+
Synthetic populations are individual-level reconstructions of real populations, built from official aggregates, registers and surveys to reproduce selected statistical characteristics. Geographically detailed representations of individuals and households provide a foundation for studying population heterogeneity, spatial inequalities and collective behaviour, connecting demographic and socioeconomic data to computational models.
+
This seminar series explores how synthetic populations are constructed, validated and used. Methods range from statistical reconstruction and maximum-entropy modelling to deep generative approaches. Generative AI opens further possibilities both for population synthesis—integrating heterogeneous sources and modelling complex dependencies—and for simulation, combining synthetic populations with agent-based models and agents powered by large language models.
+
These developments extend across urban systems and mobility, public health and medicine, economics and finance, communication and other areas of social research. They also raise scientific and computational challenges: which features of real populations can be recovered from available data? How should uncertainty, bias and model validity be assessed? How can individual attributes and simulated behaviours remain consistent across spatial scales? And how can increasingly large and complex populations be generated, simulated and analysed efficiently using high-performance computing?
+
Bringing together perspectives from physics, statistics, computer science and the social and health sciences, the series aims to connect methodological advances with applications, with particular attention to validation, reproducibility and computational scalability.
+
Population synthesis · Maximum entropy · Reproducibility · Agent-based modelling · Generative AI
+