI systems’ outputs. By examining the functionality of AI music generators through the lens of de minimis use case law, this paper will explain how the outputs of AI music generators potentially infringe the exclusive reproduction right granted to musical work and sound recording copyright owners. Going forward, courts and policymakers must not ignore AI’s capacity to undermine our incentives for human authorship, and craft rules that promote a mutually beneficial AI music ecosystem for technology companies and copyright owners alike.

65Regulating ChatGPT and other Large Generative AI ModelsOpenAlex

Philipp Hacker, Andreas Engel, Marco Mauer
Large generative AI models (LGAIMs), such as ChatGPT, GPT-4 or Stable Diffusion, are rapidly transforming the way we communicate, illustrate, and create. However, AI regulation, in the EU and beyond, has primarily focused on conventional AI models, not LGAIMs. This paper will situate these new generative models in the current debate on trustworthy AI regulation, and ask how the law can be tailored to their capabilities. After laying technical foundations, the legal part of the paper proceeds in four steps, covering (1) direct regulation, (2) data protection, (3) content moderation, and (4) policy proposals. It suggests a novel terminology to capture the AI value chain in LGAIM settings by differentiating between LGAIM developers, deployers, professional and non-professional users, as well as recipients of LGAIM output. We tailor regulatory duties to these different actors along the value chain and suggest strategies to ensure that LGAIMs are trustworthy and deployed for the benefit of society at large. Rules in the AI Act and other direct regulation must match the specificities of pre-trained models. The paper argues for three layers of obligations concerning LGAIMs (minimum standards for all LGAIMs; high-risk obligations for high-risk use cases; collaborations along the AI value chain). In general, regulation should focus on concrete high-risk applications, and not the pre-trained model itself, and should include (i) obligations regarding transparency and (ii) risk management. Non-discrimination provisions (iii) may, however, apply to LGAIM developers. Lastly, (iv) the core of the DSA's content moderation rules should be expanded to cover LGAIMs. This includes notice and action mechanisms, and trusted flaggers.

66Interpretable machine learningOpenAlex

Parliamentary Office of Science and Technology, Lorna Christie
Machine learning (ML, a type of artificial intelligence) is increasingly being used to support decision making in a variety of applications including recruitment and clinical diagnoses. While ML has many advantages, there are concerns that in some cases it may not be possible to explain completely how its outputs have been produced. This POSTnote gives an overview of ML and its role in decision-making. It examines the challenges of understanding how a complex ML system has reached its output, and some of the technical approaches to making ML easier to interpret. It also gives a brief overview of some of the proposed tools for making ML systems more accountable.

67Fairness and Abstraction in Sociotechnical SystemsOpenAlex

Andrew D. Selbst, danah boyd, Sorelle A. Friedler, et al.
A key goal of the fair-ML community is to develop machine-learning based systems that, once introduced into a social context, can achieve social and legal outcomes such as fairness, justice, and due process. Bedrock concepts in computer science---such as abstraction and modular design---are used to define notions of fairness and discrimination, to produce fairness-aware learning algorithms, and to intervene at different stages of a decision-making pipeline to produce "fair" outcomes. In this paper, however, we contend that these concepts render technical interventions ineffective, inaccurate, and sometimes dangerously misguided when they enter the societal context that surrounds decision-making systems. We outline this mismatch with five "traps" that fair-ML work can fall into even as it attempts to be more context-aware in comparison to traditional data science. We draw on studies of sociotechnical systems in Science and Technology Studies to explain why such traps occur and how to avoid them. Finally, we suggest ways in which technical designers can mitigate the traps through a refocusing of design in terms of process rather than solutions, and by drawing abstraction boundaries to include social actors rather than purely technical ones.

68Generative AI for Ethical and Bias-Free Content ModerationOpenAlex

Amisha Subhashrao Bhasme.
The growth of online platforms has led to an increase in harmful content, such as hate speech, fake news, and explicit images. While traditional content moderation techniques are human-centric, they struggle to scale effectively. Generative AI presents an opportunity to automate and enhance content moderation, offering efficiency at scale. However, generative AI models must be designed to detect harmful content while ensuring fairness and ethical behavior, avoiding biases and over-censorship. This paper explores the challenges of using generative AI for content moderation, focusing on bias detection, fairness frameworks, and solutions to prevent harm.

69Content moderation, AI, and the question of scaleOpenAlex

Tarleton Gillespie
AI seems like the perfect response to the growing challenges of content moderation on social media platforms: the immense scale of the data, the relentlessness of the violations, and the need for human judgments without wanting humans to have to make them. The push toward automated content moderation is often justified as a necessary response to the scale: the enormity of social media platforms like Facebook and YouTube stands as the reason why AI approaches are desirable, even inevitable. But even if we could effectively automate content moderation, it is not clear that we should.

70"How advertiser-friendly is my video?": YouTuber's Socioeconomic Interactions with Algorithmic Content ModerationOpenAlex

Renkai Ma, Yubo Kou
To manage user-generated harmful video content, YouTube relies on AI algorithms (e.g., machine learning) in content moderation and follows a retributive justice logic to punish convicted YouTubers through demonetization, a penalty that limits or deprives them of advertisements (ads), reducing their future ad income. Moderation research is burgeoning in CSCW, but relatively little attention has been paid to the socioeconomic implications of YouTube's algorithmic moderation. Drawing from the lens of algorithmic labor, we describe how algorithmic moderation shapes YouTubers' labor conditions through algorithmic opacity and precarity. YouTubers coped with such challenges from algorithmic moderation by sharing and applying practical knowledge they learned about moderation algorithms. By analyzing video content creation as algorithmic labor, we unpack the socioeconomic implications of algorithmic moderation and point to necessary post-punishment support as a form of restorative justice. Lastly, we put forward design considerations for algorithmic moderation systems.

71Ethical implications of AI in the MetaverseOpenAlex

Alesia Zhuk
Abstract This paper delves into the ethical implications of AI in the Metaverse through the analysis of real-world case studies, including Horizon Worlds, Decentraland, Roblox, Sansar, and Rec Room. The examination reveals recurring concerns related to content moderation, emphasising the need for a human-AI hybrid approach to strike a balance between creative freedom and user safety. Privacy and data protection emerge as crucial considerations, highlighting the importance of transparent communication and user data control for responsible AI implementation. Additionally, promoting inclusivity and diversity is emphasised, calling for transparent governance, diverse representation, and collaboration with ethics experts to ensure equitable AI practices. By addressing these specific ethical challenges, we can pave the way towards a responsible and user-centric Metaverse, maximising its potential while safeguarding user well-being and rights.

72Exploring the impacts of artificial intelligence on freedom of religion or belief onlineOpenAlex

Cameran Ashraf
Freedom of religion or belief is an essential right for building pluralistic and tolerant societies which can sustain a multiplicity of competing ideas. However, the opaqueness of artificial intelligence systems on the Internet represents a challenge to the protection and enjoyment of this and other human rights. Although AI has generated interest in the human rights literature, these studies have largely focused on AI and its impact on freedom of expression and privacy, leaving other rights such as freedom of religion or belief neglected. As part of a broader research project to expand the academic conversation about AI and human rights, this paper will examine the impact of artificial intelligence on freedom of religion or belief online. The paper will focus on the worship, teaching, observance, and practice associated with freedom of religion or belief alongside the impacts of AI in content display, content moderation, and online privacy. The paper will offer preliminary policy recommendations to encourage discussion on policy approaches to AI development and deployment which incorporate protections for freedom of religion or belief in the era of artificial intelligence.

73<p>Legal, Ethical and Practical Challenges of AI-Driven Content Moderation</p>OpenAlex

Pablo Rafael Banchio

74Who Audits the Auditors? Recommendations from a field scan of the algorithmic auditing ecosystemOpenAlex

Sasha Costanza-Chock, Inioluwa Deborah Raji, Joy Buolamwini
Algorithmic audits (or ‘AI audits’) are an increasingly popular mechanism for algorithmic accountability; however, they remain poorly defined. Without a clear understanding of audit practices, let alone widely used standards or regulatory guidance, claims that an AI product or system has been audited, whether by first-, second-, or third-party auditors, are difficult to verify and may potentially exacerbate, rather than mitigate, bias and harm. To address this knowledge gap, we provide the first comprehensive field scan of the AI audit ecosystem. We share a catalog of individuals (N=438) and organizations (N=189) who engage in algorithmic audits or whose work is directly relevant to algorithmic audits; conduct an anonymous survey of the group (N=152); and interview industry leaders (N=10). We identify emerging best practices as well as methods and tools that are becoming commonplace, and enumerate common barriers to leveraging algorithmic audits as effective accountability mechanisms. We outline policy recommendations to improve the quality and impact of these audits, and highlight proposals with wide support from algorithmic auditors as well as areas of debate. Our recommendations have implications for lawmakers, regulators, internal company policymakers, and standards-setting bodies, as well as for auditors. They are: 1) require the owners and operators of AI systems to engage in independent algorithmic audits against clearly defined standards; 2) notify individuals when they are subject to algorithmic decision-making systems; 3) mandate disclosure of key components of audit findings for peer review; 4) consider real-world harm in the audit process, including through standardized harm incident reporting and response mechanisms; 5) directly involve the stakeholders most likely to be harmed by AI systems in the algorithmic audit process; and 6) formalize evaluation and, potentially, accreditation of algorithmic auditors.

75Outsider Oversight: Designing a Third Party Audit Ecosystem for AI GovernanceOpenAlex

Inioluwa Deborah Raji, Peggy Xu, Colleen Honigsberg, et al.
Much attention has focused on algorithmic audits and impact assessments to hold developers and users of algorithmic systems accountable. But existing algorithmic accountability policy approaches have neglected the lessons from non-algorithmic domains: notably, the importance of third parties. Our paper synthesizes lessons from other fields on how to craft effective systems of external oversight for algorithmic deployments. First, we discuss the challenges of third party oversight in the current AI landscape. Second, we survey audit systems across domains - e.g., financial, environmental, and health regulation - and show that the institutional design of such audits are far from monolithic. Finally, we survey the evidence base around these design components and spell out the implications for algorithmic auditing. We conclude that the turn toward audits alone is unlikely to achieve actual algorithmic accountability, and sustained focus on institutional design will be required for meaningful third party involvement.

