%0 Conference Proceedings %T Can We Really Trust Explanations? Evaluating the Stability of Feature Attribution Explanation Methods via Adversarial Attack %A Zhao, Yang %A Yuanzhe, Zhang %A Zhongtao, Jiang %A Yiming, Ju %A Jun, Zhao %A Kang, Liu %Y Sun, Maosong %Y Liu, Yang %Y Che, Wanxiang %Y Feng, Yang %Y Qiu, Xipeng %Y Rao, Gaoqi %Y Chen, Yubo %S Proceedings of the 21st Chinese National Conference on Computational Linguistics %D 2022 %8 October %I Chinese Information Processing Society of China %C Nanchang, China %G English %F zhao-etal-2022-really %X “Explanations can increase the transparency of neural networks and make them more trustworthy. However, can we really trust explanations generated by the existing explanation methods? If the explanation methods are not stable enough, the credibility of the explanation will be greatly reduced. Previous studies seldom considered such an important issue. To this end, this paper proposes a new evaluation frame to evaluate the stability of current typical feature attribution explanation methods via textual adversarial attack. Our frame could generate adversarial examples with similar textual semantics. Such adversarial examples will make the original models have the same outputs, but make most current explanation methods deduce completely different explanations. Under this frame, we test five classical explanation methods and show their performance on several stability-related metrics. Experimental results show our evaluation is effective and could reveal the stability performance of existing explanation methods.” %U https://aclanthology.org/2022.ccl-1.82 %P 932-944