CS Tea Talk 4/6: Evaluating Explainable AI for Human-Centric Data - presented by Vinitra Swamy
Neural networks are ubiquitous in applied machine learning. Their pervasive success in predictive performance comes alongside a severe weakness, the lack of explainability of their decisions, especially relevant in human-centric fields like education. We implement five state-of-the-art methodologies for explaining black-box machine learning models (LIME, PermutationSHAP, KernelSHAP, DiCE, CEM) and examine the strengths of each approach on the downstream task of student performance prediction for five massive open online courses. Our experiments demonstrate that the families of explainers do not agree with each other on feature importance for the same Bidirectional LSTM models with the same representative set of students. We use Principal Component Analysis, Jensen-Shannon distance, and Spearman's rank-order correlation to quantitatively cross-examine explanations across methods and courses. Furthermore, we validate explainer performance across curriculum-based prerequisite relationships and examine the trustworthiness of explainers in 26 semi-structured interviews with human experts in this domain (professors!). Our results come to the concerning conclusion that the choice of explainer is an important decision and is in fact paramount to the interpretation of the predictive results, even more so than the course the model is trained on. Source code and models are released on Github. The speaker, Vinitra Swamy, is a third year PhD at EPFL in Switzerland, focusing on neural network explainability and transfer learning. Previously, she served as a machine learning lecturer at UC Berkeley and the University of Washington and an AI engineer at Microsoft AI.