Published online by Cambridge University Press: 01 January 2022
There are two senses in which a hypothesis may be said to unify evidence: (1) ability to increase the mutual information of a set of evidence statements; (2) explanation of commonalities in phenomena by positing a common origin. On Bayesian updating, only Mutual Information Unification contributes to incremental support. Defenders of explanation as a confirmatory virtue that makes independent contribution must appeal to some relevant difference between humans and Bayesian agents. I argue that common origin unification has at best a limited heuristic role in confirmation. Finally, Reichenbachian common cause hypotheses are shown to be instances of Mutual Information Unification.
I thank Michel Janssen, Marc Lange, Bill Harper, and Molly Kao for helpful discussions. I am grateful to Clark Glymour for raising the question, addressed in sec. 7, of how common-cause explanations fit into the framework. This work was supported, in part, by a grant from the Social Sciences and Humanities Research Council of Canada.