My research interest is in the field of epistemic modeling in the context of narratives and games. In particular, I try to model how actors in narratives and players in games acquire beliefs and then act on them. For example, in The Lord of the Rings, the protagonist, Frodo, is to meet Gandalf at an inn, but Gandalf is not there. Before arriving, Frodo’s belief was that Gandalf would be there, and he acted on that belief by going to the inn, but upon learning that Gandalf seems to have disappeared, he has to form a new plan. In the movie, he sees a dark, hooded figure in a corner, and forms a belief that that person is an enemy. Later this character is of course revealed as Aragorn, and Frodo has to revise his belief state again to account for the fact that they are allies. In games, there are also numerous examples, in particular in games with asymmetrical information, like most card games where each player can only see their own cards, and forms beliefs about which cards other player may or may not have over the course of the game.

One narrative genre in which character beliefs and knowledge acquisition is particularly interesting is that of Holmesian detective stories. The prototypical detective story starts by the detective being informed of some crime, and them then working to find the perpetrator. Basically, the detective (and the audience) is presented with the state of the world after the crime has been committed, and then has to figure out the chain of events that lead to that state. To do this, the detective has to plan how to acquire the clues necessary to apprehend the perpetrator. However, the perpetrator wants the detective not to find them. And more importantly, the author wants the detective to find the perpetrator, but not too quickly.

In games, I am interested in the aforementioned games with asymmetrical information, where knowledge acquisition is a key game mechanic. Examples for such games include the cooperative game Hanabi, but also social deduction games such as Werewolf/Mafia, in which players have to deduce the roles of other players.

To represent stories and be able to reason about them, I have developed the logical language Impulse, which can represent actions and events that happened in a story, has a rich representation of time, based on Interval Temporal Logic, and also has a representation for character beliefs, desires and intentions.

I am currently working on an implementation of Baltag’s Logic for Suspicious Players. However, since writing game rules in the logic itself is very cumbersome, I have developed a higher-level language that makes it easier to write them.

A language for Epistemic Actions

A high level language to express actions that update actor's mental models

Hanabi - User Interface and AIs

A browser-based UI and implementation of Hanabi for the development, exploration and evaluation of AIs

Intentional AI for Hanabi

An intentional AI for Hanabi