Answers are grounded in this paper and cite their sources.
As artificial intelligence service agents (AISAs) become increasingly widespread, anthropomorphism is often proposed as a way to foster user trust, yet empirical findings remain mixed. Drawing on competing trust-building and trust-undermining mechanisms, we conducted a three-level meta-analysis of 211 effect sizes from 136 independent samples ( N = 46,071) to estimate the anthropomorphism–trust association and examine its boundary conditions. The average association was positive and robust to alternative dependence models ( r = 0.462, 95% CI [0.420, 0.502]), but varied substantially across settings (95% prediction interval [−0.131, 0.811]). It was considerably weaker for objectively specified design cues ( r = 0.222) than for perception-based measures ( r = 0.543), most of which captured adjacent trait attributions such as competence and warmth. Restricting outcomes to global trust, reliance, or willingness to depend yielded a similar estimate ( r = 0.475), although construct overlap on the outcome side cannot be ruled out. Cognitive and emotional anthropomorphism were more strongly associated with trust than physical anthropomorphism in aggregate, but this difference was no longer significant after adjusting for operationalization; among design cues, physical and emotional anthropomorphism did not differ significantly, and no design-based cognitive effect sizes were available. The association was stronger in Eastern than in Western samples and in survey than in experimental studies, whereas no statistically significant moderation was detected for gender composition or the prespecified service-context categories. These findings caution against reading the pooled association as a uniform effect of anthropomorphic design and suggest that anthropomorphic cues should complement, rather than substitute for, demonstrable competence and reliability in AISA design and deployment.
No takes yet. Share an insight, caveat, or question.
Meng et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: