A race to belief: How evidence accumulation shapes trust in AI and human informants
Year: 2026
Type: article
Abstract: As AI systems increasingly substitute human informants across factual, social, and professional domains, understanding how people calibrate trust between these agents has become a pressing theoretical and applied question. Yet the cognitive mechanisms underlying context-dependent preferences for AI versus human informants remain poorly understood: it is unclear whether such preferences reflect prior biases, differences in deliberative caution, or differential rates of evidence accumulation during decision-making. To address this gap, we applied a Hierarchical Bayesian Sequential Sampling Model (HSSM) to trust decisions made by Colombian university students across epistemic and social scenarios, decomposing choice and response time data into latent cognitive parameters. Results indicate that context-dependent trust preferences are primarily driven by differences in evidence accumulation rate rather than by a priori bias or response caution, with drift rates diverging markedly between domains. These parameters aligned closely with participants’ explicit confidence judgments, supporting the validity of the modeling approach. We interpret these findings through the lens of epistemic vigilance theory, arguing that domain-sensitive vigilance mechanisms modulate how efficiently evidence is accumulated in favor of AI versus human informants. The study offers a mechanistic account of selective trust with implications for AI governance and the design of systems that foster well-calibrated human–AI collaboration.
Language: English
Subfield: Social Psychology
Field: Psychology
Domain: Social Sciences
Open Access status: gold
Estimated APC paid: $1,540
Funder: Universidad de La Sabana
Award: G2736742846
OpenAlex ID: W4416948213