Open logic based on total-ordered partition model (Q1286656)

One of the main problems in formalizing belief revision is that, traditionally, researchers in Artificial Intelligence (AI) have been trying to describe belief revisions based on the assumption that at any given moment of time, for any given statement \(S\), an agent either believes in \(S\), or does not believe in \(S\). When we have only two beliefs \(S_1\) and \(S_2\), and it turns out that \(S_1\&S_2\) cannot be true, then we must select between keeping \(S_1\) and keeping \(S_2\), but the belief description is absolutely symmetric with respect to \(S_1\leftrightarrow S_2\), so no such choice is possible. In real life, we often do choose \(S_1\) or \(S_2\) depending on in which of these two statements we believe more, so we have \textit{degrees} of belief. The authors formalize this idea of degrees of belief, and show that we can describe a belief revision procedure for the beliefs themselves \textit{and} for the corresponding degrees in such a way that, no matter what beliefs we start with, we can guarantee that after presenting all correct facts (in an arbitrary order), the revised set of beliefs converges to the set of true statements.