Extinction learning is a complex process in which we learn that previously acquired knowledge is no longer relevant. Accordingly, our behavior becomes adjusted. The process of extinction learning therefore involves not only forgetting old information, but also a new learning process. However, the experience of the initial learning is not simply erased; under certain circumstances, the old memory trace may reappear.

Since 2017, researchers in the SFB 1280 “Extinction Learning” have been investigating the fundamentals of extinction learning in 19 projects. In November, it was announced that funding for the research network would be extended for another 3.5 years. Sen Cheng, Professor of Computational Neuroscience, is Principal Investigator in Project A 14 “Modeling context-dependent acquisition and extinction learning” and, together with Onur Güntürkün and Metin Üngör, in the “Focus Group Learning Dynamics” (F01).

Project A 14 “Modeling context-dependent acquisition and extinction learning”

Compared to other forms of learning, extinction learning is highly context-dependent. It is known that the hippocampus plays a decisive role in this process. However, there is a lack of deeper understanding of the learning mechanisms and neuronal foundations underlying context dependence. To investigate what distinguishes individual learning signals from contextual information and where this context dependence comes from, Sen Cheng’s team uses computational modeling and virtual agents. The central assumption is that the brain does not distinguish between context and individual stimuli from the beginning, but that the different roles are due to differences in the statistics of their representations. With the help of a context-oriented task design, this assumption is tested and it is investigated under which conditions a reinforcement learning (RL) model can be made to intrinsically distinguish between stimulus and context.

“Focus Group Learning Dynamics”

The Focus Group Learning Dynamics systematically investigates learning and extinction processes based on the trial-by-trial dynamics of behavior at different levels: intra-individual, inter-individual, between task types, and between species. Behavioral variations are compared with possible model explanations, e.g., differences due to different initial states, stimulus statistics, stochasticity of the response, or due to different parameters or different models that determine the dynamics. The central hypothesis being tested is that there is a connection between the level at which differences are examined and the level of explanation.

In addition, F 01 plays a coordinating role in the SFB and helps to develop shared theories on learning dynamics, context dependency, and ecological implications of extinction learning.