• Information Dynamics

    Coasian Dynamics under Informational Robustness

    with Xiaosheng Mu (Last update: April 2022)

     

    We propose an approach to study informationally robust dynamic pricing when the seller lacks commitment.

    Learning versus Unlearning: An Experiment on Retractions

    with Duarte Gonçalves and Jack Willis (Last update: June 2022) New Version!

     

    We conduct an experiment on belief updating and document that subjects under-react to information when it is in the form of a retraction.

    Algorithm Games and Rational Play with Strategic Inference

    with In-Koo Cho (Last update: November 2021)

     

    We exhibit an algorithm which ensures that agents play a Bayesian best reply (approximately).

  • Bounded Rationality

    Evolutionarily Stable (Mis)specifications: Theory and Applications

    with Kevin He (Last update: July 2022) (New version!) Presented @ EC '21

    We introduce a selection criterion on behavioral biases in environments with learning, and show that it need not select for a bias-free worldview in some common applications.

    Learning Underspecified Models 

    with In-Koo Cho (Last update: July 2022) (New Version!)

     

    A monopolist can ensure an optimal price is used in the long run by designing an algorithm that assumes a linear demand curve, even if demand is quite nonlinear.

  • Researcher Incentives

    Research Registries and the Credibility Crisis: An Empirical and Theoretical Investigation

    with Eliot Abrams and John A. List (Last update: September 2021) R&R @ JEEA

     

    We empirically study registries, focusing mostly (but not exclusively) on the AEA RCT Registry, and theoretically discuss issues related to incentives behind registration. The Mathematica notebook referenced in the Appendix can be found here.

  • Other Papers

    Hypothetical Beliefs Identify Information

    (Last update: May 2021) R&R @ JET

     

    I demonstrate how to recover a decisionmaker's information structure from posterior beliefs over states, together with posterior beliefs that each signal could be observed. In the process, I make new observations on the geometric structure of information.

    Iterative Weak Learnability and Multi-Class AdaBoost

    with In-Koo Cho and Cheng Ding (Last update: January 2021) Rejected with encouragement to resubmit @ JMLR

     

    We propose a classification algorithm with generalizes AdaBoost to a multi-class, which requires an intuitive and simple to check condition to be valid.