76Trust Theory: A Socio-Cognitive and Computational ModelOpenAlex

Christiano Castelfranchi, Rino Falcone
This book provides an introduction, discussion, and formal-based modelling of trust theory and its applications in agent-based systems This book gives an accessible explanation of the importance of trust in human interaction and, in general, in autonomous cognitive agents including autonomous technologies. The authors explain the concepts of trust, and describe a principled, general theory of trust grounded on cognitive, cultural, institutional, technical, and normative solutions. This provides a strong base for the authors discussion of role of trust in agent-based systems supporting human-computer interaction and distributed and virtual organizations or markets (multi-agent systems). Key Features: Provides an accessible introduction to trust, and its importance and applications in agent-based systems Proposes a principled, general theory of trust grounding on cognitive, cultural, institutional, technical, and normative solutions. Offers a clear, intuitive approach, and systematic integration of relevant issues Explains the dynamics of trust, and the relationship between trust and security Offers operational definitions and models directly applicable both in technical and experimental domains Includes a critical examination of trust models in economics, philosophy, psychology, sociology, and AI This book will be a valuable reference for researchers and advanced students focused on information and communication technologies (computer science, artificial intelligence, organizational sciences, and knowledge management etc.), as well as Web-site and robotics designers, and for scholars working on human, social, and cultural aspects of technology. Professionals of ecommerce systems and peer-to-peer systems will also find this text of interest.

77Artificial Intelligence (AI) Trust Framework and Maturity Model: Applying an Entropy Lens to Improve Security, Privacy, and Ethical AIOpenAlex

Michael Mylrea, Nikki Robinson
Recent advancements in artificial intelligence (AI) technology have raised concerns about the ethical, moral, and legal safeguards. There is a pressing need to improve metrics for assessing security and privacy of AI systems and to manage AI technology in a more ethical manner. To address these challenges, an AI Trust Framework and Maturity Model is proposed to enhance trust in the design and management of AI systems. Trust in AI involves an agreed-upon understanding between humans and machines about system performance. The framework utilizes an "entropy lens" to root the study in information theory and enhance transparency and trust in "black box" AI systems, which lack ethical guardrails. High entropy in AI systems can decrease human trust, particularly in uncertain and competitive environments. The research draws inspiration from entropy studies to improve trust and performance in autonomous human-machine teams and systems, including interconnected elements in hierarchical systems. Applying this lens to improve trust in AI also highlights new opportunities to optimize performance in teams. Two use cases are described to validate the AI framework's ability to measure trust in the design and management of AI systems.

78An empirical investigation of trust in AI in a Chinese petrochemical enterprise based on institutional theoryOpenAlex

Li Jia, Yiwen Zhou, Junping Yao, et al.
Despite its considerable potential in the manufacturing industry, the application of artificial intelligence (AI) in the industry still faces the challenge of insufficient trust. Since AI is a black box with operations that ordinary users have difficulty understanding, users in organizations rely on institutional cues to make decisions about their trust in AI. Therefore, this study investigates trust in AI in the manufacturing industry from an institutional perspective. We identify three institutional dimensions from institutional theory and conceptualize them as management commitment (regulative dimension at the organizational level), authoritarian leadership (normative dimension at the group level), and trust in the AI promoter (cognitive dimension at the individual level). We hypothesize that all three institutional dimensions have positive effects on trust in AI. In addition, we propose hypotheses regarding the moderating effects of AI self-efficacy on these three institutional dimensions. A survey was conducted in a large petrochemical enterprise in eastern China just after the company had launched an AI-based diagnostics system for fault detection and isolation in process equipment service. The results indicate that management commitment, authoritarian leadership, and trust in the AI promoter are all positively related to trust in AI. Moreover, the effect of management commitment and trust in the AI promoter are strengthened when users have high AI self-efficacy. The findings of this study provide suggestions for academics and managers with respect to promoting users' trust in AI in the manufacturing industry.

79Toward an Ecologically Valid Conceptual Framework for the Use of Artificial Intelligence in Clinical Settings: Need for Systems Thinking, Accountability, Decision-making, Trust, and Patient Safety Considerations in Safeguarding the Technology and CliniciansOpenAlex

Avishek Choudhury
The health care management and the medical practitioner literature lack a descriptive conceptual framework for understanding the dynamic and complex interactions between clinicians and artificial intelligence (AI) systems. As most of the existing literature has been investigating AI's performance and effectiveness from a statistical (analytical) standpoint, there is a lack of studies ensuring AI's ecological validity. In this study, we derived a framework that focuses explicitly on the interaction between AI and clinicians. The proposed framework builds upon well-established human factors models such as the technology acceptance model and expectancy theory. The framework can be used to perform quantitative and qualitative analyses (mixed methods) to capture how clinician-AI interactions may vary based on human factors such as expectancy, workload, trust, cognitive variables related to absorptive capacity and bounded rationality, and concerns for patient safety. If leveraged, the proposed framework can help to identify factors influencing clinicians' intention to use AI and, consequently, improve AI acceptance and address the lack of AI accountability while safeguarding the patients, clinicians, and AI technology. Overall, this paper discusses the concepts, propositions, and assumptions of the multidisciplinary decision-making literature, constituting a sociocognitive approach that extends the theories of distributed cognition and, thus, will account for the ecological validity of AI.

80Content Moderation on Social Media: Does It Matter Who and Why Moderates Hate Speech?PubMed

Sai Wang, Ki Joon Kim
Cyberpsychol Behav Soc Netw. 2023 Jul;26(7):527-534. doi: 10.1089/cyber.2022.0158. Epub 2023 May 3.
Artificial intelligence (AI) has been increasingly integrated into content moderation to detect and remove hate speech on social media. An online experiment ( = 478) was conducted to examine how moderation agents (AI vs. human vs. human-AI collaboration) and removal explanations (with vs. without) affect users' perceptions and acceptance of removal decisions for hate speech targeting social groups with certain characteristics, such as religion or sexual orientation. The results showed that individuals exhibit consistent levels of perceived trustworthiness and acceptance of removal decisions regardless of the type of moderation agent. When explanations for the content takedown were provided, removal decisions made jointly by humans and AI were perceived as more trustworthy than the same decisions made by humans alone, which increased users' willingness to accept the verdict. However, this moderated mediation effect was only significant when Muslims, not homosexuals, were the target of hate speech.

81Putting AI ethics to work: are the tools fit for purpose?OpenAlex

Jacqui Ayling, Adriane Chapman
Abstract Bias, unfairness and lack of transparency and accountability in Artificial Intelligence (AI) systems, and the potential for the misuse of predictive models for decision-making have raised concerns about the ethical impact and unintended consequences of new technologies for society across every sector where data-driven innovation is taking place. This paper reviews the landscape of suggested ethical frameworks with a focus on those which go beyond high-level statements of principles and offer practical tools for application of these principles in the production and deployment of systems. This work provides an assessment of these practical frameworks with the lens of known best practices for impact assessment and audit of technology. We review other historical uses of risk assessments and audits and create a typology that allows us to compare current AI ethics tools to Best Practices found in previous methodologies from technology, environment, privacy, finance and engineering. We analyse current AI ethics tools and their support for diverse stakeholders and components of the AI development and deployment lifecycle as well as the types of tools used to facilitate use. From this, we identify gaps in current AI ethics tools in auditing and risk assessment that should be considered going forward.

82The state of the art development of AHP (1979–2017): a literature review with a social network analysisOpenAlex

Ali Emrouznejad, Marianna Marra
Although many papers describe the evolution of the analytic hierarchy process (AHP), most adopt a subjective approach. This paper examines the pattern of development of the AHP research field using social network analysis and scientometrics, and identifies its intellectual structure. The objectives are: (i) to trace the pattern of development of AHP research; (ii) to identify the patterns of collaboration among authors; (iii) to identify the most important papers underpinning the development of AHP; and (iv) to discover recent areas of interest. We analyse two types of networks: social networks, that is, co-authorship networks, and cognitive mapping or the network of disciplines affected by AHP. Our analyses are based on 8441 papers published between 1979 and 2017, retrieved from the ISI Web of Science database. To provide a longitudinal perspective on the pattern of evolution of AHP, we analyse these two types of networks during the three periods 1979–1990, 1991–2001 and 2002–2017. We provide some basic statistics on AHP journals and researchers, review the main topics and applications of integrated AHPs and provide direction for future research by highlighting some open questions.

