Machine Learning in Social Science

Period: 2024-

Social-science theories are often evaluated with a single model specification, which makes it hard to compare competing explanations and to identify where theories fail.

This line uses benchmark-style evaluation, prediction gaps, and large administrative/survey data to connect prediction to explanation: where model performance diverges, we gain substantive clues about hidden mechanisms and subgroup-specific dynamics.

Grants and Funded Projects

• NWO Veni: The Social Contexts of Upward Mobility

  NWO · EUR 320k · 2024-2027

  Role: Principal Investigator

  Population-scale modeling of educational and socioeconomic mobility using social-context networks and predictive methods that bridge explanation and forecasting.

• NWO Open Competition XS: Embedded Lives

  NWO · EUR 50k · 2025-2026

  Role: Principal Investigator

  Graph-based modeling of Dutch population life courses, including embedding methods and reusable infrastructure for context-aware social-data analysis.

• ZonMw BePrepared: Vaccination Decision Profiles

  ZonMw · EUR 146k (UU share) · 2024-2025

  Role: Co-applicant

  Data-driven modeling of vaccination decision profiles and uptake, combining survey beliefs/values with CBS microdata and network structure for targeted interventions.

Selected Publications

1. Prediction Gaps as Pathways to Explanation: Rethinking Educational Outcomes through Differences in Model Performance

   J Garcia-Bernardo, E Jaspers, W Machado, S Plach, EJ van Leeuwen (2025). arXiv preprint.

   Code PDF

2. Population-Scale Network Embeddings Expose Educational Divides in Network Structure Related to Right-Wing Populist Voting

   M Luken, J Garcia-Bernardo, S Deb, F Hafner, M Khosla (2025). arXiv preprint.

   PDF

3. Using Large Language Models for Text Annotation in Social Science and Humanities: A Hands-On Python/R Tutorial

   Q Fang, J Garcia-Bernardo, EJ van Kesteren (2025). OSF preprint.

   Software PDF

4. Combining the strengths of Dutch survey and register data in a data challenge to predict fertility (PreFer)

   E Sivak, P Pankowska, A Mendrik, T Emery, J Garcia-Bernardo, et al. (2024). Journal of Computational Social Science.

   PDF

5. The potential of benchmark challenges in the social sciences

   P Pankowska, A Mendrik, T Emery, J Garcia-Bernardo (2024). Social Science Information.

6. Avoiding Overfitting in Variable-Order Markov Models: a Cross-Validation Approach

   V Secchini, J Garcia-Bernardo, P Jansky (2025). arXiv preprint.

   Code PDF