This part looks alike this: identifying important parts of explained instances and prototypes

TL;DR

This paper introduces a method to identify the most relevant features within prototypes to enhance interpretability and diversity in prototype-based explanations, leading to better user understanding without sacrificing accuracy.

Contribution

It proposes a novel technique to highlight important overlapping features in prototypes and integrates feature importance into prototype selection, improving interpretability and diversity.

Findings

Improves user comprehension of prototype explanations

Maintains or increases predictive accuracy

Enhances prototype diversity through feature importance integration

Abstract

Although prototype-based explanations provide a human-understandable way of representing model predictions they often fail to direct user attention to the most relevant features. We propose a novel approach to identify the most informative features within prototypes, termed alike parts. Using feature importance scores derived from an agnostic explanation method, it emphasizes the most relevant overlapping features between an instance and its nearest prototype. Furthermore, the feature importance score is incorporated into the objective function of the prototype selection algorithms to promote global prototypes diversity. Through experiments on six benchmark datasets, we demonstrate that the proposed approach improves user comprehension while maintaining or even increasing predictive accuracy.