Theory of Mind in Artificial Intelligence: Agents Modelling Other Agents

Meet DTU Compute - Theory of Mind in Artificial Intelligence: Agents Modelling Other Agents. Talk by Assoc prof. Thomas Bolander, DTU Compute.

In order for an artificial intelligence system (an agent) to collaborate efficiently with other artificial intelligence systems or humans (other agents), it must be able to take the perspective of those other agents. It must be able to reason about how the world is perceived by those other agents, what they believe and what they intend to do, including higher-order reasoning (I know you believe I intend to get lunch from the Japanese street food stand). To create agents capable of that kind of reasoning, I use algorithms, logic and graphs, the three areas forming the name of the section I work in, Algolog.

In my talk, I will present the general problem of modelling other agents and the mathematical frameworks I use to model it. Then I will talk more specifically about my recent work on implicit coordination: how a group of agents (e.g. robots) can achieve their goals without any explicit negotiation or coordination. Such implicit coordination is often observed in human helpfulness: we infer the goals of others and help them achieve those goals without explicit communication or coordination (e.g. holding the door or passing the butter).

"Meet DTU Compute" is a series of lectures that offers friendly introductions to hot research topics at DTU Compute given by local experts. The lectures are co-organized together with the DTU Compute Research Academy and DTU Compute PhD School.

Each session starts with lunch (sandwiches) 12:30-13, and for that reason it is necessary to sign up by e-mail to Pia Lauridsen. Everybody is welcome to attend.



ons 21 mar 18
12:30 - 14:00


DTU Compute



DTU Compute, Building 324/040