Pseudoscientific beliefs exert a profound influence on health behaviors, political decisions, and public trust in science, yet research has primarily identified correlates of pseudoscience acceptance rather than the mechanisms by which such beliefs form and update. In this study, we leveraged computational modeling to investigate how normative social feedback shapes pseudoscientific belief revision. A total of 300 US nationally representative participants conducted a learning task where they rated a set of 20 validated pseudoscientific statements while receiving trial-by-trial feedback. Behaviorally, participants showed systematic reductions in prediction errors across trials, consistent with iterative belief updating. Computational model comparison using hierarchical Bayesian inference revealed that learning was best captured by an anchored propagation model, in which prediction errors spread across correlated beliefs but were stabilized by an anchoring parameter reflecting initial convictions. Exploratory analyses further showed that belief updating depended on the alignment between prior expectations and normative feedback, amplifying congruent information and dampening incongruent inputs. These findings provide the first mechanistic account of how pseudoscientific beliefs are simultaneously receptive to new information and resistant to change, offering an integrative framework with implications for research in belief updating, social cognition, and interventions to reduce misinformation.
García-Arch et al. (Sun,) studied this question.