Rethinking Attention-Model Explainability through Faithfulness Violation Test
This addresses a critical flaw in explainability for AI practitioners using attention-based models, though it is incremental as it builds on existing explanation techniques.
The paper identified that attention mechanisms in deep models often fail to indicate the polarity of feature impact, meaning high-attention features can suppress predictions rather than contribute positively, and proposed a faithfulness violation test to diagnose this issue, revealing that most explanation methods suffer from it.
Attention mechanisms are dominating the explainability of deep models. They produce probability distributions over the input, which are widely deemed as feature-importance indicators. However, in this paper, we find one critical limitation in attention explanations: weakness in identifying the polarity of feature impact. This would be somehow misleading -- features with higher attention weights may not faithfully contribute to model predictions; instead, they can impose suppression effects. With this finding, we reflect on the explainability of current attention-based techniques, such as Attentio$\odot$Gradient and LRP-based attention explanations. We first propose an actionable diagnostic methodology (henceforth faithfulness violation test) to measure the consistency between explanation weights and the impact polarity. Through the extensive experiments, we then show that most tested explanation methods are unexpectedly hindered by the faithfulness violation issue, especially the raw attention. Empirical analyses on the factors affecting violation issues further provide useful observations for adopting explanation methods in attention models.