The project, titled “AI-Driven Behavioural Epidemic Models for Intervention Design (Aibem),” will receive 970,139 euros over three years and will be coordinated by Professor Yamir Moreno, director of the University Institute for Biocomputation and Complex Systems Physics (BIFI) and a member of the Department of Theoretical Physics at the University of Zaragoza. The consortium also includes the research team of Joan Benach, a professor at Pompeu Fabra University (UPF), director of the Grup de Recerca en Desigualtats en Salut – Employment Conditions Network (GREDS-EMCONET) and co-director of the JHU-UPF Public Policy Center; and the team of Professor Anxo Sánchez, a researcher with the Interdisciplinary Group on Complex Systems (GISC) at Carlos III University of Madrid (UC3M).
Aibem was one of 32 projects selected from the 800 proposals submitted to the 2026 call, representing a success rate of 4%. In total, Fundación “la Caixa” will allocate 25.7 million euros to the projects selected in this edition.
The award also marks a milestone for research in Aragon. Since its first edition in 2018, Fundación “la Caixa”’s Health Research call has funded around 260 biomedical research projects in Spain and Portugal. AIBEM is the first Aragonese project led by the University of Zaragoza as the principal organization (host organization).
“For two decades, science has enabled us to accurately model the dynamics of pathogen spread through contact networks. However, the real challenge has always been the unpredictability of our own nature. Capturing the complexity of human behavior during a health crisis remains an unresolved challenge. The emergence of generative artificial intelligence changes the rules of the game in our favor: by giving autonomous computational agents the ability to reason and influence one another, we could make epidemiological models more than mere mathematical projections and turn them into a faithful reflection of the human factor at the social level,” says the project’s principal investigator, Yamir Moreno, director of the University Institute for Biocomputation and Complex Systems Physics (BIFI) and a member of the Department of Theoretical Physics at the University of Zaragoza.
MODELS DESIGNED TO EMULATE HUMAN BEHAVIOR
The Covid pandemic showed that compliance with measures such as vaccination, mask wearing and reduced mobility varied enormously between social groups and over time. It was also influenced by risk perception, trust in institutions, disinformation, socioeconomic circumstances, vulnerability and material inequalities. Nevertheless, conventional epidemiological models continue to treat human behavior as a fixed or exogenous parameter, considerably limiting their usefulness as decision-making tools.
In this regard, Aibem proposes a paradigm shift: so-called generative agent-based models (GABMs), in which each simulated individual is powered by a large language model (LLM) configured with a specific profile. This is the same technology that underpins artificial intelligence platforms such as ChatGPT and Gemini. These agents will have their own identity, goals, memory and tools. In this context, they will interact with one another, make decisions, form expectations and generate social norms. Their responses will also be systematically validated against experiments involving real people.
“Epidemics are not merely biological phenomena: their spread and consequences are deeply shaped by social inequalities, living and working conditions and power relations,” says another project researcher, Joan Benach, director of the Grup de Recerca en Desigualtats en Salut – Employment Conditions Network (GREDS-EMCONET) and co-director of the JHU-UPF Public Policy Center at Pompeu Fabra University (UPF).
In this regard, Benach believes that developing more sophisticated epidemiological models is not enough. It is also necessary to incorporate the structural causes that make certain social groups more exposed to epidemics, more vulnerable or less able to protect themselves. At UPF, researchers will contribute their expertise in health inequalities, ecosocial determinants and public policies, leading the design and evaluation of interventions that take into account both their effectiveness and their effects on equity.
“Our challenge is to help ensure that artificial intelligence is not limited to predicting epidemics more accurately, but also helps explain them and develop socially just public health responses,” Benach stresses. These responses must involve the relevant stakeholders and help reduce existing inequalities rather than reproduce them.
The project is organized around three main areas of work. UC3M will conduct behavioral experiments with socioeconomically diverse participants through the IBSEN platform, implemented as part of the European H2020 project of the same name. The University of Zaragoza, for its part, will develop generative agents that reproduce human behavior and integrate them into epidemiological models, drawing on BIFI’s supercomputing infrastructure (CESAR) and the extensive experience of the Complex Networks and Systems Research Group (COSNET) in the mathematical and computational modeling of epidemics. UPF will translate the results into decision-support tools for health authorities, including an explicit assessment of the distributive impact of each intervention.
EQUITY AS A MODEL OUTPUT
One of AIBEM’s defining features is that equity is not added at the end of the analysis, but treated as a primary output of the model itself. Age-based contact matrices will be expanded to include socioeconomic dimensions — occupation, income and neighborhood — using data from the Spanish census, labor force surveys and stratified epidemiological records, so that differences in disease burden and response capacity emerge from the material conditions of each group.
The project also incorporates an ethical and governance framework aligned with the European Union Artificial Intelligence Act (EU AI Act), including bias audits, human oversight and independent equity reviews. All models, data and validation protocols will be published openly in accordance with the FAIR principles.
Expected results include behavioral agents validated against human experimental data, with explicit socioeconomic diversity; coupled epidemiological models capable of quantifying how behavioral heterogeneity influences transmission; a systematic methodology for designing and evaluating interventions under realistic social conditions; and a decision-support toolkit aimed at Spanish and European health authorities, including policy reports, ethical guidelines and simulation software.
