Project
From Patterns to Principles: Machine Learning-Informed Theory Formation Methods for the Social Sciences

Scholars have raised concerns about the state of theory in social science for
nearly 50 years. A key concern here is that an overwhelming 89.6% of social science research is deductive (Kühberger & Scherndl, 2014), but the link between theory and hypothesis is often tenuous (Oberauer & Lewandowsky, 2019); and only 15% of deductive studies even reference specific theories (McPhetres & Albayrak-Aydemir, 2021). Moreover, many theories are sufficiently ambiguous to explain away contradictory evidence, which limits scientific progress (Scheel, 2022; Frankenhuis et al., 2023).

To address these issues, the PhD project will develop algorithmic inductive formal theory construction (AI-FTC) methods, which draft a formal theory based on patterns in quantitative tabular data. This project advances research in psychometrics and social data science, as multiple causal discovery and interpretable machine learning methods will be compared/evaluated throughout this project.

Supervisors
Dr. Caspar van Lissa
Dr. Sophie Hendrikse

Financed by
NWO: vidi grant

Period
1 October 2025 – 31 September 2029