Expectations are known to greatly affect our experience of the world. A growing idea in computational neuroscience is that perception and cognition can be successfully described using Bayesian inference models and that the brain is ‘Bayes-optimal’ under some constraints. In this context, expectations are particularly interesting, because they can be viewed as prior beliefs in the statistical inference process. Our aim is to clarify how expectations affect perception and decision-making, how long they take to build up or be unlearned, how complex they can be, and how they can inform us on the type of computations and learning that the brain performs.
I will start by reviewing psychophysical and modeling work from my team, showing how expectations about visual motion direction can be quickly and unconsciously learned through statistical learning, leading to perceptual biases and hallucinations in human observers (Chalk, Seitz & Series, Journal of Vision 2010, Gekas et al, Journal of Vision 2013).
I will also present recent work showing that the prior belief that visual objects are static or move slowly rather than fast (Weiss, Adelson & Simoncelli, Nature Neuroscience, 2002), which is thought to reflect the long-term statistics of natural stimuli and to explain a number of visual illusions such as the “aperture problem”, can be quickly unlearned and inverted (Sotiropoulos, Seitz & Series, Current Biology 2011).
I will finally describe another line of work in decision-making where we look at optimism as a prior belief on future reward. Bio: My initial background was in Computing and Electronic Engineering, then AI and Biomathematics, which I studied in Paris. I did my PhD in Computational Neuroscience, at U. N. I. C., Gif-sur Yvette, France, with Yves Fregnac and Jean Lorenceau. My work focussed on the non-classical properties (contextual modulations) of the receptive fields of primary visual cortex neurons and their perceptual correlates. From 2002 to 2004, I was a postdoctoral fellow in the laboratory of Alex Pouget at the University of Rochester, NY, USA. I then joined Gatsby Computational Neuroscience Unit, UCL, in London, UK. In 2006-07, I spent 9 months in NYC, working at NYU with Eero Simoncelli. Since July 2006, I am a lecturer (equivalent to Assistant Professor in the US) at ANC, School of Informatics, University of Edinburgh, UK. I am interested in questions related to coding in populations of neurons, Bayesian models of perception and decision-making and computational psychiatry.