AI Systems as Social Scientific Objects

Cognitive science in Vienna and Ljubljana, economics before that; PhD in computational social science at the Complexity Science Hub and the Medical University of Vienna with David Garcia

Sony Computer Science Laboratories Rome: industry postdoc, NLP on news corpora, a data-driven art and science project at MAXXI

Assistant Professor at the University of Mannheim (Data Science in the Economic and Social Sciences), then interim Professor (W2) for Social and Behavioural Data Science at the University of Konstanz

Principal Investigator at the Barcelona Supercomputing Center, leading the COMPASS group; Associate Faculty at CSH Vienna

Joining DNDS in January 2027

Classical computational social science: digital traces of humans, at scale (400M+ forum interactions, social media panels, surveys)

The data that matters for society is increasingly produced by AI systems interacting with humans or with other AI systems

My program: treat large language models as objects of measurement, with the rigor we apply to human subjects

Two complementary modes - Behavioral: present stimuli, record responses (games, surveys, classical experiments) - Representational: probe the internal geometry (embeddings, probing, steering)