Data_Sheet_1_Moral Uncanny Valley revisited – how human expectations of robot morality based on robot appearance moderate the perceived morality of robot decisions in high conflict moral dilemmas.PDF

In recent years a new sub-field of moral psychology has emerged: moral psychology of AI and robotics. In this field there are several outstanding questions on how robot appearance and other perceived properties of the robots influences the way their decisions are evaluated. Researchers have observed that robot decision are not treated identically to human decisions, even if their antecedents and consequences are identical to human decisions. To study this moral judgment asymmetry effect further, two studies with a series of high conflict moral dilemmas were conducted: Study 1 – which used photorealistic full body imagery -- revealed that utilitarian decisions by human or non-creepy (i.e., nice) looking robotic agents were less condemned than “creepy” (i.e., unease inducing) robots, whereas “creepy” robots received higher moral approval when making deontological decisions. Furthermore, an exploratory analysis demonstrated that the creepiest robot did not cause moral surprise or disappointment when making utilitarian decisions. However, Study 2 showed that mere symbolic representation of the agent’s face did not trigger the Moral Uncanny Valley (where decisions of creepy robots are perceive negatively), suggesting that the effect is dependent on the photorealistic appearance of the agent. These results are in tension with some previous findings in robot moral judgment literature. Future research should focus on creating standardized stimuli for studying moral decisions involving robots and elucidating the complex interactions between agent appearance, decision type, and pre-decision expectations. This work deepens our understanding of the relationship between a decision-making agent’s appearance and the moral judgment of their decisions. The findings have significant implications for the design and implementation of autonomous agents in morally charged situations.

近年来，道德心理学领域涌现出一个新的子领域：人工智能与机器人道德心理学。在该领域，关于机器人外观及其他感知属性如何影响对其决策的评价方式，存在若干亟待解决的问题。研究者观察到，即使机器人的决策前因后果与人类决策相同，机器人决策亦未受到与人类决策相同的对待。为了进一步研究这种道德判断不对称效应，进行了两项包含一系列高冲突道德困境的研究：研究一——采用逼真的全身影像——发现，由人类或非令人不安（即和善）外观的机器人智能体做出的功利主义决策，相较于“令人不安”（即引起不适）的机器人，受到的谴责较少，而“令人不安”的机器人在做出道义决策时获得了更高的道德认可。此外，一项探索性分析表明，在做出功利主义决策时，最令人不安的机器人并未引起道德上的惊讶或失望。然而，研究二显示，仅仅是对智能体面部的象征性表征并未触发道德的怪异谷效应（即令人不安的机器人决策被负面感知），这表明该效应依赖于智能体的逼真外观。这些结果与机器人道德判断文献中的一些先前发现相矛盾。未来的研究应集中于创建用于研究涉及机器人的道德决策的标准化刺激，并阐明智能体外观、决策类型和决策前的预期之间的复杂相互作用。这项工作深化了我们对于决策智能体外观与其决策道德评价之间关系的理解。这些发现对于在道德敏感情境中自主智能体的设计和实施具有重要的启示。