Europe supports talented early-career researchers who have demonstrated their potential to become independent research leaders. To this end, the European Research Council (ERC) opens each year the "Starting Grant" call to fund researchers with 2-7 years of experience since completion of their PhD, up to €1.5 million over five years. This year, 436 researchers have been awarded, including 38 working in France. Florent Meyniel, who works in the UNICOG laboratory directed by Stanislas Dehaene (NeuroSpin), is one of the lucky recipients.
The researcher is interested in how the human brain estimates and uses uncertainty. Several observations have led him to consider that our brain estimates the probability of the cause of an event based on the observation of already known events, using a method known as Bayesian inference. The project should make it possible to establish the mechanisms (neural codes) at the origin of our ability to correctly estimate information as being more or less certain. On the accuracy of this ability would directly depend our ability to adapt to change in an uncertain world. People who adapt easily would be people capable of correctly estimating and updating probabilities (i.e. correctly applying Bayesian inference). Their confidence level would be fair, high when the information they learn is true, low when it is false. Conversely, people who have difficulties adapting to change, such as people with certain psychiatric disorders, would misestimate the probabilities (i.e., they would not correctly apply Bayesian inference). Part of the project will aim to understand the factors that control our adaptability, particularly in learning situations, and to study whether our estimation of probabilities can be trained to be more accurate. If this is the case, applications in everyday life will be of primary importance to improve learning and decision making. To characterize the Bayesian nature of the probabilities estimated by our brain, the researcher will combine experimental psychology tests, computer models and neuroimaging.