Resolving Predictive Multiplicity for the Rashomon Set

TL;DR

This work tackles predictive multiplicity arising from the Rashomon set in binary classification by introducing three reconciliation strategies: Outlier Correction, Local Patching, and Pairwise Reconciliation. These methods can be used alone or in combination to reduce ensemble disagreement while maintaining competitive predictive performance, and reconciled predictions can be distilled into an interpretable meta-model. Across four high-stakes datasets, the approaches achieve substantial reductions in multiplicity metrics (Variance, Ambiguity, Discrepancy, Disagreement Rate) and improve neighborhood calibration (LCAE) with limited or no accuracy losses. The practical impact lies in producing more stable, fair, and accountable predictions without sacrificing overall task performance, enabling safer deployment in settings like lending, policing, and clinical decision support.

Abstract

The existence of multiple, equally accurate models for a given predictive task leads to predictive multiplicity, where a ``Rashomon set'' of models achieve similar accuracy but diverges in their individual predictions. This inconsistency undermines trust in high-stakes applications where we want consistent predictions. We propose three approaches to reduce inconsistency among predictions for the members of the Rashomon set. The first approach is \textbf{outlier correction}. An outlier has a label that none of the good models are capable of predicting correctly. Outliers can cause the Rashomon set to have high variance predictions in a local area, so fixing them can lower variance. Our second approach is local patching. In a local region around a test point, models may disagree with each other because some of them are biased. We can detect and fix such biases using a validation set, which also reduces multiplicity. Our third approach is pairwise reconciliation, where we find pairs of models that disagree on a region around the test point. We modify predictions that disagree, making them less biased. These three approaches can be used together or separately, and they each have distinct advantages. The reconciled predictions can then be distilled into a single interpretable model for real-world deployment. In experiments across multiple datasets, our methods reduce disagreement metrics while maintaining competitive accuracy.

Figures (3)

• Figure 1: Illustration of approaches for regression.
• Figure 2: Accuracy-disagreement trade-offs (mean $\pm$ std over 10 seeds; $\Delta\%$ vs. Soft Voting).
• Figure 3: Neighborhood agreement via $\mathrm{LCAE@30}$ (mean $\pm$ std over 10 seeds).