来る6月29日（月）、神戸大学人文学研究科にてKevin Kelly教授(Carnegie Mellon University)の講演会を行います。Kelly教授は現代形式認識論の代表的研究者の一人で、「オッカムのカミソリ」の数理的正当化をライフワークにされています。今回はそのプロジェクトの最新の成果をお話頂きます。皆様ふるってご参加下さい。
時間：2015/6/29（月）午後5:00 – 6:30
タイトル：A Reliabilist Justification of Ockham’s Razor
Kevin T. Kelly, Carnegie Mellon University
Konstantin Genin, Carnegie Mellon University
Hanti Lin, University of California, Davis
Scientists prefer simple, unified, sharply tested explanations over complex, diffuse explanations that rely on multiple coincidences—a bias known popularly as Ockham’s razor. One can represent that bias in terms of principles or Bayesian prior probabilities, but it is harder to say how such a bias conduces to true belief. We will present a mathematical argument, based on ideas from topology and learning theory, to the effect that Ockham’s is necessary for staying on the straightest path to the truth, even if that path cannot be perfectly straight. More precisely, choosing a simplest response to a question is necessary to rule out cycles of belief—affirming, then denying, and then re-affirming a conclusion as inquiry proceeds. Waiting for information to decide among simplest answers is necessary to minimize reversals of opinion as inquiry proceeds, even though it is impossible to eliminate reversals entirely, due to the problem of induction. The same ideas can be extended in a natural way to statistical inference, providing a new, frequentist foundation for inductive inference of theories from finite samples. One important application is the statistical inference of causal information from non-experimental data, which heretofore had no frequentist rationale.
The talk is self-contained, and is intended for a general philosophical and scientific audience.
世話人：大塚淳（神戸大学） x@y where x=jotsuka, y=lit.kobe-u.ac.jp