Faithfulness and the Notion of Adversarial Sensitivity in NLP Explanations
This work addresses a critical issue in explainable AI for NLP researchers and practitioners, though it appears incremental as it builds on existing faithfulness evaluation paradigms.
The paper tackles the problem of evaluating faithfulness in NLP explanations by introducing Adversarial Sensitivity, a method that measures explainer response under adversarial attacks, addressing limitations in current evaluation techniques.
Faithfulness is arguably the most critical metric to assess the reliability of explainable AI. In NLP, current methods for faithfulness evaluation are fraught with discrepancies and biases, often failing to capture the true reasoning of models. We introduce Adversarial Sensitivity as a novel approach to faithfulness evaluation, focusing on the explainer's response when the model is under adversarial attack. Our method accounts for the faithfulness of explainers by capturing sensitivity to adversarial input changes. This work addresses significant limitations in existing evaluation techniques, and furthermore, quantifies faithfulness from a crucial yet underexplored paradigm.