Embedded Files
Dr. Ilaria Tiddi is an Assistant Professor in Hybrid Intelligence at the Knowledge in AI (KAI) group of the Vrije Universiteit Amsterdam (NL). Her research focuses on using Knowledge Graphs (KGs) to support humans when collaborating with AI systems. She applies this in Cognitive Robotics and Scientific Discovery applications. Dr. Tiddi is the Editor-in-Chief of the CEUR-WS.org publication, is part of the Steering Committee for the Knowledge Capture (KCAP) and Hybrid Human-AI (HHAI) conferences. Coordinator of the international Staff Exchange for the Dutch Hybrid Intelligence consortium. She is active in the OCs/PCs of the major venues in Symbolic AI (ISWC/ESWC, HHAI, WWW, CIKM, IJCAI/ECAI).
Knowledge Graphs for eXplainable AI: past, present and future directions
Knowledge Graphs for eXplainable AI: past, present and future directions
The Deep Learning revolution has impacted the way we explain facts and interpret data in daily applications. This success comes at the cost of models with limited trustworthiness, as well as high energy and computational demands. Knowledge Graphs can help to tackle these shortcomings, supporting a more traceable and transparent creation of explanations that accompany the models' workings. In this talk, we will discuss on the past, present and future of the XAI-KG combination — what it is still useful and relevant, which methods still apply, and which new research directions we should aim for. We will use the case-study of Hybrid Intelligence systems where mixed teams of humans and AI agents jointly collaborate to solve tasks, showing how KGs can enable the design of transparent and accountable solutions to handle failures and unexpected events.
Page updated
Google Sites
Report abuse