Andrea Amelio

While evidence shows that cognitive biases influence individual decisions, economic outcomes often result from interactions among individuals, such as through social learning. This paper investigates the impact of social learning on a broad range of behavioral biases, reflecting economically relevant settings. Experimental evidence shows how social learning can reduce or amplify errors stemming from behavioral biases, affecting group outcomes. This suggests that social learning does not systematically eliminate biases, and can either mitigate or exacerbate their impact in settings such as the interpretation of statistical information or investment decisions.

We study the impact of contingent thinking on belief updating. Engaging in contingent thinking calls for both processing hypothetical information and contrasting multiple contingencies during the belief-updating process. Our experimental findings show that contingent thinking leads to significant deviations from Bayesian updating when signals are asymmetric in diagnosticity. We find that deviations are driven by the diminished perceived informativeness of all hypothetical signals; contrasting contingencies, however, can counteract this distortion for symmetric signals but not for asymmetric ones. These results establish contingent thinking as a distinct source of belief distortions with implications for contingent planning, information acquisition, and information design.

People who observe the same statistical evidence can nevertheless reach different conclusions because they differ in inference quality. This creates scope for beneficial social learning even when another person possesses no additional raw information. We study this possibility in a balls-and-urns experiment in which a participant and a peer face the same prior, signal structure, and realized signal. After reporting an initial posterior, participants observe the peer’s posterior and may revise. Final beliefs are more accurate on average after peer observation. Relative cognitive uncertainty—the participant’s uncertainty about their own reasoning relative to the uncertainty attributed to the peer—is aligned with objective relative accuracy. Participants also place more weight on the peer when relative confidence favors the peer, and objectively better peer posteriors receive more than twice as much weight as worse ones. A compound-inference block provides further evidence for this account: when participants become more uncertain relative to a peer solving the corresponding reduced problem, they rely substantially more on the peer, learning gains increase, and peer quality becomes correspondingly more important for final accuracy. The results show that common information does not eliminate the value of social learning: heterogeneous inference creates gains from observing others, and the calibration of relative confidence determines whether those gains are realized.

Motivated reasoning refers to the idea that people hold certain beliefs about themselves or the world due to their desire to do so, rather than striving for accuracy. This type of belief formation can lead to overconfidence and polarization, as well as facilitate immoral behavior at both the individual and collective levels. One of the supply-side mechanisms for motivated reasoning is motivated memory, or the selective retrieval of past experiences or information based on self-serving criteria. In this article, we review the still young economics literature on motivated memory. Summarizing both theoretical and empirical work, we highlight key results this literature has produced. We also discuss open questions and potentially exciting avenues for future research in this area.