Human-centric Explanations for AI Systems

Title:
Human-centric Explanations for AI Systems

Abstract:
The talk begins with an overview on the recent trends in explainable AI, followed by an introduction to our A*STAR AI Initiative efforts on this subject. The talk proceeds with the technical approach towards generating human-comprehensible explanations for certain types of machine learning models. In particular, I will highlight how to implement some key observations learned from everyday explanations that drive this research. In the end, I will also introduce an extension of this work that was incorporated in a new, long-term AI project to offer explanations via dialogues between humans and cobots.

Bio:
Dr. Jiewen Wu is a scientist at the Artificial Intelligence Initiative and the Institute for Infocomm Research under A*STAR, Singapore, where he is leading the research on leveraging domain knowledge to generate explanations for machine predictions. Dr. Wu recently also started working on explainability through dialogue in the manufacturing domain. Before joining A*STAR, Dr. Wu was a leading research scientist at Accenture Labs in Ireland, and, previous to Accenture, a research scientist at IBM Research - Ireland. In his previous roles, he built intelligent solutions to real-world problems across many domains, including transportation, health care, and finance, among others. Dr. Wu received his PhD from the University of Waterloo in 2013.