We introduce a profit-based measure of lending discrimination in loan pricing and apply it to approximately 80,000 personal loans from a major U.S. fintech platform.

We show that incorporating race and ethnicity into disease risk predictions substantially improves statistical accuracy, but that these statistical benefits lead to surprisingly modest gains in clinical utility.

We develop a taxonomy of concerns over the fairness of healthcare algorithms and critically examine seven prominent and controversial healthcare algorithms, offering a consequentialist framework for algorithm design.

Results from a large field experiment showing that automated text message reminders reduced warrants for arrest issued for missing court by over 20%.

A new theoretical framework for how to efficiently allocate limited resources to people in an equitable way.

A discussion of the apparent conflicts between formal algorithmic fairness constraints and equitable outcomes, with approaches for practitioners designing more equitable algorithms.