Distribution-Free Statistical Dispersion Control for Societal Applications
This addresses the need for explicit finite-sample guarantees on model performance dispersion to ensure equitable effects in societal applications, representing a novel extension beyond bounding expected loss or individual prediction probabilities.
The paper tackles the problem of controlling statistical dispersion in loss distributions for high-stakes societal applications, proposing a distribution-free framework that handles a richer class of statistical functionals beyond previous work, with verification through experiments in toxic comment detection, medical imaging, and film recommendation.
Explicit finite-sample statistical guarantees on model performance are an important ingredient in responsible machine learning. Previous work has focused mainly on bounding either the expected loss of a predictor or the probability that an individual prediction will incur a loss value in a specified range. However, for many high-stakes applications, it is crucial to understand and control the dispersion of a loss distribution, or the extent to which different members of a population experience unequal effects of algorithmic decisions. We initiate the study of distribution-free control of statistical dispersion measures with societal implications and propose a simple yet flexible framework that allows us to handle a much richer class of statistical functionals beyond previous work. Our methods are verified through experiments in toxic comment detection, medical imaging, and film recommendation.