Postdoc in Machine Learning and Interpretability for Cognitive Development 80 %

The successful applicant will collaborate closely with Prof. Dr. Charles Driver, head of the Quantitative Methods unit (Psychology), Prof. Dr. Martin Tomasik, head of the Research Methods unit (Educational Science), and the broader research team. Although the position is primarily research-oriented, there is potential involvement in teaching, statistical consulting, and supervision.
As scientific lead of a cross-institute workgroup, the candidate will develop, train, and compare modern neural network architectures-such as graph neural networks, recurrent neural networks, and hybrid models-to cap-ture learning trajectories, item characteristics, and contextual influences within large-scale cognitive develop-ment and educational testing datasets. They will drive methodological innovation by extending neural network approaches to integrate psychometric principles (for example, item response theory), predict individual change over time, and distinguish item properties.
In addition, the successful applicant will disseminate findings through high-impact journal articles, presentations at international conferences, and by contributing open-source code. They will also co-supervise doctoral and master's students, coordinate regular lab meetings, and foster an inclusive, collaborative culture. Optionally, the appointee may teach up to 2 SWS per semester in quantitative methods, machine learning, or education-al/psychological data science-for which additional remuneration is provided.