Rashomon Capacity: A Metric for Predictive Multiplicity in Classification
This addresses the issue of arbitrary and unjust decisions in high-stakes applications such as lending and criminal justice, offering a tool for data scientists to measure and report predictive multiplicity before deployment.
The paper tackles the problem of predictive multiplicity in classification, where models with similar performance give conflicting predictions for individual samples, and introduces Rashomon Capacity as a new metric to measure this in probabilistic classifiers, demonstrating its application across various datasets and models like neural networks.
Predictive multiplicity occurs when classification models with statistically indistinguishable performances assign conflicting predictions to individual samples. When used for decision-making in applications of consequence (e.g., lending, education, criminal justice), models developed without regard for predictive multiplicity may result in unjustified and arbitrary decisions for specific individuals. We introduce a new metric, called Rashomon Capacity, to measure predictive multiplicity in probabilistic classification. Prior metrics for predictive multiplicity focus on classifiers that output thresholded (i.e., 0-1) predicted classes. In contrast, Rashomon Capacity applies to probabilistic classifiers, capturing more nuanced score variations for individual samples. We provide a rigorous derivation for Rashomon Capacity, argue its intuitive appeal, and demonstrate how to estimate it in practice. We show that Rashomon Capacity yields principled strategies for disclosing conflicting models to stakeholders. Our numerical experiments illustrate how Rashomon Capacity captures predictive multiplicity in various datasets and learning models, including neural networks. The tools introduced in this paper can help data scientists measure and report predictive multiplicity prior to model deployment.