BCCN PhD lecture series: Stefan Kiebel (Dresden)
Prof. Dr. Stefan Kiebel, Professor of Neuroimaging, Department of Psychology, TU Dresden; member of the SFB940/3 - Volition and Cognitive Control
Join Zoom Meeting
Meeting ID: 849 5334 1083
Humans often make seemingly effortless decisions with short sub-second deadlines, for example when interacting with other people. Even though our environment is dynamic and uncertain, we are able to factor future consequences into these fast decisions. Such decisions are known to require some form of forward planning to identify the sequence of actions, among several alternatives, that will result in a favourable outcome. From research into artificial intelligence we know that such forward planning is computationally expensive. The question is how humans can minimize their computational expense in order to meet short deadlines. We address this question by combining computational modelling with behavioural experiments. Specifically, we use probabilistic inference to show that human behaviour harnesses context-specific decision making. The principled idea is to learn re-occurring contexts and store preferred action sequences or even strategies for each context. I will show that this approach has the potential to explain a large range of well-known phenomena like fast computation of the exploration-exploitation trade-off or the use of habitual behaviour.