We bring together some key findings from multiple experimental studies involving treatments where agents must learn about risks they face from experience through some form of sequential sampling.
In the first part of the talk, we revisit the widely studied description-experience (DE) gap in individual choice via an experiment investigating the causal role of sampling bias, ambiguity sensitivity, and aspects of cognition. Using a model-free approach, we elicit a DE gap similar in direction and size to the literature’s average and find that when each factor is considered in isolation, sampling bias is the only significant driver of the gap. Model-mediated analysis reveals that rare events are generally overweighted and suggests the presence of a smaller DE gap even in the absence of sampling bias.
We then discuss an investigation of the DE gap in the domain of social uncertainty, where (in an experience treatment) individuals learn about the cooperativeness of another agent. Contrary to expectations from the individual uncertainty literature, this shows that that conditional cooperators are more sensitive to rare, cooperative, events in experience relative to description. We illustrate how stronger priors under social than under individual uncertainty can account for this disparity.
In the last part of the talk, we explore decisions from experience in a setting where individuals must select one from a (possibly large) set of lotteries. The potential payoffs are initially hidden, but can be discovered through time-limited search. In this setting, which is designed to mimic aspects of decisions in consumer markets with a product space that has to be discovered, we find evidence of ‘choice overload’ arising from sub-optimal search.