DONDENA Seminar - Alberto Aletá Casas

You may follow the seminar at the following link.

“Can Generative Epidemiology be a Thing?” 

SPEAKER: Alberto Aletá Casas (Universidad de Zaragoza).

ABSTRACT:

Can Generative Epidemiology be a thing? Human behavior, data-driven models, and the promise (and limits) of LLMs.

During the COVID-19 pandemic, the use of large-scale human behavior data became a cornerstone of the mathematical models developed for managing the public health crisis. In this talk, I will first review how we used data-driven models to observe the footprints of human behavior left in many data sources. I will also discuss the complexity of coupling epidemic and economic models, highlighting the challenges of calibration and the implicit assumptions we make when decision-making relies on incomplete data. Despite all this progress, we are still far from understanding how to truly integrate human behavior into epidemic models. Furthermore, this challenge is becoming even harder. The landscape of human interaction is becoming even more complex as it is shifting rapidly with the widespread adoption of Large Language Models (LLMs). As these tools increasingly influence how we seek health information, write, code, and even speak, they force us to reconsider our modeling approaches. I will provide a technical but accessible overview of LLMs, demystifying concepts like transformers and the crucial process of human alignment. Finally, I will introduce the concept of "Generative Epidemiology". Can we use LLMs as surrogates to model complex elements of human behavior that traditional data cannot capture? Or, should we consider them as distinct agents that will interact with humans under new rules, with consequences for future crises? I will argue that regardless of whether they may serve as accurate proxies or not, understanding the interplay between algorithmic agents and human society may become essential for future pandemic preparedness.
 

BIO: 

Dr. Alberto Aleta is a Spanish physicist and computer engineer specialized in the field of Complex Systems. He is mostly known for his work on epidemic modelling, especially during the COVID-19 pandemic. However, his interests are very interdisciplinary, with contributions in fields like game theory, sustainable nutrition, and social dynamics. Alberto's work is characterised by the use of data-driven approaches to model dynamical systems and a quest for unraveling the causal mechanisms behind the observed dynamics. His works often rely on the use of agent-based models, network science, higher-order networks, and multivariate statistics. Alberto obtained his PhD in Physics from the University of Zaragoza (Spain) in 2019 and then joined the ISI Foundation (Turin, Italy) as a postdoctoral researcher until 2022. He returned to Zaragoza in 2023 with a Ramón y Cajal fellowship, rejoining the Institute for Biocomputation and Physics of Complex Systems (BIFI), where he is currently based.