Synthetic Populations

Synthetic populations are emerging as a foundational paradigm in Artificial Intelligence, addressing a growing need to model, simulate, and reason about large collections of heterogeneous individuals while preserving privacy, realism, and structural constraints. Rather than focusing on isolated agents or aggregate statistics alone, synthetic population approaches aim to generate entire populations of artificial individuals whose joint distributions, interactions, and dynamics reflect real-world systems.

Historically rooted in fields such as microsimulation, demography, and transportation modeling, synthetic populations are now undergoing a methodological and conceptual expansion driven by advances in AI. Modern applications span public policy evaluation, epidemiology, urban planning, economics, energy systems, marketing, cultural analysis, and social simulation, where access to real individual-level data is limited, sensitive, biased, or legally constrained.

From an AI perspective, synthetic populations raise a distinctive set of challenges that cut across multiple subfields. At the core lie high-dimensional combinatorial generation problems under complex global constraints, often involving consistency requirements, marginal distributions, relational structures, and fairness criteria. These problems naturally connect to combinatorial optimization, constraint programming, probabilistic graphical models, methods, from statistical physics and hybrid symbolic–statistical approaches.

Beyond static generation, synthetic populations increasingly serve as substrates for multi-agent simulation and complex dynamic systems. Individuals interact, adapt, learn, and co-evolve, giving rise to emergent collective phenomena. This brings synthetic populations into close dialogue with multi-agent systems, agent-based modeling, reinforcement learning, game theory, social choice, and network science. Questions of validation, robustness, explainability, and sensitivity analysis become central when synthetic populations are used for decision-making or policy support.

Recent developments in generative AI further enrich the field. Large language models and other foundation models enable the creation of synthetic individuals with rich behavioral, cognitive, or linguistic profiles, opening new possibilities but also raising fundamental questions about realism, bias amplification, controllability, and epistemic validity. How to combine data-driven generative models with explicit constraints, domain knowledge, and ethical guarantees remains an open research frontier.

This Research Topic aims to provide a comprehensive and interdisciplinary overview of synthetic populations as an emerging AI domain. We welcome contributions ranging from theoretical foundations and algorithmic advances to applied systems, evaluation methodologies, and critical perspectives. Topics of interest include (but are not limited to):

● Synthetic population generation under constraints
● Combinatorial and probabilistic methods for population synthesis
● Multi-agent simulations on synthetic populations
● Integration of generative AI and symbolic methods
● Validation, fairness, privacy, and interpretability
● Applications in social, economic, cultural, and policy contexts

By bringing together researchers from AI, computer science, social sciences, and applied domains, this Research Topic seeks to clarify the conceptual landscape, highlight open challenges, and establish synthetic populations as a first-class object of study in Artificial Intelligence.


Topic Editor François Pachet has shares in the company "Imagine All The People". The other Topic Editors declare no competing interests with regard to the Research Topic subject.