Speaker: Marina Agranov (Caltech)

Webpag<wbr />e: https://agran<wbr />ov.caltech.edu/<wbr />

Title: Deciphering Suboptimal Updating: Complexity, Sequencing, and Framing (with Pellumb Reshidi)

Abstrac<wbr />t: We aim to investigate the underlying reasons for the failure of individuals to adhere to Bayes' rule when updating their beliefs. We do so in a framework in which individuals often exhibit Base-Rate neglect. While our findings offer insights into the mechanisms that lead to Base-Rate neglect, they have broader implications for belief updating in general. We decompose the departure from Bayesian updating into three elements (a) framing and context, (b) timing of information delivery, and (c) complexity of the decision task at hand. We address the impact of task complexity by introducing a novel notion of complexity, wherein tasks that entail more nonlinear calculations are deemed more complex. In a series of controlled experiments, we alter all three elements and estimate the gap between the theoretically predicted and reported posterior beliefs. Through additional treatments, we further explore our notion of complexity and find empirical support for this notion. Moreover, we examine the possibility of counteracting one bias with another and find that the optimal way of releasing information depends on the underlying complexity of the environment.