Un principio di linearità per le inferenze predittive

Abstract

The determination of probability functions capable of taking into account the existing similarities among possible outcomes is still an open problem among the issues concerning predictive inference. In fact, the solutions which have been proposed up to now, though they are interesting, have a heuristic character. The present paper is a first step towards a deductive analysis of this problem: a family of probability functions is derived on the basis of a general principle of linearity. Clearly, in order to have a complete deductive analysis, we will have to study other linearity principles as well as non linear functions.