SCENE: Evaluating Explainable AI Techniques Using Soft Counterfactuals

TL;DR

SCENE introduces a new evaluation framework using large language models to generate soft counterfactual explanations, providing a standardized way to assess the stability and effectiveness of XAI methods in NLP.

Contribution

The paper proposes SCENE, a novel zero-shot evaluation method leveraging LLMs for soft counterfactuals, addressing instability issues in existing XAI techniques like LIME and SHAP.

Findings

SCENE effectively evaluates XAI methods across different NLP architectures.

It reveals strengths and limitations of popular XAI techniques.

The method offers a standardized, contextually meaningful evaluation approach.

Abstract

Explainable Artificial Intelligence (XAI) plays a crucial role in enhancing the transparency and accountability of AI models, particularly in natural language processing (NLP) tasks. However, popular XAI methods such as LIME and SHAP have been found to be unstable and potentially misleading, underscoring the need for a standardized evaluation approach. This paper introduces SCENE (Soft Counterfactual Evaluation for Natural language Explainability), a novel evaluation method that leverages large language models (LLMs) to generate Soft Counterfactual explanations in a zero-shot manner. By focusing on token-based substitutions, SCENE creates contextually appropriate and semantically meaningful Soft Counterfactuals without extensive fine-tuning. SCENE adopts Validitysoft and Csoft metrics to assess the effectiveness of model-agnostic XAI methods in text classification tasks. Applied to CNN, RNN, and Transformer architectures, SCENE provides valuable insights into the strengths and limitations of various XAI techniques.