Toward Trustworthy AI: Making Artificial Intelligence Explainable and Accountable

As Artificial Intelligence systems start making more important decisions such as hiring, medical diagnosis and parole, the question of trust becomes critical. How can we understand the reasoning behind an AI’s decision, and what does it mean for an algorithm to be "transparent"? This Lunch & Learn introduces the idea of Explainable AI (XAI), a growing area of research that seeks to make machine learning more understandable, fair and accountable. Associate professor and AI researcher Mi-Young Kim will discuss the challenges of interpreting complex AI models, the risks of “black box” decisions and the methods that aim to help humans see why AI makes certain predictions. By exploring examples from everyday life, we will consider what it takes to build AI that people can truly trust.

Mi-Young Kim, PhD, Associate Professor of Computing Science, Department of Science, Augustana Faculty, University of Alberta Mi-Young Kim is an associate professor of Computing Science at the Augustana Faculty, University of Alberta. Her research interests include Natural Language Processing (NLP), Artificial Intelligence (AI), and Explainable/Trustworthy AI. She is particularly focused on information extraction in the medical and legal domains.Since 2014, Kim has been a co-organizer of the International Competition on Legal Information Extraction and Entailment (COLIEE). Her team’s legal AI assistant achieved first place in answering Yes/No legal bar exam questions from 2014 to 2019, and again in 2022.In addition, Kim is developing AI systems for automated health assessments within Alberta’s 811 HealthLink telehealth service, designed to generate explanatory rationales for their predictions. She is also conducting research on Alberta radiology reports, extracting information related to inflammatory bowel disease and applying AI-based methods to provide interpretable explanations.