Multicalibrated Regression for Downstream Fairness
This work addresses fairness in machine learning for applications requiring multiple fairness constraints, but it is incremental as it builds on existing multicalibration and omnipredictor concepts.
The paper tackles the problem of efficiently converting a multicalibrated regression function into a classifier that minimizes error while satisfying various fairness constraints, including intersecting groups, without needing labeled data and with modest unlabeled data and computation. The result extends prior work by handling more complex group structures and generalizes omnipredictors to constrained optimization.
We show how to take a regression function $\hat{f}$ that is appropriately ``multicalibrated'' and efficiently post-process it into an approximately error minimizing classifier satisfying a large variety of fairness constraints. The post-processing requires no labeled data, and only a modest amount of unlabeled data and computation. The computational and sample complexity requirements of computing $\hat f$ are comparable to the requirements for solving a single fair learning task optimally, but it can in fact be used to solve many different downstream fairness-constrained learning problems efficiently. Our post-processing method easily handles intersecting groups, generalizing prior work on post-processing regression functions to satisfy fairness constraints that only applied to disjoint groups. Our work extends recent work showing that multicalibrated regression functions are ``omnipredictors'' (i.e. can be post-processed to optimally solve unconstrained ERM problems) to constrained optimization.