Bachelor and Master Theses

Investigating pupil-linked arousal responses to changes in environmental statistics

A growing body of study demonstrates that humans track statistical regularities in their environments and build models to cope with uncertainties in the world. However, as with every model, mental models of agents encounter uncertainties due to changes in environmental statistics. Uncertainties offer risk and/or opportunity, therefore mandating arousal, and require internal models of environmental regularities to be either updated or changed with an alternative. LC-NE is thought to regulate such arousal and model updating and replacement in the brain. Since LC-NE also has a correlational relationship with pupil dilation responses, those responses can be used to infer the time-dependent activity of the LC-NE while uncertainties occur.

In this bachelor thesis, you will present auditory and/or visual patterns to participants and measure their pupil responses to changes in environmental statistics as a result of the model maintenance process.

You will work with Hamit Basgol in order to design a new experiment and analyze its results in Python/Psychopy. If you are interested, please get in touch with Hamit Basgol,

Visual perception and visually-guided actions: How similar or different are they?

Popular theories like the Perception-Action Model / Two-visual-streams-hypothesis (Milner & Goodale, 2008) assume that the same visual information is processed in different ways for the purposes of either visual perception or visually-guided actions respectively. It is also assumed that there are different mental representations involved. In a similar vein, some theories postulate different representations for different phases of an action, planning phase or control phase (Planning-Control Model, see Glover, 2004) Using behavioural experiments, I investigate how similar or different are the underlying processing and representations in perception, action planning and action control.

This Bachelor or Master thesis will involve designing, programming and conducting an experiment and analysing the results using MATLAB and R. Reaction times and movement kinematics will be recorded and analysed.

This project is led by Kriti Bhatia. If interested, please send an email here:, with a few sentences about your background (study program) and time-frame when you would like to work on this project.

Are People Confident When Artifical Neural Networks Are Confident? (with Sascha Meyen)

A recent investigation by Prof. Wichmann and colleagues revealed that state of the art artificial neural networks that are supposed to model the human visual system are disappointingly inconsistent with how humans classify images (Beyond accuracy: quantifying trial-by-trial behaviour of CNNs and humans by measuring error consistency (Geirhos et al., 2020)). To further investigate the gap between artifical neural networks and humans, we replicated their finding. Now, we want to dig deeper in the analysis.

In this Bachelor or Master thesis, you will refine the currently existing analysis (in R) and work on methods to judge how similar two classifiers are. If you are interested, send me an email.

If you want to work with me on other topics, we can try to find something that suits you better. I have a couple of projects and am open for your suggestions as well.