Watamura Eiichiro, Liu Yichen, Ioku Tomohiro
Graduate School of Human Sciences, Osaka University, Suita, Osaka, Japan.
Center for International Education and Exchange, Osaka University, Suita, Osaka, Japan.
PLoS One. 2025 Jan 30;20(1):e0318486. doi: 10.1371/journal.pone.0318486. eCollection 2025.
Artificial intelligence (AI) is anticipated to play a significant role in criminal trials involving citizen jurors. Prior studies have suggested that AI is not widely preferred in ethical decision-making contexts, but little research has compared jurors' reliance on judgments by human judges versus AI in such settings.
This study examined whether jurors are more likely to defer to judgments by human judges or AI, especially in cases involving mitigating circumstances in which human-like reasoning may be valued.
Two pre-registered online experiments were conducted with Japanese participants (Experiment 1: N = 1,735, Mage = 48.4; Experiment 2: N = 1,731, Mage = 48.5). Participants reviewed two murder trial vignettes and made sentencing decisions (1 = suspended sentence; 8 = prison sentence) under two conditions: trials with and without mitigating circumstances.
Across both experiments, participants showed no preference for deferring to human judges' or AI judgments when making sentencing decisions. While suspended sentences were more common in cases with mitigating circumstances, this tendency was unrelated to the judgment source. These findings suggest that jurors do not inherently avoid algorithmic judgments and may consider AI opinions on par with those of human judges in certain contexts. However, whether this leads to improved decision-making quality remains an open question, as objectivity (a strength of AI) and emotional considerations (a safeguard for fairness) may interact in complex ways during juror deliberations. Future research should further explore how these factors influence juror attitudes and decisions in diverse trial scenarios, taking into account potential biases in existing literature.
人工智能（AI）预计将在涉及公民陪审员的刑事审判中发挥重要作用。先前的研究表明，在道德决策背景下，人工智能并未得到广泛青睐，但很少有研究比较陪审员在这种情况下对人类法官判断与人工智能判断的依赖程度。
本研究考察了陪审员是否更倾向于听从人类法官或人工智能的判断，尤其是在涉及从轻情节的案件中，这类案件中类似人类的推理可能会受到重视。
对日本参与者进行了两项预先注册的在线实验（实验1：N = 1735，年龄中位数 = 48.4；实验2：N = 1731，年龄中位数 = 48.5）。参与者阅读了两个谋杀案审判的 vignettes，并在两种情况下做出量刑决定（1 = 缓刑；8 = 监禁）：有从轻情节和没有从轻情节的审判。
在两项实验中，参与者在做出量刑决定时，对听从人类法官或人工智能的判断没有表现出偏好。虽然缓刑在有从轻情节的案件中更为常见，但这种趋势与判断来源无关。这些发现表明，陪审员并非天生就回避算法判断，在某些情况下可能会将人工智能的意见与人类法官的意见同等看待。然而，这是否会导致决策质量的提高仍是一个悬而未决的问题，因为客观性（人工智能的优势）和情感因素（公平性的保障）在陪审员审议过程中可能会以复杂的方式相互作用。未来的研究应进一步探讨这些因素如何在不同的审判场景中影响陪审员的态度和决策，同时考虑现有文献中的潜在偏差。