83Dissecting racial bias in an algorithm used to manage the health of populationsOpenAlex

Ziad Obermeyer, Brian W. Powers, Christine Vogeli, et al.
Racial bias in health algorithms The U.S. health care system uses commercial algorithms to guide health decisions. Obermeyer et al. find evidence of racial bias in one widely used algorithm, such that Black patients assigned the same level of risk by the algorithm are sicker than White patients (see the Perspective by Benjamin). The authors estimated that this racial bias reduces the number of Black patients identified for extra care by more than half. Bias occurs because the algorithm uses health costs as a proxy for health needs. Less money is spent on Black patients who have the same level of need, and the algorithm thus falsely concludes that Black patients are healthier than equally sick White patients. Reformulating the algorithm so that it no longer uses costs as a proxy for needs eliminates the racial bias in predicting who needs extra care. Science , this issue p. 447 ; see also p. 421

84Lack of Transparency and Potential Bias in Artificial Intelligence Data Sets and AlgorithmsOpenAlex

Roxana Daneshjou, Mary P. Smith, Mary Sun, et al.
IMPORTANCE: Clinical artificial intelligence (AI) algorithms have the potential to improve clinical care, but fair, generalizable algorithms depend on the clinical data on which they are trained and tested. OBJECTIVE: To assess whether data sets used for training diagnostic AI algorithms addressing skin disease are adequately described and to identify potential sources of bias in these data sets. DATA SOURCES: In this scoping review, PubMed was used to search for peer-reviewed research articles published between January 1, 2015, and November 1, 2020, with the following paired search terms: deep learning and dermatology, artificial intelligence and dermatology, deep learning and dermatologist, and artificial intelligence and dermatologist. STUDY SELECTION: Studies that developed or tested an existing deep learning algorithm for triage, diagnosis, or monitoring using clinical or dermoscopic images of skin disease were selected, and the articles were independently reviewed by 2 investigators to verify that they met selection criteria. CONSENSUS PROCESS: Data set audit criteria were determined by consensus of all authors after reviewing existing literature to highlight data set transparency and sources of bias. RESULTS: A total of 70 unique studies were included. Among these studies, 1 065 291 images were used to develop or test AI algorithms, of which only 257 372 (24.2%) were publicly available. Only 14 studies (20.0%) included descriptions of patient ethnicity or race in at least 1 data set used. Only 7 studies (10.0%) included any information about skin tone in at least 1 data set used. Thirty-six of the 56 studies developing new AI algorithms for cutaneous malignant neoplasms (64.3%) met the gold standard criteria for disease labeling. Public data sets were cited more often than private data sets, suggesting that public data sets contribute more to new development and benchmarks. CONCLUSIONS AND RELEVANCE: This scoping review identified 3 issues in data sets that are used to develop and test clinical AI algorithms for skin disease that should be addressed before clinical translation: (1) sparsity of data set characterization and lack of transparency, (2) nonstandard and unverified disease labels, and (3) inability to fully assess patient diversity used for algorithm development and testing.

85Balancing Innovation and Regulation in the Age of Generative Artificial IntelligenceOpenAlex

Xukang Wang, Ying Cheng Wu
Abstract The emergence of generative artificial intelligence (AI), exemplified by models like ChatGPT, presents both opportunities and challenges. As these technologies become increasingly integrated into various aspects of society, the need for a harmonized legal framework to address the associated risks becomes crucial. This article presents a comprehensive analysis of the disruptive impact of generative AI, the legal risks of AI-generated content, and the governance strategies needed to strike a balance between innovation and regulation. Employing a three-pronged methodology—literature review, doctrinal legal analysis, and case study integration—the study examines the current legal landscape; synthesizes scholarly works on the technological, ethical, and socioeconomic implications of generative AI; and illustrates practical challenges through real-world case studies. The article assesses the strengths and limitations of US governance strategies for AI and proposes a harmonized legal framework emphasizing international collaboration, proactive legislation, and the establishment of a dedicated regulatory body. By engaging diverse stakeholders and identifying critical gaps in current research, the study contributes to the development of a legal framework that upholds ethical principles, protects individual rights, and fosters responsible innovation in the age of generative AI.

86The ethics of ChatGPT – Exploring the ethical issues of an emerging technologyOpenAlex

Bernd Carsten Stahl, Damian Eke
This article explores ethical issues raised by generative conversational AI systems like ChatGPT. It applies established approaches for analysing ethics of emerging technologies to undertake a systematic review of possible benefits and concerns. The methodology combines ethical issues identified by Anticipatory Technology Ethics, Ethical Impact Assessment, and Ethical Issues of Emerging ICT Applications with AI-specific issues from the literature. These are applied to analyse ChatGPT's capabilities to produce humanlike text and interact seamlessly. The analysis finds ChatGPT could provide high-level societal and ethical benefits. However, it also raises significant ethical concerns across social justice, individual autonomy, cultural identity, and environmental issues. Key high-impact concerns include responsibility, inclusion, social cohesion, autonomy, safety, bias, accountability, and environmental impacts. While the current discourse focuses narrowly on specific issues such as authorship, this analysis systematically uncovers a broader, more balanced range of ethical issues worthy of attention. Findings are consistent with emerging research and industry priorities on ethics of generative AI. Implications include the need for diverse stakeholder engagement, considering benefits and risks holistically when developing applications, and multi-level policy interventions to promote positive outcomes. Overall, the analysis demonstrates that applying established ethics of technology methodologies can produce a rigorous, comprehensive foundation to guide discourse and action around impactful emerging technologies like ChatGPT. The paper advocates sustaining this broad, balanced ethics perspective as use cases unfold to realize benefits while addressing ethical downsides.

87China's Interim Measures on generative AI: Origin, content and significanceOpenAlex

Sara Migliorini

88Attention is not all you need: the complicated case of ethically using large language models in healthcare and medicineOpenAlex

Stefan Harrer
Large Language Models (LLMs) are a key component of generative artificial intelligence (AI) applications for creating new content including text, imagery, audio, code, and videos in response to textual instructions. Without human oversight, guidance and responsible design and operation, such generative AI applications will remain a party trick with substantial potential for creating and spreading misinformation or harmful and inaccurate content at unprecedented scale. However, if positioned and developed responsibly as companions to humans augmenting but not replacing their role in decision making, knowledge retrieval and other cognitive processes, they could evolve into highly efficient, trustworthy, assistive tools for information management. This perspective describes how such tools could transform data management workflows in healthcare and medicine, explains how the underlying technology works, provides an assessment of risks and limitations, and proposes an ethical, technical, and cultural framework for responsible design, development, and deployment. It seeks to incentivise users, developers, providers, and regulators of generative AI that utilises LLMs to collectively prepare for the transformational role this technology could play in evidence-based sectors.

89The Artificial Intelligence Assessment Scale (AIAS): A Framework for Ethical Integration of Generative AI in Educational AssessmentOpenAlex

Mike Perkins, Leon Furze, Jasper Roe, et al.
Recent developments in Generative Artificial Intelligence (GenAI) have created a paradigm shift in multiple areas of society, and the use of these technologies is likely to become a defining feature of education in coming decades. GenAI offers transformative pedagogical opportunities, while simultaneously posing ethical and academic challenges. Against this backdrop, we outline a practical, simple, and sufficiently comprehensive tool to allow for the integration of GenAI tools into educational assessment: the AI Assessment Scale (AIAS). The AIAS empowers educators to select the appropriate level of GenAI usage in assessments based on the learning outcomes they seek to address. The AIAS offers greater clarity and transparency for students and educators, provides a fair and equitable policy tool for institutions to work with, and offers a nuanced approach which embraces the opportunities of GenAI while recognising that there are instances where such tools may not be pedagogically appropriate or necessary. By adopting a practical, flexible approach that can be implemented quickly, the AIAS can form a much-needed starting point to address the current uncertainty and anxiety regarding GenAI in education. As a secondary objective, we engage with the current literature and advocate for a refocused discourse on GenAI tools in education, one which foregrounds how technologies can help support and enhance teaching and learning, which contrasts with the current focus on GenAI as a facilitator of academic misconduct.

90The role of ChatGPT in higher education: Benefits, challenges, and future research directionsOpenAlex

Tareq Rasul, Sumesh Nair, Diane Robyn Kalendra, et al.
This paper examines the potential benefits and challenges of using the generative AI model, ChatGPT, in higher education, in the backdrop of the constructivist theory of learning. This perspective-type study presents five benefits of ChatGPT: the potential to facilitate adaptive learning, provide personalised feedback, support research and data analysis, offer automated administrative services, and aid in developing innovative assessments. Additionally, the paper identifies five challenges: academic integrity concerns, reliability issues, inability to evaluate and reinforce graduate skill sets, limitations in assessing learning outcomes, and potential biases and falsified information in information processing. The paper argues that tertiary educators and students must exercise caution when using ChatGPT for academic purposes to ensure its ethical, reliable, and effective use. To achieve this, the paper proposes various propositions, such as prioritising education on the responsible and ethical use of ChatGPT, devising new assessment strategies, addressing bias and falsified information, and including AI literacy as part of graduate skills. By balancing the potential benefits and challenges, ChatGPT can enhance students’ learning experiences in higher education.

91Generative Artificial Intelligence: Implications and Considerations for Higher Education PracticeOpenAlex

Tom Farrelly, Nick Baker
Generative Artificial Intelligence (GAI) has emerged as a transformative force in higher education, offering both challenges and opportunities. This paper explores the multifaceted impact of GAI on academic work, with a focus on student life and, in particular, the implications for international students. While GAI, exemplified by models like ChatGPT, has the potential to revolutionize education, concerns about academic integrity have arisen, leading to debates on the use of AI detection tools. This essay highlights the difficulties in reliably detecting AI-generated content, raising concerns about potential false accusations against students. It also discusses biases within AI models, emphasizing the need for fairness and equity in AI-based assessments with a particular emphasis on the disproportionate impact of GAI on international students, who already face biases and discrimination. It also highlights the potential for AI to mitigate some of these challenges by providing language support and accessibility features. Finally, this essay acknowledges the disruptive potential of GAI in higher education and calls for a balanced approach that addresses both the challenges and opportunities it presents by emphasizing the importance of AI literacy and ethical considerations in adopting AI technologies to ensure equitable access and positive outcomes for all students. We offer a coda to Ng et al.’s AI competency framework, mapped to the Revised Bloom’s Taxonomy, through a lens of cultural competence with AI as a means of supporting educators to use these tools equitably in their teaching.

92ChatGPT and Generative Artificial Intelligence for Medical Education: Potential Impact and OpportunityOpenAlex

Christy Boscardin, Brian C. Gin, Polo Black Golde, et al.
ABSTRACT: ChatGPT has ushered in a new era of artificial intelligence (AI) that already has significant consequences for many industries, including health care and education. Generative AI tools, such as ChatGPT, refer to AI that is designed to create or generate new content, such as text, images, or music, from their trained parameters. With free access online and an easy-to-use conversational interface, ChatGPT quickly accumulated more than 100 million users within the first few months of its launch. Recent headlines in the popular press have ignited concerns relevant to medical education over the possible implications of cheating and plagiarism in assessments as well as excitement over new opportunities for learning, assessment, and research. In this Scholarly Perspective, the authors offer insights and recommendations about generative AI for medical educators based on literature review, including the AI literacy framework. The authors provide a definition of generative AI, introduce an AI literacy framework and competencies, and offer considerations for potential impacts and opportunities to optimize integration of generative AI for admissions, learning, assessment, and medical education research to help medical educators navigate and start planning for this new environment. As generative AI tools continue to expand, educators need to increase their AI literacy through education and vigilance around new advances in the technology and serve as stewards of AI literacy to foster social responsibility and ethical awareness around the use of AI.

