Signal Mind perception shapes how people judge AI companionability

Summary

This preprint asks whether perceiving an AI companion as having a mind comes before people accepting it as a companion. In a field-sampled experiment, participants read descriptions that either attributed or denied a mind to an AI companion before rating its companionability. Mind-attributive descriptions somewhat reduced baseline skepticism for connective coordination, but did not change eudaimonic exchange. Pre-interaction beliefs predicted companionability along two dimensions: agentic mind perception predicted eudaimonic potential, while experiential mind perception predicted connective potential. The agentic link was moderated by meaning in life, but only among people already high in existential purpose. The authors read the results as two pathways to companionability, both more accessible to people who already flourish.