Presenter: Markus Eger, Assistant Professor in Computer Science at California State Polytechnic University-Pomona

Description: As the use of AI techniques keeps proliferating, more and more humans interact with intelligent agents every day. However, this interaction is not always without friction, as humans have very particular ways of communicating that do not always match what a computer would prefer. An additional challenge for AI agents is that communication happens in many, often implicit, ways, which are not always well-defined or universal.

In this talk, Markus will discuss an approach to improving the interaction between intelligent agents and humans in games. In order to do so, the agent needs a model of its human partner's knowledge and beliefs, and be able to determine long-term and immediate goals of the interaction. He will discuss logical representations of knowledge and belief that can be used by an agent, but also point out challenges that still remain. To actually convey information, the agent can use several different means of realizing its communication. Noting that not all communication is based on speech acts or words, he will present several different modalities, and contexts that an agent can use and operate in. These modalities include explicit communication, but also other means of expression, like architecture and other generative problems. To complete the communication model, the human user also has to be able to provide feedback, to be able to respond to the AI agent in some form. The overall model he will present has applications in many different domains, and he will present various research projects that have already been completed, or are currently in progress, based on this general approach. He will also provide a vision for what remains to be done, and how we might be able to work towards solving some of the challenges that are still open.

Bio: Markus Eger is an assistant professor at Cal Poly Pomona, where he currently mentors six graduate and over 30 undergraduate students on various projects. He earned his Ph.D. from NC State University in 2018, and then spent 2 years in Costa Rica where he was a visiting professor at the University of Costa Rica. His research interests focus around making AI agents for games that communicate and/or interact with humans. He has published articles at AIIDE, FDG, CIG and other venues, and is actively involved in the game AI research community, currently serving as General Chair for AIIDE 2023.

Hosted by: Professor Noah Wardrip-Fruin