My research mainly builds on two streams. One, how we can learn and make good decisions when the data we observe are themselves shaped by earlier decisions. The other asks how markets, platforms, and institutions should be designed when people respond to prices, contracts, rules, and new technologies.

Learning & Decision-Making

How can we learn, evaluate, and optimize policies when decisions determine what we observe?

  • Selective Labels and Counterfactuals: Why Diagnostics Are Not Enough

    Neha Sharma and Hamsa Bastani

    What can we learn about counterfactual outcomes when past human decisions determine which outcomes are observed?

  • Learning with Selectively Observed Treatments

    What can we learn when the treatment itself is observed only when the outcome is observed?

  • The Winner's Curse: Why Optimizing Structural Models Overpromises and Underdelivers

    Neha Sharma and Hamsa Bastani

    How should we evaluate a policy generated by an optimized structural model?

  • Rethinking Algorithmic Fairness for Human-AI Collaboration

    Haosen Ge, Neha Sharma, Hamsa Bastani, and Osbert Bastani

    How should algorithmic recommendations account for biased human compliance so the combined human–AI system remains fairer than human decisions alone?

Markets, Platforms & Institutions

How should rules, prices, contracts, and technologies shape participation, access, coordination, and welfare?