Using Bayes' Rule to Shave with Einstein's Razor

Saturday, February 16, 2013
Room 306 (Hynes Convention Center)
Randy Gallistel , Rutgers University, Piscataway, NJ
Using Bayes rule to decide between competing encodings of one's experience implements Einstein's razor; it picks out the description that is as simple as possible but no simpler. This principle illuminates psychological phenomena ranging from the figure-ground phenomenon in perception to the extinction phenomenon is associative learning. What this principle does not do is generate the competing descriptions. Genetically specified domain-specific principles do much of that work. These principles are often analytic truths, such as that position is the integral of velocity. The justification for assuming that they have been incorporated into the Bayesian inference mechanism does not rest on speculation about conditions that prevailed eons ago.