Tokyo Forum for Analytic Philosophy
日時：2018年10月3日（水）18:00-20:00
場所：東京大学駒場キャンパス18号館4階コラボレーションルーム2

Trusting the Predictions of a Hypothesis vs Believing that the Hypothesis is True

Speaker: Olav Benjamin Vassend
From: Nanyang Technological University, Singapore
URL: https://sites.google.com/site/olavbvassend/home

Abstract: It is often reasonable to trust the predictions of a hypothesis that is known to be false. Indeed, as Forster and Sober (1994) emphasize, in the context of statistical model selection it is sometimes rational to trust the predictions of a known false model over the predictions of a known true model. Hence, the goal of identifying true hypotheses is distinct from the goal of identifying hypotheses that can be expected to be predictively accurate. The aim of my presentation is to show that the pursuit of truth and the pursuit of predictive accuracy in fact are governed by different rational norms. In particular, although degrees of belief ought to be updated through Bayesian conditionalization, degrees of trust ought instead to be updated through a different method that I call "Brier updating" (after (Brier 1950)). I argue that Brier updating has strong theoretical support and I show that it results in more accurate predictions than Bayesian conditionalization in simple simulation experiments. I then explore the philosophical implications of the arguments, including the upshot for scientific realism.

より詳細な情報については、TFAPのウェブサイトをご覧ください。