UCLIC Research Seminar Series
Convolutional neural networks (CNNs) offer great machine learning performance over a range of applications, but their operation is hard to interpret, even for experts. Various explanation algorithms have been proposed to address this issue, yet limited research effort has been reported concerning their user evaluation. In this talk, I will share the findings of our recent IUI paper that investigates this area. In this paper, we report on an online between-group user study designed to evaluate the performance of "saliency maps" - a popular explanation algorithm for image classification applications of CNNs. Our results indicate that saliency maps produced by the LRP algorithm helped participants to learn about some specific image features the system is sensitive to. However, the maps seem to provide very limited help for participants to anticipate the network's output for new images. Drawing on our findings, we highlight implications for design and further research on explainable AI. In particular, we argue the HCI and AI communities should look beyond instance-level explanations.
Ahmed Alqaraawi, Martin Schuessler, Philipp Weiß, Enrico Costanza, and Nadia Berthouze. 2020. Evaluating saliency map explanations for convolutional neural networks: a user study. In Proceedings of the 25th International Conference on Intelligent User Interfaces (IUI '20). Association for Computing Machinery, New York, NY, USA, 275-285. DOI:https://doi.org/10.1145/3377325.3377519
Ahmed AlQaraawi is a PhD student at UCL's Interaction Centre (UCLIC). His main research is centred around the area of "explainable AI". He is particularly interested in developing and evaluating explanation techniques and tools that target non-ML experts and to explore how interactive techniques can be exploited in that space. Before joining UCLIC, he completed an MS degree in Electrical Engineering at the University Of Southern California (USC), then worked at King Abdulaziz City for Science and Technology (KACST) in a couple of research and development projects in the field of signal processing and machine learning.