Causal basis for probabilistic belief change: distance vs. closeness

In probabilistic accounts of belief change, traditionally Bayesian conditioning is employed when the received information is consistent with the current knowledge, and imaging is used otherwise. It is well recognised that imaging can be used even if the received information is consistent with the current knowledge. Imaging assumes, inter alia, a relational measure of similarity among worlds. In a recent work, Rens and Meyer have argued that when, in light of new evidence, we no longer consider a world ω to be a serious possibility, worlds more similar to it should be considered relatively less plausible, and hence more dissimilar (distant) a world is from ω, the larger should be its share in the original probability mass of ω. In this paper we argue that this approach leads to results that revolt against our causal intuition, and propose a converse account where a larger share of ω’s mass move to worlds that are more similar (closer) to it instead.