93Generative artificial intelligence augmenting SME financial managementOpenAlex

Michael Metzger, Seán O’Reilly, Ciarán Mac an Bhaird
This study investigates the potential for entrepreneurs to leverage advances in technological innovation, specifically generative Artificial Intelligence (AI), to build management capability to mitigate business and financial risks. Drawing on theories of Technology Affordances and Constraints and the Resource-Based View (RBV) of the firm, recognising that small and medium-sized enterprises (SMEs) are inherently resource-constrained. We examine how AI-generated financial diagnostics can empower SMEs by generating accessible, real-time analysis and insights, thus bolstering the management function and increasing chances of survival and growth. Using a dataset of 1,150 UK SMEs spanning eight years of financial statements, we test a large language model (LLM) prediction assessment and analyse the potential for SMEs to utilise the technology, notwithstanding enterprise-specific constraints. We conclude that AI may be a very effective tool for smaller enterprises to augment the financial management function, although its efficacy hinges on organisational readiness, competence in interpreting data, and the will to act on automated red-flag alerts. These findings offer practical guidance for SMEs seeking to enhance their financial management processes in today's digital era.

94Empirical Study on the Effectiveness of Generative AI in Financial Risk Management for Small and Medium EnterprisesOpenAlex

Xinge Li
With the rapid development of the digital economy, small and medium enterprises (SMEs) face increasingly complex and diverse financial risks. Traditional financial risk management methods exhibit significant limitations in processing massive data and predicting complex risk patterns. The emergence of generative artificial intelligence technology provides innovative solutions for financial risk management in SMEs. This study focuses on the application effectiveness of generative AI technology in SME financial risk management. Through constructing a theoretical analytical framework and employing a combined research methodology of questionnaire surveys, case analysis, and empirical testing, we conducted an in-depth investigation of 235 SMEs. The research findings reveal that generative AI technology significantly outperforms traditional methods in financial risk identification accuracy, prediction precision, and management efficiency, with risk identification accuracy improving by 27.3% and risk prediction precision increasing by 34.5%. Meanwhile, factors such as technology acceptance, data quality, and organizational support have significant impacts on the effectiveness of generative AI. The research results provide theoretical guidance and practical reference for SMEs to rationally utilize generative AI technology to enhance their financial risk management capabilities.

95The Impacts of Artificial Intelligence on Business Innovation: A Comprehensive Review of Applications, Organizational Challenges, and Ethical ConsiderationsOpenAlex

Rubén Machucho-Cadena, Ortíz González
This review synthesizes current knowledge on the transformative impacts of artificial intelligence (AI)—computational systems capable of performing tasks requiring human-like reasoning—on business innovation. It addresses the potential of AI to reshape strategies, operations, and value creation across various industries. Key themes include AI-driven business model innovation, human–AI collaboration, ethical governance, operational efficiency, customer experience personalization, organizational capability development, and adoption disparities. AI enables scalable product development, personalized service delivery, and data-driven strategic decisions. Successful implementations hinge on overcoming technical, cultural, and ethical barriers, with ethical AI adoption enhancing consumer trust and competitiveness, positioning responsible innovation as a strategic imperative. For practitioners, this review offers evidence-based frameworks for aligning AI with business objectives. For academics, it identifies research frontiers, including longitudinal impacts, context-specific roadmaps for small- and medium-sized enterprises, and sustainable innovation pathways. This review conceptualizes AI as a driver of systemic organizational transformation, requiring continuous learning, ethical foresight, and strategic ability for competitive advantage.

96Artificial Intelligence Adoption in SMEs: Survey Based on TOE–DOI Framework, Primary Methodology and ChallengesOpenAlex

Elsa Delgado-Sánchez, Reyes Calderón, Francisco Herrera
Despite the transformative potential of artificial intelligence (AI), small and medium-sized enterprises (SMEs) continue to face significant challenges in its effective adoption. While prior studies have emphasized strategic benefits and readiness models, there remains a lack of operational guidance tailored to SME realities—particularly regarding implementation barriers, resource constraints, and emerging demands for responsible AI use. This study presents an analysis of AI adoption in SMEs by integrating the technology–organization–environment (TOE) framework with selected attributes from the diffusion of innovations (DOI) theory to examine adoption dynamics through a dual structural and perceptual lens. Empirical insights from sectoral and regional contexts are also incorporated. Ten critical challenges are identified and analyzed across the TOE dimensions, ranging from data access and skill shortages to cultural resistance, infrastructure limitations, and weak governance practices. Notably, the framework is expanded to incorporate responsible AI governance and democratized access to generative AI—particularly open-weight large language models (LLMs) such as LLaMA, DeepSeek-R1, Mistral, and FALCON—as emerging technological and ethical imperatives. Each challenge is paired with actionable, context-sensitive solutions. The paper is a structured, literature-based conceptual analysis enriched by empirical case study insights. As a key contribution, it introduces a structured, six-phase roadmap methodology to guide SMEs through AI adoption—offering step-by-step recommendations aligned with technological, organizational, and strategic readiness. While this roadmap is conceptual and has yet to be validated through field data, it sets a foundation for future diagnostic tools and practical assessments. The resulting study bridges theoretical insight and implementation strategy—empowering inclusive, responsible, and scalable AI transformation in SMEs. By offering both analytical clarity and practical relevance, this study contributes to a more grounded understanding of AI integration and calls for policies, ecosystems, and leadership models that support SMEs in adopting AI not merely as a tool, but as a strategic enabler of sustainable and inclusive innovation.

97Unlocking the Power of Digital Commons: Data Cooperatives as a Pathway for Data Sovereign, Innovative and Equitable Digital CommunitiesOpenAlex

Michael Max Bühler, Igor Calzada, Isabel Cane, et al.
Network effects, economies of scale, and lock-in-effects increasingly lead to a concentration of digital resources and capabilities, hindering the free and equitable development of digital entrepreneurship, new skills, and jobs, especially in small communities and their small and medium-sized enterprises (“SMEs”). To ensure the affordability and accessibility of technologies, promote digital entrepreneurship and community well-being, and protect digital rights, we propose data cooperatives as a vehicle for secure, trusted, and sovereign data exchange. In post-pandemic times, community/SME-led cooperatives can play a vital role by ensuring that supply chains to support digital commons are uninterrupted, resilient, and decentralized. Digital commons and data sovereignty provide communities with affordable and easy access to information and the ability to collectively negotiate data-related decisions. Moreover, cooperative commons (a) provide access to the infrastructure that underpins the modern economy, (b) preserve property rights, and (c) ensure that privatization and monopolization do not further erode self-determination, especially in a world increasingly mediated by AI. Thus, governance plays a significant role in accelerating communities’/SMEs’ digital transformation and addressing their challenges. Cooperatives thrive on digital governance and standards such as open trusted application programming interfaces (“APIs”) that increase the efficiency, technological capabilities, and capacities of participants and, most importantly, integrate, enable, and accelerate the digital transformation of SMEs in the overall process. This review article analyses an array of transformative use cases that underline the potential of cooperative data governance. These case studies exemplify how data and platform cooperatives, through their innovative value creation mechanisms, can elevate digital commons and value chains to a new dimension of collaboration, thereby addressing pressing societal issues. Guided by our research aim, we propose a policy framework that supports the practical implementation of digital federation platforms and data cooperatives. This policy blueprint intends to facilitate sustainable development in both the Global South and North, fostering equitable and inclusive data governance strategies.

98Generative AI for BIM-based Digital Construction Cost Management: A Qualitative Sentiment Analysis ApproachOpenAlex

Temitope Omotayo, Jiamei Deng, Md. Shohrab Hossain, et al.
This study aimed to understand the sentiments and acceptance of Generative AI (GenAI) for digital cost management in UK construction businesses by investigating the ethical, technical, market entry, and operational requirements of GenAI in this context. Using a qualitative approach that covered various themes, a multiple case study research strategy was employed, involving micro, small, and large organisations. Sentiment analysis, a branch of natural language processing, was utilised to analyse interview findings, providing insights into participants' emotional undertones and opinions. The study involved four case studies with nine participants from micro, small, large, and academic organisations. These participants provided insights into the ethical considerations, regulations, maintenance, and operations of a GenAI platform for digital cost management. The findings were presented descriptively across four themes: ethics, market entry, technical operations, and operations. The study found a need for a balanced approach to ethics, emphasising transparency and regulatory compliance. Market entry, adaptability, regulatory compliance, and affordability were identified as key factors influencing the adoption of GenAI tools. The technical operations theme revealed a positive sentiment towards the operational benefits of GenAI, such as improved efficiency and decision-making, but also emphasised the need for professional oversight. Operational challenges included workforce training and quality assurance. The implications of these findings are significant for the adoption of BIM and GenAI in the construction sector, especially among SMEs. The integration of these technologies promises to revolutionise operations, offering enhanced efficiency and collaboration. However, challenges such as perceived complexity, initial investment, and the need for skilled personnel must be addressed. The study suggests that overcoming these barriers requires a concerted effort from industry stakeholders, policymakers, and academia to ensure the effective adoption and implementation of BIM and GenAI in the construction industry.

