Misinformation – such as fake reviews or biased recommendations – poses a challenge for consumers that can be conceptualized as a two-step problem: First, they need to judge the credibility of the signal. Second, they need to update their prior beliefs conditional on their judgment and the signal realization. This paper reports on an experiment that disentangles both steps utilizing a novel approach that requires neither belief elicitation nor structural estimation of utility parameters. We utilize a mixed design, comparing subjects’ decisions in one of three treatments to their individually optimal decision. The sequential nature of our experiment allows us to design a task that elicits the individually optimal decision directly. While we find no overall treatment effects, we uncover substantial heterogeneity: the interventions improved decision accuracy for male participants and for those with higher cognitive ability and education, but had no significant effect on others. • Misinformation complicates decision-making for consumers in many markets. • Experiment disentangles errors in belief formation and Bayesian updating. • No significant effects in the pooled sample. • Effects are heterogeneous across several dimensions. • Findings inform policies to counteract misinformation effects.
Knapp et al. (2026) studied this question.