Ashesh Rambachan

My work combines machine learning and AI with ideas from econometrics and statistics about how we learn about the world.

I’m interested in using algorithms to discover things we don’t yet know to look for. What can a model that makes good predictions teach us about the world? I study how to use these models to generate new hypotheses and identify gaps in economic theories. I also study how to assess what the models themselves have learned.

I also study what makes an empirical conclusion credible. I develop methods for causal inference when our assumptions may be wrong, and for evaluating decisions made by people and algorithms. I ask when we can draw reliable conclusions from AI predictions and measurements.