99Acceptance and integration of Artificial intelligence and machine learning in the construction industry: Factors, current trends, and challengesOpenAlex

Nitin Liladhar Rane, Pravin Desai, Jayesh Rane
The construction industry, historically hesitant in adopting new technologies, is undergoing significant transformation with the integration of artificial intelligence (AI). This research delves into the various elements influencing AI acceptance and implementation within this sector. The study applies well-established models and theories of technology acceptance, including the Technology Acceptance Model (TAM), Unified Theory of Acceptance and Use of Technology (UTAUT), and Innovation Diffusion Theory (IDT), specifically adapted to the unique context of the construction industry. Critical factors driving AI acceptance encompass perceived usefulness, ease of use, organizational readiness, top management support, and external pressures. Furthermore, the research highlights essential elements such as workforce skills, data availability, and cybersecurity concerns that considerably affect AI adoption. Current trends reveal an increasing utilization of AI in project management, predictive maintenance, and design optimization, with a notable surge in the adoption of AI-powered Building Information Modeling (BIM) and robotics. Despite these advancements, the construction industry encounters significant challenges, including high implementation costs, resistance to change, and a lack of standardization. This research offers a comprehensive review of the current state of AI in the construction industry, providing insights into evolving trends and ongoing challenges. Keywords: Construction Industry, Artificial Intelligence, Project Management, Decision Support Systems, Decision Making, Machine Learning, Construction Projects. Citation: Rane, N. L., Desai, P., & Rane, J. (2024). Acceptance and integration of Artificial intelligence and machine learning in the construction industry: Factors, current trends, and challenges. In Trustworthy Artificial Intelligence in Industry and Society (pp. 134-155). Deep Science Publishing. https://doi.org/10.70593/978-81-981367-4-9_4 4.1 Introduction The construction industry, one of the oldest and most essential sectors worldwide, has continually adapted to technological progress (Irani & Kamal, 2014; Oprach et al., 2019; Whitlock-Glave et al., 2019). Recently, artificial intelligence (AI) has emerged as a transformative element, poised to significantly enhance efficiency, safety, and overall project outcomes in construction (Mohammadpour et al., 2019; Patil, 2019; Akinosho et al., 2020). The adoption and integration of AI in this industry, however, are shaped by various factors and face notable challenges. The acceptance of AI in the construction industry is driven by several pivotal factors. A major incentive is the potential for substantial cost savings and efficiency improvements (Mohammadpour et al., 2019; Patil, 2019). AI technologies, including machine learning algorithms and predictive analytics, can optimize resource allocation, reduce waste, and streamline project management processes, resulting in considerable cost reductions. This is particularly appealing to construction firms operating in a highly competitive market with narrow profit margins. Enhancing safety on construction sites is another critical factor. AI-powered systems can monitor site conditions in real-time, predict potential hazards, and alert workers to dangerous situations (Darko et al., 2020; Sacks et al., 2020; Abioye et al., 2021). For instance, AI can analyze data from wearable devices to detect signs of worker fatigue or stress, thus preventing accidents. This proactive approach not only protects workers but also minimizes project delays and financial losses due to accidents. The increasing complexity of construction projects further drives AI adoption. Modern construction often involves intricate designs and sophisticated engineering requirements (Mohamed & Mohamad, 2021; Heo et al., 2021; Bolpagni & Bartoletti, 2021). AI can assist in managing these complexities through advanced modeling and simulation capabilities. Integrating Building Information Modeling (BIM) with AI enhances the accuracy of project planning and execution, ensuring all project elements are well-coordinated and executed as planned (Momade et al., 2021; Chen & Ying, 2022; Saeed et al., 2022). Several trends illustrate the growing presence of AI in the construction industry. One significant trend is AI’s role in project management (Regona et al., 2022; Mendoza et al., 2022; Regona et al., 2024). AI algorithms can analyze historical project data to provide insights into project timelines, budget forecasts, and resource needs. This data-driven approach enables construction managers to make informed decisions, anticipate potential issues, and adjust plans proactively. Another trend is AI’s application in design and engineering. Generative design, an AI-driven approach, allows engineers to input design parameters and constraints, with the AI generating multiple design alternatives. This accelerates the design phase and often results in innovative and optimized solutions that might not emerge from traditional methods. AI is also advancing in construction robotics. Autonomous machines, such as drones and robots, are increasingly used for tasks like site surveys, bricklaying, and concrete pouring. These AI-driven machines can work continuously without fatigue, boosting productivity and ensuring consistent quality. Drones equipped with AI capabilities are particularly valuable for conducting aerial site inspections and monitoring progress, providing real-time data to keep projects on track. Predictive maintenance powered by AI is becoming a standard practice. By analyzing data from equipment sensors, AI can predict when machinery is likely to fail or require maintenance, allowing for timely interventions (Regona et al., 2022; Regona et al., 2024). This not only extends equipment lifespan but also reduces downtime and maintenance costs. Despite the promising benefits and trends, AI adoption in the construction industry faces significant challenges. One major barrier is the high initial cost of implementing AI technologies. Integrating AI requires substantial investments in hardware, software, and training, which can be prohibitive for many construction firms, particularly small and medium-sized enterprises (SMEs) (Liang et al., 2024; Liu et al., 2024; Adeloye et al., 2023). The industry also confronts a skills gap. Successful AI implementation necessitates a workforce proficient in both construction practices and advanced technological solutions. There is a growing need for training programs to equip construction professionals with the necessary AI-related skills. Without such training, the full potential of AI cannot be realized. Data management presents another critical challenge (Mohapatra et al., 2023; Oluleye et al., 2023). AI systems depend on large volumes of high-quality data to function effectively. In construction, data is often fragmented and stored in disparate systems. Integrating these data sources to create a cohesive and accessible data environment is a complex task that many firms struggle with. Additionally, the conservative nature of the construction industry can hinder AI adoption. The industry has traditionally been slow to embrace new technologies, and there is often resistance to change. This cultural barrier can impede AI implementation, as stakeholders may be reluctant to deviate from established practices and workflows. 4.2 Methodology A thorough literature review was conducted to compile existing knowledge and insights on the acceptance and implementation of AI in the construction industry. The review encompassed academic journals, conference papers, industry reports, and other relevant publications from the past such as and to a and comprehensive of The was on factors influencing AI acceptance, current AI application trends, and by the construction industry in AI technologies. This analyzing and from various to provide a of the existing research insights into the in the a was This and the and of of significant within the the the most and This a of the and research and further This was in the and trends in AI acceptance and implementation within the construction industry. was conducted to the and into providing a of the of research and in AI adoption in the construction industry. By algorithms to the the and into These major of and the and factors with The the of factors influencing AI acceptance, such as technological readiness, organizational and as as and trends within This approach a and comprehensive of the complex of AI implementation in the construction industry. and and of the In the such as of and in the research The of and the research on AI into construction to enhance efficiency, safety, and The these highlights growing and to through with and are essential that training algorithms to from data and make or in the application in the construction industry for predictive maintenance, safety and project of and are also in to devices that and for real-time monitoring and in of and with AI a trend these to create construction These to the of AI in managing construction Decision systems AI to assist in resource allocation, and The role in project outcomes and in 4.1 of the in literature AI and This and These the various AI and in construction, the of AI research on to industry and like and this role in safety and AI is used to predict hazards, monitor safety and analyze industry safety Construction and of and this on construction through and These to data and to construction processes, efficiency, and in engineering Project and like and this highlights application in project AI-driven systems assist project managers in informed decisions, and managing and costs. and such as and the of knowledge management and in adopting AI in practices and are for professionals with the necessary skills to AI solutions. The and provide insights into trends and in adopting AI in The of machine and a on advanced to and enhance The of AI with and a trend construction AI integration with devices enables real-time predictive maintenance, and project outcomes and costs. The of a growing on AI to enhance construction safety AI algorithms predict and monitor and analyze data for AI-driven systems optimize project management processes, in data-driven resource allocation, and and cost data management and integration are for AI implementation in and of data that need and to AI in construction is by a lack of and training programs are essential to this and equip professionals with the necessary AI implementation requires significant in software, and Construction need to the and for AI adoption. AI in construction and such as data and these concerns is for and AI and theories of technology acceptance in construction industry The construction industry, often as traditional and slow to to change, has a significant of artificial intelligence (AI) technologies. The adoption and integration of AI in this are by various models and that the of technology acceptance and the factors influencing to embrace these Technology Acceptance Model The Technology Acceptance Model (TAM), by in is a for technology to factors acceptance of perceived and perceived ease of In the construction industry, perceived to the to which AI technology enhances construction processes, project or competitive ease of to construction professionals can and AI For AI-driven project management benefits such as cost and safety to be perceived as Additionally, these be to adoption construction may lack advanced skills. 4.1 the models and theories of technology acceptance in construction industry. 4.1 and theories of technology acceptance in construction industry in Construction Technology Acceptance Model acceptance of technology on perceived and ease of of the adoption of construction management Unified Theory of Acceptance and Use of Technology elements from multiple models to and technology adoption construction Diffusion of the and factors influencing the of new technologies. Innovation the adoption of new construction like Theory of planned on and perceived to construction that technology adoption is likely when the technology task Technology construction to project needs. Theory and learning and Learning, construction workers in new safety through Model of Innovation on resistance to adopting new technologies. for resistance to new construction Model and of technology on initial and with construction management Theory of on the and to predict the adoption of on Model technology adoption on and factors the adoption of construction robotics. Unified Theory of Acceptance and Use of Technology Building on the Unified Theory of Acceptance and Use of Technology (UTAUT), by et in of technology and In the construction to the that AI to such as project efficiency or to perceived ease of in the of the complexity of AI the of or the of in the industry, from industry can significantly conditions the of and support, including training programs and which are for the implementation of AI in Diffusion of Theory Diffusion of in offers a comprehensive for new and within a or industry. to the adoption of AI in the construction industry can be through and is the perceived of AI existing methods. For instance, AI-driven predictive maintenance can significant cost savings by equipment thus a AI within the existing and practices of construction the perceived of and AI technologies. to the to which AI can be on a involves the of AI such as through projects or AI in Theory of The Theory of by in the role of and perceived in technology In the construction industry, AI are shaped by benefits and are when AI is perceived to enhance safety, and streamline to the of and industry on an to stakeholders and industry for AI a environment for to the of to AI effectively. training and can enhance this of for construction professionals to Innovation Theory