研究：机器能造成伤害，但惩罚机器无法带来同等正义感

Algorithms, robots, and artificial intelligence are increasingly making morally salient decisions (e.g., medical triage) and taking actions (e.g., terminate employees). Extant research suggests that we can and do perceive machines to be responsible for moral harm. Unlike human perpetrators, however, we suggest that we do not experience the same level of justice satisfaction after punishing machine perpetrators. Through a series of experimental paradigms, we first show that people do perceive machines as capable of wrongdoings and punish them accordingly (Pilot Study), but that people experience lower levels of justice satisfaction after punishing machines compared to human perpetrators (Study 1). Studies 2 and 3 replicate some of these findings, whereas Study 4 tests two mechanisms that might account for these effects—punishment as communication and punishment as retribution. A final supplemental study explores whether the dissatisfaction from punishing machines would cause people to displace their aggression onto innocent human bystanders. Our findings demonstrate that machines can commit wrongs and that people attempt to punish them, yet such punishment does not consistently restore a sense of justice relative to punishing humans. This pattern underscores the need for further research on how people understand and respond to machine wrongdoing. (PsycInfo Database Record (c) 2026 APA, all rights reserved)

中文翻译还没有生成，先显示原文。