Autonomous robots increasingly perform functions that are potentially hazardous and could cause injury to people (e.g., autonomous driving). When this happens, questions will arise regarding responsibility, although autonomy complicates this issue – insofar as robots seem to control their own behaviour, where would blame be assigned? Across three experiments, we examined whether robots involved in harm are assigned agency and, consequently, blamed. In Studies 1 and 2, people assigned more agency to machines involved in accidents when they were described as ‘autonomous robots’ (vs. ‘machines’), and in turn, blamed them more, across a variety of contexts. In Study 2, robots and machines were assigned similar experience, and we found no evidence for a role of experience in blaming robots over machines. In Study 3, people assigned more agency and blame to a more (vs. less) sophisticated military robot involved in a civilian fatality. Humans who were responsible for robots' safe operation, however, were blamed similarly whether harms involved a robot (vs. machine; Study 1), or a more (vs. less; Study 3) sophisticated robot. These findings suggest that people spontaneously conceptualise robots' autonomy via humanlike agency, and consequently, consider them blameworthy agents.

自主机器人越来越多地执行具有潜在危险并可能对人员造成伤害的功能（例如自动驾驶）。当这种情况发生时，就会出现关于责任的问题，尽管自主性使这个问题变得复杂——就机器人似乎控制自己的行为而言，责任归咎于哪里？通过三个实验，我们检查了参与伤害的机器人是否被指定了代理机构，并因此受到指责。在研究 1 和 2 中，当涉及事故的机器被描述为“自主机器人”（相对于“机器”）时，人们会为它们分配更多的代理权，反过来，在各种情况下，人们也会更多地指责它们。在研究 2 中，机器人和机器被分配了类似的经验，但我们没有发现经验在指责机器人而不是机器方面发挥作用的证据。在研究 3 中，人们将更多的代理权和责任归咎于涉及平民死亡事件的更先进（相对于不太先进）的军用机器人。然而，无论伤害涉及机器人（与机器；研究 1），还是更复杂（与更少；研究 3）的机器人，负责机器人安全操作的人类都会受到类似的指责。这些发现表明，人们自发地通过类似人类的代理机构来概念化机器人的自主性，因此认为它们是应受谴责的代理。