🌱 I am a computational cognitive scientist.

I am a computational cognitive scientist at the University of Tübingen. My research develops and evaluates process models of learning, generalisation, and decision-making, with a central commitment to model adequacy: whether models capture the right computational mechanisms, not merely the right outcomes. Much of my work approaches this through failure — irrational generalisation across supervised and unsupervised task structures, and the computational principles that persist across radically different learning conditions (10.48550/arXiv.2608.06149). Failures constrain the space of admissible mechanisms for biological intelligence far more sharply than successes do, since many mechanisms produce correct behaviour but few produce the right error under the right conditions. I am increasingly interested in extending this to any learning systems, where the same logic applies: recent work of mine locates an instability in a attention-learning architecture (used for theorising about human learning) and the change that resolves it — a result about the neural network architecture, derived rather than observed (10.48550/arXiv.2609.08615). Analysing learning dynamics in tractable models this way seems to me the most promising route to formal accounts that hold for learning in minds and machines alike.

I develop things mostly for my research, but that doesn't mean that they should only be available to me. 🌝

psp: a tool for cognitive scientists to explore the complete parameter space of computational models and to discover their global behavior - all their predictions.

catlearn: catlearn is an archive of formal models of categorization and learning, plus benchmark datasets to test them against. It's also an archive of simulations using those models. It's free and open-source software … and always will be.

cpm: cpm is a fundamental scientific modelling library for Computational Psychiatry.

Computational Modelling: I am always implementing formal models from Psychology and attempting to replicate simulations, so you don't have to 🌝 Most of my model implementations can be found in catlearn. If you want to make your or someone else's model available for everyone and you think I can help with that, feel free to reach out! 🌝

I will always share what I do for free. You will be able to edit, remix, and use my code and software, given you share yours under the same license. So, if you are using any of my software, consider buying me a coffee ☕ I will go to my favorite coffee shop and continue to code endlessly 💻