Abstract
AI systems designed for human interaction often include anthropomorphic elements, intended to enhance usability and acceptability by tapping into, or exploiting, human social cognition. However, this anthropomorphic design can also lead users to mistakenly view and treat AI as if it has emotions, intentions or moral agency. Common solutions, such as explicitly notifying users they are interacting with an AI, may not effectively prevent these unintended perceptions. This talk critically examines such strategies, highlighting their limitations and emphasizes the need to develop approaches that directly engage with the psychological mechanisms behind our tendency to anthropomorphize AI systems.

Dorna Behdadi is a TAIGA postdoctoral researcher in philosophy. Their current research project investigates the conceptual and normative implications of perceiving and treating AI systems as social and moral agents. Their broader research interests include moral agency and responsibility, moral psychology and applied ethics, particularly the ethical challenges emerging from interactions between humans and artificial intelligence.