Professor Barton L. Anderson will be visiting from The University of Sydney to present a talk questioning the increasingly popular 'Bayesian' approach to studying perception. This should have relevance to a wide range of contemporary approaches to different problems in Psychology.
How Bayesian is our experience of the world? Insights gleaned from material and surface perception
A core problem of mid-level vision involves understanding how the visual system disentangles different physical contributions to image structure. The image contains a conflated mixture of 3D scene geometry, surface properties (lightness, color, gloss, translucency), and the illumination field, which our visual systems somehow parse into distinct dimensions of experience. The problem of decomposing images into their underlying physical sources is universally regarded as an ill-posed inference problem that can only be solved approximately with the aid of the tools of probability theory (Bayes theorem). One common assumption of Bayesian vision models is that natural selection shaped our visual systems to generate veridical scene representations, based on the belief that that veridicality is synonymous with evolutionary fitness. I will discuss logical flaws and problems with this conceptualization, describe a number of different research projects we have conducted into material that reveal the shortcomings of this approach, and offer an alternative conception of how to study and model perceptual experience that avoids the pitfalls inherent in this approach.
Barton Anderson is a vision scientist currently residing in Sydney Australia who studies a variety of problems in vision. The main focus of his research is on "mid-level" vision, particularly issues of perceptual organization, segmentation, grouping, and the recovery of surface attributes (such as their color, material composition and shape).
Professor Anderson is widely recognised as a world leader in his field. He completed his PhD at Vanderbilt University in the United States, before taking up faculty positions at MIT and The University of Sydney. He has published extensively in the world's premier scientific outlets, including Nature, Science, Nature Neuroscience, Current Biology and Psychological Review.