Explaining Search Results to Enhance User Experience

Abstract: Information access systems, such as search engines, are part of everyone's daily life and also provide information for decision-making in societally critical tasks, such as healthcare. Recent information access systems, similar to many other systems, have been designed with deep neural rankers and trained using data-driven algorithms. Although these systems have shown high potential in detecting and amplifying biases in training data, they have remained black-box to humans.

In this talk, I will highlight the unique and major challenges of explaining black-box models in the context of users' information needs as well as explaining ranked lists of items. I will then show how search results can be explained to users to help them make better and faster decisions on item relevance. Moving beyond explaining results with respect to one information need, I will talk about how the behavior of black-box models can be explained as a whole. I conclude with what explanations with machine-automated decisions can help to improve users' critical thinking.

Bio: Razieh Rahimi is a postdoc researcher and lecturer in the College of Information and Computer Sciences at the University of Massachusetts Amherst. Prior to UMass, she was a researcher at the Chinese University of Hong Kong. She obtained her Ph.D. from the University of Tehran. Her research focuses on the explanation of search results and designing interpretable ranking models. She has also experience in designing and developing models for knowledge extraction from unstructured text and cross-language information retrieval. Her work has been published in multiple peer-reviewed venues.

To obtain the Zoom information for this talk, please see the announcements on the CICS email lists. For more information on this event, please contact Joyce Mazeski.