Stefania Merone

Under what conditions should people verify the trustworthiness of an information source? I consider a Bayesian decision maker who receives signals from sources of uncertain quality, and can pay a cost to learn whether a given source is reliable. I find that verification is only valuable for an intermediate range of prior beliefs for which uncertainty about source reliability is sufficiently high and substantial updating is still possible. More importantly, verification is selective: when unreliable sources are noisy but still informative, signals that contradict prior beliefs are systematically more valuable to verify; hence, asymmetric fact-checking can arise from rational behavior. I then explore departures from Bayesian updating, and show that confirmatory bias attenuates this asymmetry.

Commitment is widely used as evidence of temptation: if individuals restrict their future options, they are presumed to anticipate self-control problems. Empirically, however, commitment is typically measured as a binary decision to remove an option, collapsing rich preference information into a single statistic and risking systematic misclassification of underlying motives. We address this limitation in a field experiment on a food-ordering platform. Participants rank all subsets of three meal categories, and one chosen menu is later implemented for a binding order. Observing both singleton preferences and the full lattice of menus allows us to disentangle normative tastes, flexibility motives, and temptation channels. We show that temptation is identified only relative to normative preferences and that unidimensional self-control models impose sharp restrictions frequently violated in the data.