Provably Robust Model-Centric Explanations for Critical Decision-Making

TL;DR

This paper advocates for a model-centric, SAT-based approach to generate robust explanations for AI models, providing more reliable insights than traditional data-centric methods like LIME and SHAP, especially in critical decision-making scenarios.

Contribution

It introduces a formal SAT-based framework for model-centric explanations, highlighting its advantages over existing data-centric explanation tools in terms of robustness and practical utility.

Findings

Model-centric explanations are more robust than data-centric methods.

SAT formalism enables actionable insights into AI model risks.

Data-centric explanations can be brittle and less reliable.

Abstract

We recommend using a model-centric, Boolean Satisfiability (SAT) formalism to obtain useful explanations of trained model behavior, different and complementary to what can be gleaned from LIME and SHAP, popular data-centric explanation tools in Artificial Intelligence (AI). We compare and contrast these methods, and show that data-centric methods may yield brittle explanations of limited practical utility. The model-centric framework, however, can offer actionable insights into risks of using AI models in practice. For critical applications of AI, split-second decision making is best informed by robust explanations that are invariant to properties of data, the capability offered by model-centric frameworks.