|290-3||Graduate Seminar: Belief, Credence, and Reasons||MacFarlane||W 2-4||TBA|
We can describe an agent’s doxastic state by saying which propositions the agent believes. Alternatively, we can assign credences or confidence levels between 0 and 1 to each proposition. The first approach is “digital,” positing a binary, on/off relation of belief, while the second is “analog,” recognizing continuous variation in the strength of belief. What is the relation between these two ways of describing a doxastic state? Is one of the two modes of description accurate, and the other inaccurate? Or are they just different ways of describing the same underlying reality? Is one of the two modes of description fundamental, and the other derivative? Why do we have these two modes of description? Do they serve different purposes?
The divide between digital and analog descriptions of doxastic states corresponds to a divide in theories of reasoning. Traditional accounts of reasoning give a central role to the notion of a reason. We decide what to believe, and what do do, by balancing up the reasons for and against and evaluating their weight. Much traditional epistemology is shaped by this conception of reasoning. But this notion does not play a role in Bayesian epistemology or in decision theory, which focuses more on a holistic notion of rationality. We will consider arguments that the constraints of rationality make sense only for credences, not for beliefs. We will also consider whether the notion of a reason makes sense only in the context of the full-belief ontology. Can we think of actions rationalized by states of partial belief as done for reasons? If not, is that a problem for the idea that reasons are central to reasoning? Do we care about reasons because we want to make better decisions, or for some other reason?
We will read and discuss some of the recent literature bearing on these topics.