Research

Cognitive dissonance is the internal tension individuals feel when their past actions are inconsistent with their beliefs or attitudes. I propose an intrapersonal game to model a decision-maker (DM) who distorts her beliefs to mitigate cognitive dissonance from past choices. Two selves make sequential, observable decisions, with unobservable belief manipulation occurring in the interim stage between them. The subgame perfect Nash equilibria are characterized by tractable axioms on choice patterns, with parameters identifiable from choice data. I demonstrate that the model is a useful tool for studying path dependence in decision-making. It matches a variety of experimental and real-world evidence consistent with cognitive dissonance theory. Applying the model provides new insights into existing topics, including add-on selling, behavioral poverty traps, and the value of information.

In the context of aggregating the preferences of Bayesian experts under subjective uncertainty, Commutativity is a normative criterion stating that the aggregation of expert posteriors should be identical to the update of the aggregated priors. I propose thought experiments that raise questions about the normative appeal of Commutativity and motivate relaxations of the condition. Under suitable assumptions, I establish equivalences between these (non-)commutativity conditions and classic axioms for decision-making under uncertainty, including Independence, C-Independence, and Ambiguity Aversion. I show that one of the relaxations of Commutativity plays the key role in characterizing an aggregation rule which can be viewed as the outcome of a game played by "dual-selves," Pessimism and Optimism. 

I characterize an overprecise updating model, in which an agent systematically overestimates the informativeness of their posterior beliefs. The model captures the intrapersonal trade-off between mitigating risk and avoiding distorted posteriors. My characterization establishes a connection between laboratory evidence and the psychological mechanism driving the desire for uncertainty reduction. I demonstrate that overprecision contributes to cognitive biases such as overinference, confirmation bias, and overoptimism. For an agent receiving a sequence of signals that contradict her initial belief, overprecision leads to initially updating too slowly, followed by overly rapid updates.