Lifelong Explainer for Lifelong Learners

Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing, pages 2933–2940 November 7–11, 2021. c⃝2021 Association for Computational Linguistics 2933 Lifelong Explainer for Lifelong Learners Xuelin Situ† Sameen Maruf† Ingrid Zukerman† Cecile Paris‡ Gholamreza Haffari† †Department of Data Science and Artificial Intelligence, Monash University, Australia ‡CSIRO Data61, Australia †{firstname.lastname}@monash.edu ‡{firstname.lastname}@data61.csiro.au Abstract Lifelong Learning (LL) black-box models are dynamic in that they keep learning from new tasks and constantly update their parameters. Owing to the need to utilize information from previously seen tasks, and capture commonal- ities in potentially diverse data, it is hard for automatic explanation methods to explain the outcomes of these models. In addition, exist- ing explanation methods, e.g., LIME (Ribeiro et al., 2016), which are computationally expen- sive when explaining a static black-box model, are even more inefficient in the LL setting. In this paper, we propose a novel Lifelong Ex- planation (LLE) approach that continuously trains a student explainer under the supervi- sion of a teacher – an arbitrary explanation al- gorithm – on different tasks undertaken in LL. We also leverage the Experience Replay (ER) mechanism to prevent catastrophic forgetting in the student explainer. Our experiments com- paring LLE to three baselines on text classifi- cation tasks show that LLE can enhance the stability of the explanations for all seen tasks and maintain the same level of faithfulness to the black-box model as the teacher, while be- ing up to 102 times faster at test time. Our ablation study shows that the ER mechanism in our LLE approach enhances the learning ca- pabilities of the student explainer. Our code is available at https://github.com/situsnow/LLE. 1 Introduction Explaining a model’s predictions to practitioners and end users, especially in the case of a black- boxmodel, is non-trivial. Recent research on eX- plainable Artificial Intelligence usually considers feature attribution as a local explanation, i.e., how much each feature contributes to the outcome of the model. Related works include backpropagation- based methods, where the influence of model out- come is backpropagated according to gradients or layer-wise rules (Bach et al., 2015; Sundarara- jan et al., 2017; Smilkov et al., 2017; Erion et al., 2021); perturbation-based methods, which observe changes in model performance after feature per- turbation (Schwab and Karlen, 2019; Kim et al., 2020), or approximate the local decision boundary through perturbed samples (Ribeiro et al., 2016; Lundberg and Lee, 2017); and model-based meth- ods, which train an explainer model by optimizing an explanation-meritorious objective,1 such as ro- bustness/stability (Lakkaraju et al., 2020; Alvarez- Melis and Jaakkola, 2018) that requires similar examples to have similar explanations. All these methods aim to explain static black-box models, whereas explaining dynamic ones, as in the life- long learning (LL) (Silver et al., 2013) setting, is under-explored. We propose a Lifelong Explanation (LLE) ap- proach that learns to explain the outcome of a LL black-box under the supervision of a teacher ex- planation algorithm. The key challenge in LL is to prevent catastrophic forgetting (McCloskey and Cohen, 1989) of knowledge learnt …

Content truncated to 500 words