Abstract:The rise of powerful Large Language Models (LLMs) provides a compelling opportunity to investigate the consequences of anthropomorphism, particularly regarding how their exposure may influence the way individuals view themselves (self-perception) and other people (other-perception). Using a mind perception framework, we examined attributions of agency (the ability to do) and experience (the ability to feel). Participants evaluated their agentic and experiential capabilities and the extent to which these features are uniquely human before and after exposure to LLM responses. Post-exposure, participants increased evaluations of their agentic and experiential qualities while decreasing their perception that agency and experience are considered to be uniquely human. These results indicate that anthropomorphizing LLMs impacts attributions of mind for humans in fundamentally divergent ways: enhancing the perception of one's own mind while reducing its uniqueness for others. These results open up a range of future questions regarding how anthropomorphism can affect mind perception toward humans.

Large language models have divergent effects on self-perceptions of mind and the attributes considered uniquely human