A hierarchical decomposition for explaining ML performance discrepancies
This work addresses the need for more targeted interventions to close ML performance gaps across domains, offering a method that is more detailed than existing approaches but is incremental in its methodological advancement.
The paper tackles the problem of understanding why machine learning algorithms perform differently across domains by introducing a nonparametric hierarchical framework that provides both aggregate and detailed variable-level decompositions of performance gaps, without requiring causal knowledge, and includes debiased estimators and statistical inference for valid confidence intervals.
Machine learning (ML) algorithms can often differ in performance across domains. Understanding $\textit{why}$ their performance differs is crucial for determining what types of interventions (e.g., algorithmic or operational) are most effective at closing the performance gaps. Existing methods focus on $\textit{aggregate decompositions}$ of the total performance gap into the impact of a shift in the distribution of features $p(X)$ versus the impact of a shift in the conditional distribution of the outcome $p(Y|X)$; however, such coarse explanations offer only a few options for how one can close the performance gap. $\textit{Detailed variable-level decompositions}$ that quantify the importance of each variable to each term in the aggregate decomposition can provide a much deeper understanding and suggest much more targeted interventions. However, existing methods assume knowledge of the full causal graph or make strong parametric assumptions. We introduce a nonparametric hierarchical framework that provides both aggregate and detailed decompositions for explaining why the performance of an ML algorithm differs across domains, without requiring causal knowledge. We derive debiased, computationally-efficient estimators, and statistical inference procedures for asymptotically valid confidence intervals.