Innovation Theory by and in offers insights into the resistance to new technologies. that resistance can from and and In the construction industry, might concerns the of AI high implementation costs, or potential to existing workflows. of in AI or to established these is for to such as through AI and the role of AI in skills Model The by and in that technology is likely to be with the tasks is to In the construction industry, this that AI with the and tasks of construction For AI for project site and safety management be to the unique of these AI are with a of construction and are likely to be perceived as valuable and by industry The by et in extends the by the and technological factors. In the construction industry, factors the skills, and of construction workers and factors encompass the and of construction factors the and integration of AI systems. Successful AI adoption in construction requires a approach that these For instance, a of providing comprehensive training and ensuring integration of AI with existing systems can enhance the and acceptance of AI technologies. influencing artificial intelligence acceptance in construction industry and Innovation The of technology is a in the acceptance of AI in the construction industry (Mohapatra et al., 2023; Oluleye et al., 2023). such as Building Information Modeling and the of the for AI These the of of which AI systems can analyze to optimize construction For instance, with AI can predict potential project and project management The and of AI algorithms further in AI’s to and acceptance within the industry. and benefits are in the acceptance of AI in AI can significantly reduce with waste, and project By AI allows construction to to substantial cost Additionally, AI-driven predictive maintenance can the lifespan of downtime and costs. The potential for through these AI an for construction firms, driving and The of AI on the workforce presents both and challenges. one AI can by and allowing workers to on the other there is and the need for new Successful acceptance of AI on these workforce training and for workers to and with AI systems can ease the AI can enhance safety and create new can of and a AI adoption. and and a significant role in AI acceptance in The industry is with safety and AI with these to be and for AI implementation can provide a that safety, and and industry need to and these to in AI systems. Additionally, to data and in AI-driven construction projects is essential for and and management AI with a that and technological are likely to AI solutions. a pivotal role in AI adoption by benefits and into the Furthermore, a of and to embrace new technologies. to is a but and of AI’s can Project and The complexity and of construction projects also AI projects with high of complexity the most from AI integration due to the significant of data and the need for AI can enhance project management by resource allocation, and For the on in AI might not be as a challenge to the and of AI solutions to various project can in acceptance the industry. and The construction industry is multiple stakeholders such as and The acceptance of AI on to and these AI-powered that with existing systems and data can enhance a that AI construction firms, and industry is for driving AI adoption. and can the and acceptance of AI technologies. and in AI systems is to Construction professionals need to that AI solutions are and in Building this requires in AI algorithms make and ensuring that AI systems are and of AI’s capabilities and can Additionally, practices and for AI implementation in construction can in a environment for AI adoption. and with A challenge in the construction involves AI with current systems and workflows. The industry on and systems for project management, design, and For AI solutions to be with these existing systems. Without there is a of data and AI solutions be to and integration with established and adoption. and of AI The and of AI are essential for acceptance within the construction industry. the in project and AI solutions be to need to to a of from small to and AI that can be to project are likely to be the to AI solutions in with project and the project and a growing on the construction industry to AI has the potential to enhance these by resource waste, and for AI to acceptance, with AI solutions that construction as the of of and of likely to be the benefits of AI can from stakeholders a role in the adoption of AI including such as data and are In construction, this to ensuring of and data AI systems be to and to worker and proactive of concerns can in AI technologies. and for AI in construction can further these concerns and and The design of AI technologies, particularly and significantly Construction professionals in and complex or systems can hinder adoption. AI solutions that are accessible and to comprehensive training and is also critical to utilization of AI technologies. A can enhance overall and acceptance of AI in the construction industry. 4.2 the factors influencing artificial intelligence acceptance in construction industry. 4.2 influencing artificial intelligence acceptance in construction industry Current trends in artificial intelligence adoption in construction industry Project The adoption of AI in construction is driven by the for project management, safety, and cost efficiency (Mohamed & Mohamad, 2021; Heo et al., 2021; Bolpagni & Bartoletti, 2021). A notable trend in the construction industry is the application of AI in project AI algorithms analyze from projects to potential delays and cost By historical and real-time AI project managers with informed and optimized resource such as predictive and machine learning models are becoming essential in and project to project Building Information Modeling (BIM) Building Information Modeling (BIM) has construction planning and Integrating AI with is an trend that the capabilities of both technologies. the design detect design and optimize This integration and and Additionally, AI data to insights on and and construction and and adoption in construction is significantly by AI-powered tasks such as bricklaying, concrete and site with high the on and Autonomous construction by in worker safety and project AI is increasingly used to optimize and Generative design algorithms AI to create multiple design on parameters and This enables and engineers to various design and the most and solutions. AI-driven design enhances and that designs are and with particularly in the of project Construction and AI significantly enhances site monitoring and safety AI-powered and drones monitor construction sites in real-time, potential and ensuring safety These systems detect and alert site Additionally, AI data from wearable devices to monitor and fatigue to a Predictive Predictive maintenance is a growing trend in AI adoption within the construction industry. AI algorithms analyze data from in construction equipment to predict maintenance preventing and Predictive maintenance equipment lifespan and project are particularly valuable for AI in The construction complex involves stakeholders and AI management, efficiency and costs. AI algorithms optimize and effectively. real-time AI construction in decisions, waste, and ensuring timely preventing delays and cost and Building is a growing in construction, and AI a role in AI data on and to for the of construction For instance, AI designs for efficiency, and construction to Integrating AI into enables construction to and requirements effectively. with AI AI in the construction industry by providing data-driven insights and AI-powered of data from various including project management software, financial and to These insights construction managers to make informed project resource allocation, and AI for project reduces costs, and overall The with a of the construction industry, the AI adoption increasing to substantial This adoption is several AI-driven design with design, Building Information Modeling and Predictive maintenance with on equipment and resource Construction planning and for and resource monitoring and management in and with and Autonomous equipment and management and Construction site monitoring and efficiency, and the role of real-time progress and The the of AI various of the construction industry, growing and to illustrate the current trends in AI adoption in the construction industry of artificial intelligence adoption in construction industry Data and A major to AI adoption in construction is the and of AI systems depend on to and effectively. In the construction industry, data is often and fragmented due to the of and For instance, data may be in various by the industry. This can significantly impede the and of AI which on data for machine learning with The construction industry a of systems and technologies, from project management and Building Information Modeling (BIM) to various machinery and Integrating AI solutions with these existing systems can be complex and These systems are often not to with one to the advanced data for AI into existing without ongoing presents a significant challenge that planning and AI technology involves substantial initial costs. These not only the AI technology but also the to AI such as advanced and data Additionally, significant investments may be in training and existing to new AI-driven methods. For many construction particularly small to medium-sized these can be AI adoption. The construction industry faces a notable in and to and AI systems. The current workforce is often with traditional and may lack the necessary skills to and AI technologies. this requires substantial in training and new with in data and which can be a slow and and Integrating AI into construction also various and For AI for monitoring and managing construction sites concerns data and with and Furthermore, as AI systems in critical processes, of and when an AI-driven or is complex and not by existing to is a significant barrier in many including There can be and resistance from to traditional methods. This cultural resistance can slow the adoption of new technologies, as may be reluctant to or on AI solutions. this challenge requires the benefits of training programs that the for the and is in new including that can enhance or not safety the of AI in critical tasks that are that AI systems are and various conditions is essential these systems can be are in the adoption of AI in AI-driven can significant from due to to project management that might that AI systems and is for and acceptance from all The integration of artificial intelligence (AI) within the construction industry a by various technology acceptance models and such as the Technology Acceptance Model and the Unified Theory of Acceptance and Use of Technology been in the factors driving AI adoption in These models perceived usefulness, ease of use, and as of AI the Diffusion of highlights the of and AI adoption These insights valuable on construction professionals and AI technologies. The factors influencing AI acceptance in the construction industry are and factors the perceived benefits of AI in boosting safety, and costs. readiness, the of and a Additionally, external factors such as and competitive significantly affect AI adoption. The of these factors a complex environment AI acceptance and implementation depend on both and Current trends in AI adoption within the construction industry reveal a growing of AI transformative AI-powered are for project management, predictive maintenance, and design The integration of AI with Building Information Modeling (BIM) and the of as real-time data and project efficiency and Additionally, AI-driven and are construction processes, and costs. These trends a and construction Despite these promising trends, AI adoption in the construction industry faces several challenges. to change, lack of and AI and concerns data and are significant Furthermore, the high initial and the for training and of the workforce considerable challenges. The construction fragmented with stakeholders and of technological further the adoption to resistance to change, data and substantial requirements be critical to potential in the construction industry. these trends and the the for a innovative and in L., J. & Artificial intelligence in the construction industry: A review of and challenges. of Building & of Artificial Intelligence (AI) in the construction industry: A review of & Deep learning in the construction industry: A review of and of Building & Bartoletti, Artificial intelligence in the construction industry: benefits and In of the P., & Ying, Artificial intelligence in the construction industry: and P., & Artificial intelligence in the industry: and of research in construction, & of data the artificial intelligence projects in the engineering and construction industry. & Kamal, systems research in the construction industry. with & J. J. (2024). of artificial intelligence and in the and construction industry. in L., & (2024). Artificial intelligence for and management in construction industry: A literature Information J. & A review on the role of artificial intelligence in the construction industry. & Mohamad, The implementation of artificial intelligence (AI) in the construction industry. In Publishing. & Artificial intelligence to design and In of the on and in Construction & of Artificial Intelligence in the Construction J. & review of application of artificial intelligence in engineering and in and & Artificial Intelligence for the implementation of in the construction industry: A critical and & Building the of the construction industry through artificial intelligence and Patil, of artificial intelligence in construction of in & (2024). Artificial Intelligence and of the Construction and & Artificial for the construction industry: are perceived and in of and & and adoption of AI in the construction industry: A of and & Building artificial intelligence and construction in the & Artificial Intelligence and in Construction L., & The technology in the construction industry: the of artificial

