A subjective interpretation of conditional probability (Q794045)

The aim of this paper is to give, within the framework of (subjective) probability changes over time, a new foundation or justification for the rule of conditioning, i.e. the standard reduction of conditional probability to (a quotient of) unconditional probabilities. The basic concept involved is that of an ''information partition'' which, besides forming a disjoint partition of the set, \(\Omega\), of possible worlds (or ''states of nature''), is meant to ''specify that the person's knowledge at time t will amount to knowledge of which element of the partition contains the true state of nature'' (p. 453). The mathematical results of section 3 and 4 seem to be right, but some of the underlying philosophical assumptions must be judged rather dubious: i) the notion of ''knowledge'' is, somewhat misleadingly, used as synonymous with ''full belief''; ii) the author leaves unclear whether the ''states of nature'' which are designed to provide a model for the epistemic state of the agent G are to be allowed to contain themselves information about G's epistemic state; iii) consequently it remains doubtful whether an unproblematic sense of ''specification'' may be given to the author's stipulation that ''each element \(\omega\) of \(\Omega\) specifies that Gracchus's knowledge at time t will be the element of [the partition] G that contains \(\omega\) '' (p. 455): the statement of proposition 1 suggests that ''specification'' merely amounts to the existence of a function assigning to each \(\omega\in \Omega\) a subset \(G_ t(\omega)\) of \(\Omega\) such that (1) \(\omega \in G_ t(\omega)\) and (2) \(''G_ t(\omega)\) will be Gracchus's knowledge at time t''; in the subsequent proof, however, the author seems to assume that condition (2) is not only true but also contained in (or derivable from) the description of the ''state of nature'', i.e. from \(\omega\) itself.