Following a call for applications for junior team leaders in artificial intelligence—which attracted 96 applicants—the Paris Brain Institute has selected two young researchers, Valeria Fascianelli and Jonathan Nicholas, to build their own research teams within its Center for Artificial Intelligence and Data Science. Their work will focus on how the brain organizes information, retrieves memories, and makes decisions. Their teams will join the Institute in February and September 2027, respectively.
Mathematical modelling of brain mechanisms is essential to understanding how the brain functions in both healthy and pathological states. Scientists at the Center for Artificial Intelligence (AI) and Data Science— launched in 2025—are developing computational methods to analyse and interpret data from the Institute’s research teams, with the aim of improving the diagnosis and prognosis of neurological and psychiatric disorders.
By developing computational models that reproduce specific cognitive mechanisms, neuroscientists can now explore processes that were long difficult to observe, such as how neural representations form and how our emotions and memories shape decision-making.
Within this context, the Paris Brain Institute recruited two junior team leaders. Launched in the spring of 2026, this call for applications aimed to attract young scientists from around the world interested in developing an independent research program at the intersection of artificial intelligence, data science, and neurosciences.
Over the 96 applications reveived, two projects were selected through a rigorous, multi-step selection process, involving internationally renowned experts in the fields of neuroscience, artificial intelligence, and data science, who carefully evaluated each candidate's scientific excellence, originality, and potential for impact. Led by Valeria Fascianelli and Jonathan Nicholas, these projects will enhance the scientific expertise of the Center for Artificial Intelligence and Data Science and contribute to a better understanding of the brain mechanisms underlying human cognition.
Valeria Fascianelli
How past experiences shape neural geometry and bias cognitive functions in brain and machines ?
Scheduled to join in February 2027 – [METIS: Modeling the gEometry of inTernal states In brains and artificial Systems]
How does the brain represent the information to learn, decide, and perceive the world? This is the central question Valeria Fascianelli’s research program propose to solve, focusing on neural geometry study.
This concept refers to the way the activity of billions of neurons organizes itself into mathematical structures. These configurations, which reflect our thoughts and cognitive strategies, provide a new framework to understand brain functions, particularly the mechanisms of decision-making. However, several questions remain unanswered: how do these structures form during learning? How do they evolve under the influence of our internal states, such as emotions or motivation?
To answer these questions, Valeria Fascianelli (currently an Associate Research Scientist at Columbia University in New York, USA) will develop an approach at the interface of neuroscience and artificial intelligence. She will develop computational models, including artificial neural networks, as well as methods comparing brain activity across different individuals, brain regions, and species.
Attention will be paid to the influence of emotions on the organization of neural activity. Fear, for example, could distort neural geometry and bias our decisions, attention, or visual perception.
By linking mathematical models of brain activity to observable behaviours, this research should provide a better understanding of how emotional states influence cognitive processes and guide our daily choices.
The team’s name echoes mêtis, the ancient Greek word for adaptive, strategic intelligence.
Jonathan Nicholas
Memories recalled and decision making
Expected arrival in September 2027 – [CMDLab - Computational Memory and Decision Lab]
During the process of decision-making, our brain searches through our memories to evaluate different possibilities. But how does it determine which memories to retrieve, and when? At what point does it decide that it has enough information to make a choice? And how does it know what information to store in a memory in the first place?
Jonathan Nicholas’s research program (currently a postdoc at New York University, New York, USA) aims to unravel these deliberation and storage mechanisms by drawing on a remarkable property of gaze: even when people are looking at a blank screen, their eyes spontaneously return to the locations where information was originally encoded.
These eye movements thus provide a window onto the memories consulted during decision-making, making it possible to observe memory retrieval in real time as people deliberate. By combining eye tracking with brain imaging (MEG and fMRI), his team will be able to see not only which memories are retrieved, but also how the brain uses them to reach a decision.
Jonathan Nicholas links this approach to a computational framework in which memory is treated as a limited and costly resource: the brain should retrieve memories only when they are worth the effort, and it should store information when it is likely to pay off later. Deliberation thus becomes a sequence of micro-decisions in which the brain repeatedly chooses between retrieving another memory or making a decision using the information it already has.
His work has notably shown that an artificial neural network trained to make these micro-decisions optimally, without ever being exposed to human data, reproduces the eye movements and choices of human participants. This suggests that the human brain retrieves memories in a near-optimal way.
Jonathan Nicholas’s future team will build on these results along three research tracks: revealing the causal role of gaze itself in decision-making, understanding how the fidelity of our memories shapes the quality of our choices, and decoding the brain mechanisms underlying this storage and this deliberation processes.
Beyond the fundamental implications for understanding human cognition, this research could open concrete avenues for understanding the decision-making deficits observed in many neurological and psychiatric disorders.
New international recruitments to Support the AI Center’s Scientific Strategy
These two hires are part of the Paris Brain Institute’s Center for Artificial Intelligence and Data Science development strategy. Launched in Janurary 2025, this center brings together expertise in artificial intelligence, data science, neuroscience, and computational disciplines, aiming to strengthen collaboration across these fields.
The recruitment of Valeria Fascianelli and Jonathan Nicholas underscores the Institute’s commitment to support new scientific approaches linking mathematical models, artificial intelligence, and experimental observation of the brain. Their research programs address complementary questions: how do neural representations form and evolve under the influence of emotions? How does the brain draws on memory to make decisions?
By building their own teams at the Institute, these two researchers will help advance research on cognitive functions and strengthen the connections between basic neuroscience and artificial intelligence.
The Paris Brain Institute thanks all members of the interview and selection committees, as well as the administrative staff, for their dedication throughout this demanding recruitment process.
The Institute welcomes Valeria Fascianelli and Jonathan Nicholas and looks forward to the upcoming arrival of their teams, who will contribute to the scientific momentum of the Center for Artificial Intelligence and Data Science.
Opened in July 2025, the Center for Artificial Intelligence (AI) and Data Science of the Paris Brain Institute federates unique multidisciplinary expertise to advance brain understanding and improve brain health. It is housed in a dedicated facility located a 5-minute walk from the main building. It currently gathers 100 people with the objective to scale up to 150.
Click here