Computational characterization of perceptual decision-making as contextual inference

The jury will be composed of:

• Floris de Lange (rapporteur, Radboud U Nijmegen)
• Mathias Pessiglione (rapporteur, ICM, Paris)
• Rufin VanRullen (rapporteur, CerCo, Toulouse)
• Claire Sergent (examinatrice, U Paris Descartes)
• Vincent de Gardelle (examinateur, CES, Paris)
• Catherine Tallon-Baudry (examinatrice, ENS, Paris)
• Etienne Koechlin (examinateur, ENS, Paris)
• Nicolas Baumard (tuteur HDR, ENS, Paris)

Abstract:

A central question that has motivated research in perception arises from a simple observation: we do not perceive the world around us as it is, nor even as others perceive it. What, then, is the origin of this 'subjective' character of human perception? By seeking to answer this question in terms of cognitive computations and neural mechanisms, my research has shown how human perception can be described as the product of a noisy inference process that is subject to multiple contextual influences and neurobiological constraints. I will first give a selective overview of my previous work, focusing on the main findings which have shaped my current view on the issue. I will then present in more detail ongoing work in my group which uses this computational framework for bridging research in perceptual and reward-guided decision-making, and for understanding psychiatric conditions associated with dysfunctions of inference.