Unprocessing Seven Years of Algorithmic Fairness
This work addresses methodological errors in fairness research, providing a critical evaluation for researchers and practitioners in machine learning and AI ethics.
The authors tackled the problem of evaluating algorithmic fairness methods by empirically testing claims of improvement over a seven-year-old postprocessing baseline, finding that the fairness-accuracy Pareto frontier achieved by postprocessing contains all other evaluated methods.
Seven years ago, researchers proposed a postprocessing method to equalize the error rates of a model across different demographic groups. The work launched hundreds of papers purporting to improve over the postprocessing baseline. We empirically evaluate these claims through thousands of model evaluations on several tabular datasets. We find that the fairness-accuracy Pareto frontier achieved by postprocessing contains all other methods we were feasibly able to evaluate. In doing so, we address two common methodological errors that have confounded previous observations. One relates to the comparison of methods with different unconstrained base models. The other concerns methods achieving different levels of constraint relaxation. At the heart of our study is a simple idea we call unprocessing that roughly corresponds to the inverse of postprocessing. Unprocessing allows for a direct comparison of methods using different underlying models and levels of relaxation.