WISCA: A Consensus-Based Approach to Harmonizing Interpretability in Tabular Datasets

TL;DR

This paper introduces WISCA, a novel consensus method that harmonizes interpretability explanations across models and datasets, enhancing reliability in high-stakes domains.

Contribution

WISCA is a new approach that combines class probability and normalized attributions to improve interpretability consensus in machine learning.

Findings

WISCA aligns closely with the most reliable interpretability method.

Consensus explanations improve explanation reliability in synthetic datasets.

WISCA outperforms individual interpretability techniques in consistency.

Abstract

While predictive accuracy is often prioritized in machine learning (ML) models, interpretability remains essential in scientific and high-stakes domains. However, diverse interpretability algorithms frequently yield conflicting explanations, highlighting the need for consensus to harmonize results. In this study, six ML models were trained on six synthetic datasets with known ground truths, utilizing various model-agnostic interpretability techniques. Consensus explanations were generated using established methods and a novel approach: WISCA (Weighted Scaled Consensus Attributions), which integrates class probability and normalized attributions. WISCA consistently aligned with the most reliable individual method, underscoring the value of robust consensus strategies in improving explanation reliability.