Scoring Rules and the Inevitability of Probability 论文

摘要

Abstract : Let a person express his uncertainty about an event E, conditional upon an event F, by a number x and let him be given, as a result, a score which depends on x and the truth or falsity of E when F is true. It is shown that if the scores are additive for different events and if the person chooses admissible values only, then there exists a known transform of the values x to values which are probabilities. In particular, it follows that values x derived by significance tests, confidence intervals or by the rules of fuzzy logic are inadmissible. Only probability is a sensible description of uncertainty.