Judged by humans, comforted by machines: How psychological expectations shape experiences and perceptions of AI support

Key Findings

• — Six studies, five preregistered, total N = 3,341, plus a pilot showing people hold different initial expectations of AI's versus humans' experiential capacity
• — Receiving support from AI rather than a human lowered fear of negative evaluation, which in turn improved well-being outcomes (Studies 1 to 3)
• — People perceived AI as exceeding their expectations for empathy, which drove perceived support effectiveness (Studies 4 to 6)
• — Effects emerged in both controlled simulations and real interactions with actual human and AI support givers
• — The paper frames initial psychological expectations about AI's experiential capacity as the central mechanism shaping AI support experiences

Methodology Notes

Journal of Personality and Social Psychology, DOI 10.1037/pspa0000494, published online 2026-09-03 (Crossref and PubMed ArticleDate; PMID 42691276), no volume or issue yet. Affiliations from PubMed: Darden School of Business, University of Virginia, and Warrington College of Business, University of Florida. The support context is minor everyday aversive experiences, not clinical distress; the abstract was read via PubMed and the article is paywalled, so preregistration and sample details beyond the abstract are not independently confirmed.

Sources