Human-grounded Evaluations of Explanation Methods for Text Classification

Proceedings of the 2019 Conference on Empirical Methods in Natural Language Processing and the 9th International Joint Conference on Natural Language Processing, pages 5195–5205, Hong Kong, China, November 3–7, 2019. c⃝2019 Association for Computational Linguistics 5195 Human-grounded Evaluations of Explanation Methods for Text Classification Piyawat Lertvittayakumjorn and Francesca Toni Department of Computing Imperial College London, UK {pl1515, ft}@imperial.ac.uk Abstract Due to the black-box nature of deep learn- ing models, methods for explaining the mod- els’ results are crucial to gain trust from hu- mans and support collaboration between AIs and humans. In this paper, we consider sev- eral model-agnostic and model-specific expla- nation methods for CNNs for text classifica- tion and conduct three human-grounded eval- uations, focusing on different purposes of ex- planations: (1) revealing model behavior, (2) justifying model predictions, and (3) help- ing humans investigate uncertain predictions. The results highlight dissimilar qualities of the various explanation methods we consider and show the degree to which these methods could serve for each purpose. 1 Introduction Explainable Artificial Intelligence (XAI) is aimed at providing explanations for decisions made by AI systems. The explanations are useful for sup- porting collaboration between AIs and humans in many cases (Samek et al., 2018). Firstly, if an AI outperforms humans in a certain task (e.g., Al- phaGo (Silver et al., 2016)), humans can learn and distill knowledge from the given explanations. Secondly, if an AI’s performance is close to hu- man intelligence, the explanations can increase humans’ confidence and trust in the AI (Symeoni- dis et al., 2009). Lastly, if an AI is duller than humans, the explanations help humans verify the decisions made by the AI and also improve the AI (Biran and McKeown, 2017). One of the challenges in XAI is to explain prediction results given by deep learning mod- els, which sacrifice transparency for high predic- tion performance. This type of explanations is called local explanations as they explain individ- ual predictions (in contrast to global explanations which explain the trained model independently of any specific prediction). There have been several methods proposed to produce local explanations. Some of them are model-agnostic, applicable to any machine learning model (Ribeiro et al., 2016; Lundberg and Lee, 2017). Others are applicable to a class of models such as neural networks (Bach et al., 2015a; Dimopoulos et al., 1995) or to a specific model such as Convolutional Neural Net- works (CNNs) (Zhou et al., 2016). With so many explanation methods available, the next challenge is how to evaluate them so as to choose the right methods for different settings. In this paper, we focus on human-grounded evalu- ations of local explanation methods for text classi- fication. Particularly, we propose three evaluation tasks which target different purposes of explana- tions for text classification – (1) revealing model behavior to human users, (2) justifying the predic- tions, and (3) helping humans investigate uncer- tain predictions. We then use the proposed tasks to evaluate nine explanation methods working on a standard CNN for text classification. These ex- …

Content truncated to 500 words