Brendan O'Connor, M.S. candidate, Symbolic Systems Program, "Does belief revision go wrong? Bayesian rationality and confirmation bias"
ABSTRACT:
When I look at news editorials, I like to read what I already agree with. When I do experiments I select samples that favor my hypothesis. I find support for my beliefs everywhere. Psychologists might say I am exhibiting confirmation or evidence assimilation bias -- a tendency to seek out evidence in favor of one's currently held beliefs, or overinterpret evidence as favoring them. These biases have large implications for human rationality.
This totally unbiased talk will review several classic experiments, and discuss cognitive and affective explanations for why confirmation bias occurs. Some relevant connectionist, cognitive dissonance, and behavioral economics models will be presented, along with an introduction to Bayesian probability theory -- a model for rational belief revision -- which, it will be argued, can help define and empirically test how confirmation bias works.