Place: 4274 Chamberlin (refreshments will be served)
Speaker: Tim Rogers, UW Department of Psychology
Abstract: Theories of human knowledge acquisition (ie learning) vary in many of their particulars but typically embrace the common assumption that learning is rational: through learning, people acquire reasonably accurate statistical models of the environment that allow them, given some new information, to make approximately optimal probabilistic inferences about unobserved states of the world. My own work on knowledge acquisition resides firmly in this tradition, but I have always found it difficult to reconcile this view with the everyday observation that many people appear to pretty firmly believe some pretty crazy things. We can see that this is true even without having to agree what the crazy beliefs are. For instance, the President either was or was not born in Hawaii. These are the only two logical possibilities, and there is a fact of the matter. Of the two groups prepared to vociferously argue each side of the proposition, one must be wrong. The incorrect belief persists in this group despite the fact that we all live to some extent in the same world and are presumably applying largely similar reasoning mechanisms to bear on largely the same evidence. The same point can be made with reference to controversies about global warming, evolution, whether vaccines cause autism, the efficacy of trickle-down economics or gun control policy, the relative payscales of public and private sector workers, and any number of other important issues facing public life. If we are all such optimal learners, why do people arrive at such starkly opposing sets of beliefs?
There is a long tradition of research addressing aspects of this problem. One idea is that human reasoning is "motivated"--there are emotional costs associated with different beliefs, and in deciding which beliefs to endorse, people jointly minimize an error cost (ie, fit of the beliefs to evidence) and the emotional cost associated with the belief. But this approach fails to address the central question of where the emotional cost comes from, or why people should be "motivated" to entertain incorrect beliefs in the first place. A second hypothesis is that the cognitive mechanisms that support human learning and inference were only optimal in an evolutionary context, and are not suited to the modern environment in which we now find ourselves. But such accounts seem similarly underconstrained without some specific accounting of what the learning mechanisms are and how and why specifically they are unsuited to our current environment.