Reframing the Performance and Ethics of Empathic AI: Wisdom of the Crowd and Placebos

Abstract

Recently, claims have emerged that artificial intelligence (AI) is better at providing empathy than humans. These claims are based on experiments in which large language models were prompted to generate empathic responses to short emotional passages. These responses, as well as analogous responses generated by human participants, were judged by third-party human raters. In several cases, the AI-generated responses were preferred to human responses. Such findings have led to suggestions that people should use empathic AI to supplement human empathy. This article critically examines these positions by drawing analogies to two well-established psychological effects. First, I argue that the apparent preferability of AI-generated empathy reflects an analogue of the “wisdom-of-the-crowd” effect. This reframes the performance of empathic AI in a more mundane and less dehumanizing way. Second, I consider whether people should use AI for empathy. Here I draw an analogy to placebo effects, suggesting that even clear utilitarian benefits may not justify the adoption of empathic AI. Through these analogies between AI and well-known psychological effects, this article equips readers with new conceptual tools for grappling with empathic AI, its performance, and the morality of its use.