Niehaus Susanne, Hartwig Matthias, Rosen Patricia H, Wischniewski Sascha
Unit Human Factors, Ergonomics, Federal Institute of Occupational Health and Safety (BAuA), Dortmund, Germany.
Front Artif Intell. 2022 Jul 6;5:868382. doi: 10.3389/frai.2022.868382. eCollection 2022.
The continuous and rapid development of AI-based systems comes along with an increase in automation of tasks and, therewith, a qualitative shift in opportunities and challenges for occupational safety and health. A fundamental aspect of humane working conditions is the ability to exert influence over different aspects of one's own work. Consequently, stakeholders contribute to the prospect of maintaining the workers' autonomy albeit increasing automation and summarize this aspiration with the human in control principle. Job control has been part of multiple theories and models within the field of occupational psychology. However, most of the models do not include specific technical considerations nor focus on task but rather on job level. That is, they are possibly not able to fully explain specific changes regarding the digitalization of tasks. According to the results of a large-scale study on German workers (DiWaBe), this seems to be the case to some extend: the influence of varying degrees of automation, moderated by perceived autonomy, on workers' wellbeing was not consistent. However, automation is a double-edged sword: on a high level, it can be reversely related to the workers' job control while highly autonomous and reliable systems can also create opportunities for more flexible, impactful and diverse working tasks. Consequently, automation can foster and decrease the factor of job control. Models about the optimal level of automation aim to give guidelines on how the former can be achieved. The results of the DiWaBe study indicate that automation in occupational practice does not always happen in line with these models. Instead, a substantial part of automation happens at the decision-making level, while executive actions remain with the human. From an occupational safety and health perspective, it is therefore crucial to closely monitor and anticipate the implementation of AI in working systems. Constellations where employees are too controlled by technology and are left with a high degree of demands and very limited resources should be avoided. Instead, it would be favorable to use AI as an assistance tool for the employees, helping them to gather and process information and assisting them in decision-making.
基于人工智能的系统持续快速发展，带来了任务自动化程度的提高，进而使职业安全与健康面临的机遇和挑战发生了质的转变。人道工作条件的一个基本方面是对自身工作不同方面施加影响的能力。因此，尽管自动化程度不断提高，但利益相关者仍致力于维护工人的自主权，并以“人在控制中”原则来概括这一愿望。工作控制一直是职业心理学领域多种理论和模型的一部分。然而，大多数模型并未纳入具体的技术考量，也未聚焦于任务，而是侧重于工作层面。也就是说，它们可能无法充分解释任务数字化方面的具体变化。根据一项针对德国工人的大规模研究（DiWaBe）的结果，在某种程度上似乎确实如此：不同程度的自动化在感知自主权的调节下对工人幸福感的影响并不一致。然而，自动化是一把双刃剑：在较高层面上，它可能与工人的工作控制呈反向关系，而高度自主且可靠的系统也能为更灵活、更有影响力和更多样化的工作任务创造机会。因此，自动化既能促进也能降低工作控制因素。关于自动化最佳水平的模型旨在就如何实现前者提供指导方针。DiWaBe研究的结果表明，职业实践中的自动化并不总是与这些模型相符。相反，很大一部分自动化发生在决策层面，而执行行动仍由人来完成。从职业安全与健康的角度来看，因此密切监测和预测人工智能在工作系统中的实施至关重要。应避免员工受技术过度控制、需求高度集中且资源非常有限的情况。相反，将人工智能用作员工的辅助工具，帮助他们收集和处理信息并协助他们进行决策，会是比较有利的做法。
