Rule of Thumb: Explaining Artificial Intelligence Systems using Partial Information

2026-08-11

AI Summary

This work addresses the challenge of explaining AI decisions in complex scenarios—such as model-agnostic settings, zero-shot learning, and scientific discovery—by proposing a “Rule of Thumb” (RoT) approach. RoT generates concise, intuitive local explanations by identifying the most critical features for a given prediction. It introduces partial information modeling into explainable AI (XAI) for the first time, enabling effective auditing of black-box models, interpretation of zero-shot large language models, and support for scientific discovery, all while aligning with prevailing AI regulations. The method employs a model-agnostic feature importance assessment coupled with efficient computational strategies, achieving high fidelity without sacrificing speed. To facilitate reproducibility and real-world deployment, the authors have open-sourced the implementation.

Abstract

Explainable Artificial Intelligence (XAI) seeks to explain how an Artificial Intelligence (AI) system arrived at a particular decision. We propose ''Rule of Thumb'' (RoT) explanations, a new approach to XAI based upon a novel formulation that identifies the most relevant features for predicting the behaviour of an AI system, for a particular datapoint. We show how RoT is well-suited to enable XAI in: (a) zero-shot classification using large language models (LLMs), (b) auditing of opaque AI systems without model access, and (c) the use of AI in scientific discovery. Additionally, RoT meets specific requirements from leading AI regulations, provides a familiar interface and visualisations for XAI practitioners, is model-agnostic, and is substantially faster than alternatives. Code available at: https://github.com/KaiRawal/Rule-of-Thumb-Explaining-Artificial-Intelligence-Systems-using-Partial-Information

Problem

• Explainable Artificial Intelligence
• AI decision explanation
• model-agnostic explanation
• zero-shot classification
• AI auditing

Innovation

• Rule of Thumb
• Explainable AI
• Model-agnostic
• Zero-shot classification
• Feature relevance