A Problem for Credal Consequentialism

Beliefs, we’ll assume, come in degrees. As a short-hand, we’ll refer to these graded doxastic attitudes as credences, and to the totality of an agent’s credences as their credal state. Plausibly, there are certain epistemically good-making features that an agent’s credal state may have. Credences, for example, may be closer or further from the truth, and it is at least prima facie plausible that the closer an agent’s credences are, on average, to the truth the better such credences are epistemically. Given an appropriately well-defined notion of the epistemic good for credal states, it is tempting to think that the facts concerning which credal states are rationally permissible or obligatory for an agent may be grounded in facts about the extent to which such credal states appear, by the lights of the agent, to be conducive to the attainment of the epistemic good. Call this sort of view credal consequentialism. For example, it is tempting to think that an agent ought rationally to adopt those credences that, by her lights, would be most conducive to the attainment of the epistemic good. Credal rationality, on this picture, consists in maximizing the expected epistemic utility of one’s credal state. Like consequentialist accounts in ethics, credal consequentialism has the virtue of being simple and principled. However, like consequentialist accounts in ethics, credal consequentialism has some surprising consequences that are at odds with our pre-theoretic judgments. First, given plausible accounts of the epistemic good for credal states, this account councils the acceptance of so-called epistemic bribes; in certain cases, by adopting apparently irrational credences in particular propositions an agent may ensure that their other credences have certain epistemic good-making features, and so, in certain cases, an agent may maximize their overall epistemic utility by adopting credences in particular propositions that seem irrational. Second, given plausible accounts of the epistemic good for credal states, ∗Forthcoming in Epistemic Consequentialism, Jeffrey Dunn and Kristoffer Ahlstrom-Vij (eds.), OUP. Thanks to the editors and anonymous referees of this volume. Special thanks to Jason Konek and Cian Dorr for very helpful comments on a draft of this paper. See Joyce (1998) and Joyce (2009) for an assessment of different ways of developing the idea that the principal good-making feature for credal states consists in their closeness to the truth. For alternative proposals according to which the principal good-making feature of credal states consists in their closeness to chances see Hájek (n.d.), Pettigrew (2012), and Caie (2015). See, however, Konek and Levinstein (forthcoming) for an account that attempts to ground facts about credal rationality in facts about the epistemic good for credal states in a distinctively non-consequentialist manner. For arguments to this effect see, for example, Jenkins (2007), Greaves (2013), and Berker (2013).