Understanding challenges to the interpretation of disaggregated evaluations of algorithmic fairness
This addresses a critical issue for practitioners who rely on disaggregated evaluation to assess fairness in machine learning models, highlighting fundamental limitations and proposing improvements.
The paper tackles the problem that disaggregated evaluations of algorithmic fairness can be misleading when data reflect real-world disparities or selection bias, showing that equal performance across subgroups is unreliable and alternative approaches may be invalid without causal assumptions. It proposes complementing disaggregated evaluations with causal analysis to control for confounding and distribution shift.
Disaggregated evaluation across subgroups is critical for assessing the fairness of machine learning models, but its uncritical use can mislead practitioners. We show that equal performance across subgroups is an unreliable measure of fairness when data are representative of the relevant populations but reflective of real-world disparities. Furthermore, when data are not representative due to selection bias, both disaggregated evaluation and alternative approaches based on conditional independence testing may be invalid without explicit assumptions regarding the bias mechanism. We use causal graphical models to characterize fairness properties and metric stability across subgroups under different data generating processes. Our framework suggests complementing disaggregated evaluations with explicit causal assumptions and analysis to control for confounding and distribution shift, including conditional independence testing and weighted performance estimation. These findings have broad implications for how practitioners design and interpret model assessments given the ubiquity of disaggregated evaluation.