Can we tell from the data when and how treatment effects generalize across different individuals and contexts? Exciting new working paper, joint with Emily Breza and Arun G. Chandrasekhar that studies generalizability for applications to external validity, effect heterogeneity and model discovery