On The Stability of Interpretable Models

TL;DR

This paper investigates the stability of interpretable classification models like decision trees and linear models, emphasizing the importance of assessing how data and model choices impact their reliability and accountability.

Contribution

It provides an experimental analysis of the stability of interpretable models with respect to feature, instance, and model selection, highlighting the need for stability impact assessment.

Findings

Interpretable models' stability varies with data and model choices.

Biases in data collection can affect model stability.

Stability assessment is crucial for trustworthy interpretability.

Abstract

Interpretable classification models are built with the purpose of providing a comprehensible description of the decision logic to an external oversight agent. When considered in isolation, a decision tree, a set of classification rules, or a linear model, are widely recognized as human-interpretable. However, such models are generated as part of a larger analytical process. Bias in data collection and preparation, or in model's construction may severely affect the accountability of the design process. We conduct an experimental study of the stability of interpretable models with respect to feature selection, instance selection, and model selection. Our conclusions should raise awareness and attention of the scientific community on the need of a stability impact assessment of interpretable models.