Best Practices for Using AI- and ML-generated Variables in Economics

It has become common practice across all fields of economics for researchers to use variables generated by AI or machine learning methods as inputs to econometric models. However, naively treating AI- and ML-generated variables as "data" leads to biased estimates and invalid inference. This project will help further develop best practices for working with AI- and ML-generated variables. In particular, the project will develop R (one position) and Python (one position) packages for implementing methods to correct bias and perform valid inference.