Keynotes | K&Ts | GACs | Talks | Posters | Search

Poster B in Poster Session B: Tuesday, August 4, 2:00 – 3:45 pm, Kimmel Center, Shorin & Rosenthal Rooms

How Prior Beliefs Distort the Integration of Information in Sequential Decision-Making

Imogen Scott¹, Lucie Charles¹, Keiji Ota¹, Thibault Gajdos Preuss²; ¹Queen Mary, University of London, ²CNRS

Presenter: Imogen Scott

A central challenge for human cognition is integrating information from sources that vary in reliability with our prior beliefs. Normative Bayesian accounts provide a benchmark on how those should be combined, but it remains unclear how humans compare to this ideal, particularly when unreliable evidence supports prior beliefs. Here, we examined how prior beliefs influence the integration of explicit reliability samples in a sequential decision-making task. Participants observed sequences of evidence samples (red and blue squares), each associated with an explicit reliability expressed as a percentage (50%, 55%, or 65%), and judged which option was more likely to be correct. We manipulated prior belief by introducing a base rate bias in the overall likelihood of each colour that participants could learn through feedback or were explicitly told. We fitted a computational model of sequential evidence integration in log-odds space capturing key sources of suboptimality in the decision process and the learning of base rate. Each sample was assumed to be weighted according to its position in the sequence (recency parameter) and its potentially distorted reliability (probability distortion parameter). Belief about the base rate was updated through biased learning (learning parameter). Participants treated all sources as more reliable than they were, replicating previous findings (Ota et al., 2025). Participants also showed a consistent tendency to choose the option aligned with their belief about the base rate. Awareness of the bias in the base rate was associated with more adaptive belief updating. Importantly, there was no evidence of an interaction between prior belief and evidence reliability, suggesting these distortions operate independently. We also found no evidence for selective amplification of belief-congruent evidence. Together, these findings indicate that prior beliefs bias decision-making primarily by shifting decision criteria rather than altering the weighting of information, revealing multiple, dissociable sources of bias in human inference.

Topic Area: Decision-Making, Cognitive Control & Event Cognition