100Developing Scalable Compliance Architectures for Cross-Industry Regulatory AlignmentOpenAlex

Emmanuel Cadet, Lawal Abdulmutalib Babatunde, Joshua Oluwagbenga Ajayi, et al.
The rapid globalization of digital business ecosystems and the proliferation of complex, sector-specific regulations have amplified the challenge of achieving unified compliance across diverse industries. Organizations operating in multi-sector environments face fragmented regulatory obligations, often resulting in redundant processes, inefficiencies, and increased operational risk. This paper presents a comprehensive approach to developing scalable compliance architectures designed to enable cross-industry regulatory alignment while maintaining agility, cost-effectiveness, and resilience. The proposed architecture integrates modular, interoperable components capable of mapping and harmonizing overlapping regulatory requirements from finance, healthcare, manufacturing, energy, and other highly regulated sectors. By leveraging cloud-native infrastructure, artificial intelligence, machine learning, and regulatory technology (RegTech) solutions, the architecture supports automated rule interpretation, dynamic compliance control mapping, and continuous monitoring. Key features include a multi-layered governance model, a unified regulatory taxonomy, and an adaptive control library capable of aligning with evolving legal mandates and industry standards such as GDPR, HIPAA, PCI DSS, ISO 27001, and NERC CIP. A central innovation is the deployment of an intelligent compliance orchestration engine that enables real-time risk scoring, policy enforcement, and cross-sector reporting while reducing audit preparation times and minimizing compliance fatigue. The scalability of the framework is achieved through microservices architecture and API-driven interoperability, allowing seamless integration with existing enterprise resource planning (ERP), governance, risk, and compliance (GRC) platforms, and security information and event management (SIEM) systems. Using simulated enterprise deployment scenarios and multi-sector compliance datasets, the proposed architecture demonstrates significant improvements in regulatory coverage, operational efficiency, and cost optimization. Furthermore, the study explores governance models for maintaining ethical AI usage, data privacy, and cross-border compliance consistency. This work provides a blueprint for organizations seeking to unify fragmented compliance operations, enabling them to transition from reactive, sector-specific adherence toward proactive, enterprise-wide regulatory alignment that enhances trust, resilience, and competitive advantage in the global digital economy.

101Digital access and inclusion for SMEs in the financial services industry through Cybersecurity GRC: A pathway to safer digital ecosystemsOpenAlex

Oluwatosin Yetunde Abdul-Azeez, Alexsandra Ogadimma Ihechere, Courage Idemudia
The integration of digital technologies into the financial services industry has revolutionized how small and medium-sized enterprises (SMEs) access and utilize financial services. However, this digital transformation also brings heightened cybersecurity risks, making robust governance, risk management, and compliance (GRC) frameworks essential for fostering a safer digital ecosystem. This paper explores the pivotal role of cybersecurity GRC in enhancing digital access and inclusion for SMEs within the financial sector. By analyzing current challenges and opportunities, we propose a comprehensive approach to fortifying cybersecurity measures that align with the unique needs of SMEs. Firstly, the paper identifies the primary cybersecurity threats facing SMEs, including data breaches, phishing attacks, and ransomware, which can severely impact their operations and financial stability. It underscores the importance of a proactive GRC strategy that encompasses risk assessment, policy development, and continuous monitoring to mitigate these threats effectively. Moreover, the paper highlights the necessity for regulatory compliance, stressing how adherence to standards such as GDPR, PCI DSS, and ISO/IEC 27001 can bolster SMEs' defenses and enhance their credibility with customers and partners. Secondly, the research delves into the benefits of enhanced digital access and inclusion facilitated by a strong cybersecurity GRC framework. These benefits include improved financial inclusion for underbanked SMEs, streamlined access to digital financial services, and the promotion of innovation and competitiveness. The paper argues that by ensuring a secure digital environment, SMEs can confidently adopt emerging technologies such as blockchain, artificial intelligence, and cloud computing, driving growth and efficiency. Lastly, the paper presents case studies of successful cybersecurity GRC implementations in the financial services sector, showcasing best practices and lessons learned. It provides practical recommendations for SMEs to develop and maintain robust GRC frameworks, including leveraging automated tools for threat detection, fostering a culture of cybersecurity awareness, and engaging with cybersecurity experts for continuous improvement. In conclusion, the paper asserts that a comprehensive cybersecurity GRC strategy is crucial for enhancing digital access and inclusion for SMEs in the financial services industry. By addressing cybersecurity risks and ensuring compliance, SMEs can safely navigate the digital landscape, unlocking new opportunities for growth and innovation while contributing to a more secure and inclusive digital economy. Keywords: Digital Access, Inclusion, SMEs, Financial Services, Cybersecurity.

102Artificial Intelligence for Sustainability: Evidence from select Small and Medium Enterprises in the PhilippinesOpenAlex

Alexander A. Hernandez, Arlene R. Caballero, Erlito M. Albina, et al.
Artificial intelligence (AI) is an emerging technology in small and medium enterprises (SMEs). Recently, SMEs have growing interests in using artificial intelligence to improve business performance. However, SMEs are confronted with sustainability, which remains a research gap. This paper presents the first evidence of select SMEs using artificial intelligence in business for sustainability in the Philippines, through a qualitative study involving managers. Results show that a few SMEs have AI applications in business that contributes to sustainability. While there are on-going sustainability efforts, most SMEs are in the incremental and situational development levels. Also, this study confirms that insufficient physical and technological infrastructure, availability of data, customers privacy and security, insufficient legal frameworks, management support, and lack of AI adoption strategy are evident issues and challenges that limits the progress AI application in business for sustainability. This study presents some implications to SMEs, policy-making and future work to progress AI for sustainability.

103Strategic Integration of Generative AI in Organizational Settings: Applications, Challenges, and Adoption RequirementsOpenAlex

Mousa Al-kfairy
Generative AI is revolutionizing the way organizations operate, offering transformative capabilities that span automated content creation, strategic decision-making, and customer engagement through AI-driven chatbots. This article conducts a comprehensive literature review to explore the applications, challenges, and strategic requirements for adopting generative AI in organizational contexts, focusing on the distinct needs of small and medium enterprises (SMEs) and large organizations. The findings reveal that generative AI can improve efficiency, drive innovation, and improve customer satisfaction, but its adoption pathways differ significantly between organizational sizes. For SMEs, the emphasis lies on cost-effective and scalable solutions that optimize resource-constrained operations. At the same time, large organizations leverage their extensive resources to scale AI applications, manage complex systems, and address ethical and regulatory challenges. The study highlights critical barriers, including data privacy concerns, integration with legacy systems, and resistance to change, alongside actionable recommendations for overcoming these challenges. By synthesizing insights from 38 high-quality studies, this research bridges the gap between theory and practice. It provides a roadmap for organizations of varying scales to harness generative AI as a cornerstone of their digital transformation journey. It also identifies key areas for future exploration, ensuring relevance in this rapidly evolving field.

104Leveraging Generative AI Tools Like ChatGPT for Startups and Small Business GrowthOpenAlex

David W. Townsend
Artificial intelligence (AI) is transforming the way businesses operate, especially with the advent of generative AI tools such as OpenAI's ChatGPT.These cutting-edge technologies offer an array of opportunities for startups and small businesses to optimize their processes, enhance customer engagement, and drive growth.This article delves into the world of generative AI, highlighting its potential applications, opportunities, and risks, and concludes with practical steps entrepreneurs can take to implement these tools in their businesses.

105Human resource management in the age of generative artificial intelligence: Perspectives and research directions on ChatGPTOpenAlex

Pawan Budhwar, Soumyadeb Chowdhury, Geoffrey Wood, et al.
Abstract ChatGPT and its variants that use generative artificial intelligence (AI) models have rapidly become a focal point in academic and media discussions about their potential benefits and drawbacks across various sectors of the economy, democracy, society, and environment. It remains unclear whether these technologies result in job displacement or creation, or if they merely shift human labour by generating new, potentially trivial or practically irrelevant, information and decisions. According to the CEO of ChatGPT, the potential impact of this new family of AI technology could be as big as “the printing press”, with significant implications for employment, stakeholder relationships, business models, and academic research, and its full consequences are largely undiscovered and uncertain. The introduction of more advanced and potent generative AI tools in the AI market, following the launch of ChatGPT, has ramped up the “AI arms race”, creating continuing uncertainty for workers, expanding their business applications, while heightening risks related to well‐being, bias, misinformation, context insensitivity, privacy issues, ethical dilemmas, and security. Given these developments, this perspectives editorial offers a collection of perspectives and research pathways to extend HRM scholarship in the realm of generative AI. In doing so, the discussion synthesizes the literature on AI and generative AI, connecting it to various aspects of HRM processes, practices, relationships, and outcomes, thereby contributing to shaping the future of HRM research.

106An Explainable AI Tool for Operational Risks Evaluation of AI Systems for SMEsOpenAlex

The Anh Han, Debasısh Pandıt, Sina Joneidy, et al.
With the surge in artificial intelligence (AI) adoption by Small and Medium-sized Enterprises (SMEs), ensuring their safety, fairness, and operational assurance has become paramount. Since many SMEs operate with limited resources, they face unique challenges in ethically and securely deploying AI systems. This research delves into the core principles of AI governance, risk management, and testing, specifically tailored for SMEs, emphasising making these concepts accessible and understandable. Through collaborative efforts, including interactive workshops, meetings and surveys with twenty SME participants, we identified vital AI application areas and challenges and conceptualised an evaluation tool leveraging explainable AI. This tool assesses AI-driven systems' robustness, potential biases, and other software and hardware vulnerabilities It also addresses ethical considerations and legal compliance, emphasising establishing trust and accountability with stakeholders as a foundation for successful AI integration. In conclusion, the paper presents a pilot study that conducts a risk analysis of prevalent AI applications, specifically AI-driven language models, for SMEs. This study illustrates how the proposed evaluation tool will integrate risk levels across different application domains.

