University of Groningen

Pau is a PhD candidate at the Interuniversity Center for Social Science Theory and Methodology (ICS), based at the University of Groningen.

His project focuses on a type of deep learning models known as Graph Neural Networks (GNNs) and their potential applications in social network analysis. The main goal is to evaluate the extent to which GNNs can help us predict outcomes in social networks, as well as to determine whether these predictions can provide causal insights. To this end, he will combine GNNs with novel, state-of-the-art explainability methods to better understand what the models are learning—that is, the factors enabling their predictions.

Some examples of how GNNs could be applied to social network data include predicting the adoption of behaviors based on the influence of neighbors, estimating the likelihood that two individuals are connected, and identifying those at higher risk of occupying marginal positions.

He will apply these models to data collected by WP2 of GUTS, as well as to other datasets such as KiVa NL—a large longitudinal dataset containing network information for over ten thousand students across five hundred classrooms.