Unlike most of the neural representations of cognitive variables currently being studied, the representation of uncertainty is directly linked to the observer's beliefs about the world around him, which poses specific methodological challenges. Two scientific communities are working on this neural representation of uncertainty. Their very different approaches may seem irreconcilable. But what is the reality? Researchers from both communities, including Florent Meyniel from Unicog (NeuroSpin), sat down together to see if it was possible to bring them closer together, or even to reconcile them.
The fruit of this collective reflection has been published in Nature Neuroscience. In it, the authors identify the methodological foundations of the two approaches. One approach assumes the level of uncertainty represented in the brain at a given moment, using appropriate models and experiments, and looks for patterns of brain activity that vary according to this uncertainty. The other approach uses models from theoretical neuroscience that presuppose a certain neural code for the representation of cognitive variables and the associated uncertainty, and compares these predictions with the data.
The article shows that these approaches are highly complementary and would benefit from being used synergistically to understand how the brain represents uncertainty, probably in different ways at different stages of information processing.
Frédéric-Joliot Institute Researcher contact: