From Patterns to Principles: Using Machine Learning to Construct Social Scientific Theories 

Theories reflect scientists' understanding of phenomena. Ideally, they are unambiguous, and continuously updated based on new findings. However, social scientific theories can be sufficiently ambiguous to accommodate contradictory findings. This limits progress. But how can we construct better theories? Patterns in data are often an inspiration, the way a falling apple inspired Newton’s gravitational theory. This Vidi-project develops methods to translate data patterns, detected with machine learning, into unambiguous formal theories, and make those theories shareable so they can be updated (by others). As proofs-of-concept, we construct formal theories of adolescent emotion regulation, moral concern, and cooperation with existing collaborators.

• dr. Caspar van Lissa