Pieter Verbeke

Postdoc
Department of Experimental Psychology
Ghent University - Belgium
pjverbek.verbeke@ugent.be

Latest news:

The Compass toolbox is now published in Behavior Research Methods!
https://doi.org/10.3758/s13428-023-02165-7
How much data do we need to answer our research question?
For standard F- or T-tests, a closed-form solution and useful toolboxes (e.g., G-power) are available.
However, when using computational models this question gets more complex. Therefore, we propose a more general definition of statistical power and accompanying Python-based toolbox to perform power computations under the Rescorla-Wagner model of decision making.
** We keep working on this project! **
Since acceptance an extra module has been added to allow for fast parameter estimations with the RW model on your own data.
Also, we are working hard to extend the toolbox to the drift diffusion model.

Also read our preprint: Humans adaptively select different computational strategies in different learning environments.
doi.org/10.1101/2023.01.27.525944
In this paper, we evaluated multiple models of different hierarchical complexity on multiple reinforcement learning problems (also of different complexity).
Results demonstrate that different environments are best solved with different learning strategies; and that humans adaptively select the optimal learning strategy.

About me

I'm a Postdoctoral researcher at the Department of Experimental Psychology at Ghent University.
My research focusses on how humans and artificial agents can balance shared versus separated task representations to optimize continual learning.
Here, separated representations are useful to avoid (catastrophic) interference and shared representations are useful to speed up learning via generalization.
In our project, we argue that humans learn compositional/ modular task representations.
At a hierarchically higher level, relations between tasks are learned and used to decide which lower-level modules get control over behavior.
The appropriate modules can be bound via the synchronization of oscillations (in biological agents) or via multiplicative gating (in artificial agents).
To investigate this, we use multiple tools such as computational modelling, EEG, fMRI and behavioral studies.
All of this is done with intellectual support of Tom Verguts and his research team.
When I'm not busy doing research, I try to spend most of my time with my family (2 kids, my girlfriend and a dog).
If there is still some time left, I try to keep up to date with the latest news on sports (mainly cycling) and belgian politics.