The research group Computational Statistics with Applications in Psychology, led by Prof. Dr. Marjolein Fokkema, investigates how complex statistical, machine learning, and artificial intelligence methods can contribute to scientific understanding of human behavior.

Machine learning models are often valued for their ability to identify complex patterns and generate accurate predictions. However, high predictive performance alone does not necessarily provide researchers with reliable scientific knowledge. Many models are difficult to interpret, and the uncertainty associated with their results is often not adequately quantified. The group develops methods that address these limitations and enable transparent scientific inference with machine learning and AI.

A central focus lies on tree-based methods and ensemble approaches. Individual decision trees are based on simple and understandable rules, but they may not be flexible enough to capture the nuances of human behavior. Combining many trees can improve predictive accuracy and model complex relationships, but it also makes the resulting models more difficult to interpret. The group examines how the flexibility of these methods can be combined with understandable results and valid uncertainty quantification.

Further core interest of the group are methods for multi-source data, subgroup detection and psychological assessment (also known as psychometrics). By studying and developing methods for detecting heterogeneous patterns in data, the group aims to improve the analysis of psychological and behavioral processes. Applications include psychological assessment, mental health, education, personalized treatment, and other fields in which understanding individual differences is essential.

The group combines methodological research with the development of open-source statistical software. Its tools are designed to make new methods accessible to researchers and support reproducible data analysis. Through interdisciplinary collaboration with psychologists, statisticians, computer scientists, and researchers from applied fields, the group contributes to data science that is not only powerful, but also transparent about its assumptions, limitations, and uncertainty.