392241 Applied Cognitive Computing: Deep reinforcement learning with bounded rationality (S) (SoSe 2025)

Information: We will try to hold the seminar in person, if necessary we will switch to live video conferences (the official tool is zoom.us).

Agents (e.g. people) are generally considered to be rational, i.e. that they choose their actions in such a way as to maximize their expected utility. However, finding the optimal action is often infeasible with limited, or bounded resources.
Since all cognitive agents (including) humans, only have limited resources available to them, the best they can do is to achieve what is called „bounded-rationality“, which can lead to (systematic) deviations from „optimal behavior“.

In this seminar, we look more closely into these deviations and the different possible causes for them. We will further discuss ways of dealing with these deviations from the point of artificial systems that either themselves face
an infeasible problem and/or aim to interact with people.