Summit
2025 APS Global Psychological Science Summit · 2025
Theory of Mind about Large Language Models: Experiencing and Thinking Agent or Glorified Autocomplete?
- Brian Stone
Boise State University
Abstract
People (n=529) ascribe mental capacities to LLMs: ChatGPT was rated higher in the factor Thinking than infants, animals, and robots but not adults, and lower in the factor Experience than all except robots. Manipulating news story framing (anthropomorphic vs. computational metaphor) had no effect on theory of mind about LLMs.
Human Factors