107Analytical Study of the World's First EU Artificial Intelligence (AI) Act, 2024OpenAlex

Junaid Sattar Butt
The world's first law governing "artificial inelegance" has arrived!The emergence of Artificial Intelligence (AI) technologies has prompted a global discourse on the necessity of regulatory frameworks to govern their development and deployment responsibly.With the escalating integration of Artificial Intelligence (AI) technologies into various facets of human life, the imperative for regulatory frameworks has become paramount.On March 13, 2024, the European Parliament formally adopted the EU Artificial Intelligence Act, 2024 1 ("AI Act, 2024") with a large majority of 523-46 votes in favor of the legislation, the first horizontal and standalone legislation dedicated exclusively to AI governance.The AI Act, 2024 represents a watershed moment in global governance, aiming to establish comprehensive guidelines and safeguards for the development, deployment, and use of AI systems across diverse sectors.Through rigorous analysis of the Act's key components, including definitions, principles, obligations, and enforcement mechanisms, this research seeks to elucidate its potential impact on stakeholders, innovation ecosystems, and societal dynamics worldwide.This study employs a multidisciplinary approach to scrutinize the intricate provisions and implications of the AI Act, 2024 encompassing legal, ethical, socio-economic, and technological dimensions.A crucial aspect of this research will be a deep dive into the specific provisions and regulations outlined in the AI Act, 2024 and will explore how the Act tackles the identification and mitigation of "inelegant biases" within AI systems.Additionally, the research will analyze the AI Act, 2024's requirements for explain-ability in "inelegant" AI decisions, ensuring transparency and accountability.The mechanisms established for enforcement and oversight will also be under scrutiny to understand their effectiveness in upholding the Act's regulations.Furthermore, this research endeavors to identify the strengths, weaknesses, opportunities, and threats inherent in the AI Act, 2024 considering its adaptability to evolving technological landscapes, its alignment with fundamental human rights principles, and its capacity to foster responsible AI innovation while mitigating risks and disparities.This research will contribute valuable insights to ongoing discussions about navigating the complexities of artificial intelligence in a responsible and ethical manner.

108The Brussels Effect and Artificial IntelligenceOpenAlex

Charlotte Siegmann, Markus Anderljung
The European Union is likely to introduce among the first, and most comprehensive AI regulatory regimes of the world’s major jurisdictions. We ask whether the EU’s upcoming regulation for AI will diffuse globally, producing a so-called “Brussels Effect”. Extending Anu Bradford’s work, we outline the mechanisms by which such regulatory diffusion may occur. We consider both the possibility that the EU’s AI regulation will incentivise changes in products offered in non-EU countries (a de facto Brussels Effect) and the possibility it will influence regulation adopted by other jurisdictions (a de jure Brussels Effect). Focusing on the proposed EU AI Act, we tentatively conclude that both de facto and de jure Brussels effects are likely. A de facto effect is particularly likely to arise in large US tech companies with “high-risk” AI systems. The upcoming regulation might be important in offering the first operationalisation of developing and deploying trustworthy AI.

109Regulatory and Compliance Requirements for SMEs Operating AI Systems through Data Centers in the EU, with a Focus on Data Protection Challenges in GermanyOpenAlex

Thomas Joswig, Walter Kurz
This research examines the regulatory challenges encountered by small and medium-sized enterprises (SMEs) operating artificial intelligence (AI) systems through data centres in the European Union (EU), with a particular focus on data protection issues in Germany. The study analyses the interaction between the General Data Protection Regulation (GDPR) and the proposed EU AI Act, emphasising the compliance barriers faced by SMEs. Methods: A mixed-method approach was employed, combining qualitative analysis of regulatory frameworks and scholarly literature with quantitative survey data from SMEs across key industries. This methodology ensured a comprehensive examination of both regulatory requirements and their practical implications. The findings indicate that SMEs demonstrate high familiarity with GDPR (mean score 82.24) but lower awareness of the AI Act (mean score 56.24), with significant intersectoral variation. Challenges include resource limitations, ambiguous ”high-risk” AI classifications, and the complexity of dual compliance. Notably, government and healthcare sectors reported substantial regulatory burdens, while energy and finance sectors exhibited lower preparedness for AI Act requirements. The study reveals the fragmented implementation of GDPR across member states, complicating compliance for cross-border SMEs. The dual demands of GDPR and the AI Act necessitate streamlined regulatory processes and tailored support mechanisms, such as simplified guidelines and financial assistance. Explainability and transparency obligations, while essential for trust, introduce additional administrative burdens that may impede innovation. Harmonising GDPR and AI Act requirements is crucial to enabling SMEs to comply without inhibiting innovation. Policy recommendations include regulatory sandboxes, targeted training, and increased financial support for SMEs to foster legally compliant yet innovative AI applications.

110European AI Standards – Technical Standardisation and Implementation Challenges under the EU AI ActOpenAlex

Robert Kilian, Linda Jäck, Dominik Ebel
Abstract Harmonised standards are the cornerstone of efficient EU AI Act compliance. This paper presents one of the first systematic analyses of European technical and soon to be harmonised standardisation for organisations providing AI systems. Based on in-depth qualitative interviews with twenty-three leading European organisations developing AI applications across different sectors, such as Mistral and Helsing, and providing transparency regarding the status quo of draft standards, it examines how companies, especially start-ups and SMEs, are dealing with the contemplated standardisation under the EU AI Act and sectoral standardisation. Industry sectors covered include mobility, finance, manufacturing, healthcare, as well as defense and legal tech. Key challenges identified comprise an insufficient effective implementation period of likely less than 6 months compared to at least 12 months actually required for around thirty (partially referenced) technical standards, an imbalance of participation and influence in standardisation committees, double regulation and technical implementation hurdles as well as significant annual costs for harmonised standards compliance. Technical standards are currently reshaping global AI competition and will have a massive influence on the AI landscape as market entry barriers, particularly on start-ups. Hence, the paper offers policy recommendations based on the revealed challenges for AI providers.

111An Open Knowledge Graph-Based Approach for Mapping Concepts and Requirements between the EU AI Act and International StandardsOpenAlex

Julio Hernández, Delaram Golpayegani, David Lewis
The many initiatives on trustworthy AI result in a confusing and multipolar landscape that organizations operating within the fluid and complex international value chains must navigate in pursuing trustworthy AI. The EU's AI Act will now shift the focus of such organizations toward conformance with the technical requirements for regulatory compliance, for which the Act relies on Harmonized Standards. Though a high-level mapping to the Act's requirements will be part of such harmonization, determining the degree to which standards conformity delivers regulatory compliance with the AI Act remains a complex challenge. Variance and gaps in the definitions of concepts and how they are used in requirements between the Act and harmonized standards may impact the consistency of compliance claims across organizations, sectors, and applications. This may present regulatory uncertainty, especially for SMEs and public sector bodies relying on standards conformance rather than proprietary equivalents for developing and deploying compliant high-risk AI systems. To address this challenge, this paper offers a simple and repeatable mechanism for mapping the terms and requirements relevant to normative statements in regulations and standards, e.g., AI Act and ISO management system standards, texts into open knowledge graphs. This representation is used to assess the adequacy of standards conformance to regulatory compliance and thereby provide a basis for identifying areas where further technical consensus development in trustworthy AI value chains is required to achieve regulatory compliance.

112Scalable Agile Framework for Execution in AI for Medical AI Ethics Policy Design in Small- and Medium-Sized Enterprises.PubMed

Ion Nemteanu, Adir Mancebo, Leslie Joe, et al.
J Med Internet Res. 2026 Feb 25;28:e80028. doi: 10.2196/80028.
Artificial intelligence (AI) is transforming patient care, but it also raises ethical questions, such as bias and transparency. While a range of well-established frameworks exist to guide responsible AI practice, most were designed for academic or regulatory settings and can be hard to operationalize within fast-moving, resource-limited small and medium-sized enterprises (SMEs). We report on the collaborative design of the SAFE-AI (Scalable Agile Framework for Execution in AI), an approach that embeds ethical safeguards, including fairness, transparency, responsibility metrics, and continuous monitoring, directly into standard Agile development cycles. In keeping with established Agile principles, SAFE-AI provides "just enough structure" to integrate ethical oversight into existing workflows without prescribing extensive new governance layers. Similar to other Agile frameworks, such as Scrum, which is described as a "lightweight framework" designed to help teams solve complex problems through iterative learning and minimal process overhead, SAFE-AI aims to remain practical for organizations that may not have dedicated ethics or compliance staff. Rather than simplifying technical methods, SAFE-AI simplifies when and how ethical review is triggered and documented, making responsible AI practices feasible even in environments with limited ethics, governance, or compliance resources. SAFE-AI assumes the presence of qualified data scientists and engineers, and it does not replace the need for statistical or technical expertise but instead provides a lightweight structure for coordinating and documenting work that those experts already perform. We followed a design-science, practice-oriented approach over 20 weeks. After a discovery workshop, a cross-functional team was assembled that included SME employees, ethics researchers, and academic partners. The SME's role was limited to informing design constraints and feasibility considerations during the co-design phase. No operational pilot or production deployment was conducted as part of this study. To reduce the risk of internal design bias and improve generalizability, we also consulted external stakeholders through structured feedback sessions, including clinicians, health care domain experts, and regulatory specialists. Their feedback was incorporated into each prototype-feedback cycle, ensuring that priorities reflected not only the SME's immediate context but also broader clinical and regulatory perspectives. The co-design process produced a 4-phase SAFE-AI life cycle: discovery, assessment, development, and monitoring. SAFE-AI's phase-specific checklists meld acceptance, fairness, and transparency metrics into each Agile sprint. A novel scenario-based probability analogy mapping method was added to translate model risk and uncertainty into plain-language narratives for nontechnical stakeholders, forming the framework's core "responsibility metrics" layer. SAFE-AI is presented as a proposed framework showing that meaningful ethical safeguards can be embedded easily within common workflows used by SMEs that already use basic Agile or iterative development practices. Its checklist-driven phases and automatic review triggers provide a defensible way to track fairness, transparency, and responsibility throughout the model lifecycle.