Hi, I'm Alessandra Brondetta!

I am a Ph.D. student at the Institute of Cognitive Science, Osnabrück University, in Prof. Sebastian Musslick's research group, where I work at the intersection of cognitive science, artificial intelligence and complex systems.
Before the Ph.D., I completed a M.Sc. in Physics of Complex Systems and a B.Sc. in Physics at the University of Turin, and carried out my master’s thesis research at Brown University and at CENTAI Institute in Turin.
I explore cognitive processes through AI and theoretical modeling. By connecting theory and computation, I am interested in better understanding adaptive cognition at multiple levels and accelerating large-scale experiments with behavioral simulators.

News

• 11/2025 Poster presentation at AI in Science Summit 2025 in Copenhagen (DK).
• 09/2025 Attendee at Automated Scientific Discovery of Mind & Brain (2‑week workshop), Princeton University.
• 09/2025 Poster presentation at COCOFLEX seminar in Ghent (BE).
• 08/2025 Published on arXiv our commentary "Not Yet AlphaFold for the Mind".
• 03/2025 🎉 Officially started as a PhD student with Prof. Sebastian Musslick.
• 02/2025 Released Mentevo, a compact Python library for multi-agent task-switching modeling.
• 11/2024 Research talk (online) for NTU Singapore's Dynamical Systems group.
• 09/2024 Joined Institute of Cognitive Science (Osnabrück University) as Research Associate.
• 08/2024 Heterogeneity paper published in the COGSCI 2024 proceedings, poster presented in Rotterdam (NL).
• 07/2024 🎉 Completed M.Sc. in Physics of Complex Systems, Università di Torino.
• 07/2023 Research presentation at Princeton University (Prof. Jonathan D. Cohen's group).
• 04/2023 Visiting Research Fellow at Carney Institute, Brown University (through 07/2023).
• 01/2023 Research Trainee with Prof. Giovanni Petri NPLab at CENTAI Institute, Torino (continued into 2024).

Publications

• 

  Automated Prototyping of Behavioral Experiments with Large Language Models

  Alessandra Brondetta, Sebastian Musslick

  Paper AISS2025

  Piloting behavioral experiments is slow and costly. We introduce an LLM-driven framework to prototype experiments in silico before running them with humans. An “experimentalist” LLM iteratively proposes designs while “participant” LLMs engage with them, optimizing for desired behavioral outcomes. In the Wisconsin Card Sorting Test, the method discovers task framings that systematically shift behavior along the stability–flexibility axis. This can speed up study design and reduce pilot costs before human testing.

• 

  Not Yet AlphaFold for the Mind: Evaluating Centaur as a Synthetic Participant

  Sabrina Namazova, Alessandra Brondetta, Younes Strittmatter, Matthew Nassar, Sebastian Musslick

  Paper arXiv

  Simulators speed science by generating realistic data—think AlphaFold for protein structures. Behavioral science seeks a comparable participant simulator that produces human-like behavior across tasks. We review Centaur (an LLM fine-tuned on 160 psychological experiments) against core criteria: it predicts well, but its generative behavior diverges from human data. Takeaway: a strong direction with clear potential, but more work is needed before it can serve as a reliable participant simulator or comprehensive model of cognition.

• 

  Mentevo: A Python Library for Cognitive Stability and Flexibility Simulation

  Alessandra Brondetta

  Github

  Mentevo is a compact Python library for simulating and analyzing a dynamical system model of cognitive stability and flexibility in task-switching environments, within groups of agents. Originally developed for the study presented in Brondetta et al. (2024), it provides tools to explore how environmental and task conditions (e.g., task-switching rates) affect individual and group task-switching performance. The library also includes parametrization options, performance metrics and visualization methods.

• 

  On the Benefits of Heterogeneity in Cognitive Stability and Flexibility for Collaborative Task Switching

  Alessandra Brondetta, Anastasia Bizyaeva, Maxime Lucas, Giovanni Petri, Sebastian Musslick

  Paper CogSci 2024

  We model how the trade-off between cognitive stability and flexibility shapes group task-switching. In our multi-agent simulations, heterogeneous teams (mix of more stable and more flexible subjects) outperform homogeneous ones in fast-changing environments—especially when the most flexible members handle switch cues. The findings suggest hypotheses about how to compose teams and design supports in educational and clinical settings.