Towards Faithful Explanations: Boosting Rationalization with Shortcuts Discovery

TL;DR

This paper introduces SSR, a method that enhances neural network explanations by discovering and leveraging shortcuts, while mitigating their misleading influence to produce more faithful rationales.

Contribution

The paper proposes a novel SSR approach that detects shortcuts, mitigates their effects, and improves rationale quality with data augmentation, addressing limitations of existing methods.

Findings

SSR effectively discovers potential shortcuts in data.

Mitigation strategies reduce reliance on shortcuts in rationales.

Experimental results show improved explanation faithfulness.

Abstract

The remarkable success in neural networks provokes the selective rationalization. It explains the prediction results by identifying a small subset of the inputs sufficient to support them. Since existing methods still suffer from adopting the shortcuts in data to compose rationales and limited large-scale annotated rationales by human, in this paper, we propose a Shortcuts-fused Selective Rationalization (SSR) method, which boosts the rationalization by discovering and exploiting potential shortcuts. Specifically, SSR first designs a shortcuts discovery approach to detect several potential shortcuts. Then, by introducing the identified shortcuts, we propose two strategies to mitigate the problem of utilizing shortcuts to compose rationales. Finally, we develop two data augmentations methods to close the gap in the number of annotated rationales. Extensive experimental results on real-world datasets clearly validate the effectiveness of our proposed method.