Abstract

How should we understand the notion of sufficient evidence? How should we deal with the worry that any attempt to use a threshold of support to characterise this notion would be arbitrary? In this paper, I argue that the most familiar way of understanding a threshold is unworkable but also argue that it would be misguided to do away with a threshold-centred account. I argue that we should approach sufficiency in the theoretical realm much in the way that we would in the practical realm where we could, if we wished, say that the evidence for choosing some option is sufficient when given this information this option is no less choiceworthy than its alternatives. Applied to belief, the idea would be that the evidence is sufficient when the expected epistemic desirability of belief is not less than that of suspension. Whilst we might have to acknowledge that the precise weights of reasons that bear on whether to believe are unknowable (making the threshold unknowable), the mechanism that determines a threshold is known and clearly non-arbitrary. The paper concludes with some reflections on why this comparativist approach has been overlooked by epistemologists and reflections on the normative significance of naked statistical evidence. Contrary to an increasingly popular view, I think that naked statistical evidence can both confer justification upon a belief and defeat the justification a belief has, but only if it's evidence _of_ the properties that make belief desirable or undesirable. Whilst naked statistical evidence of truth will not make it rational to believe, this is not because naked statistical evidence is normatively irrelevant but because it is evidence that indicates that only one aspect of desirable belief might be present. Any evidence that warrants a high degree of confidence that a belief is knowledge will make it rational to believe and this includes naked statistical evidence of the properties that turn a belief into knowledge. I also argue that it is an objection to nearly every extant theory of justification that they recognise the factors that determine the threshold I discuss in this paper since these theories invariably treat the threshold as normatively irrelevant. Bad on them. The real threshold problem is not (contrary to prevailing opinion) the problem of explaining how it might be non-arbitrary (that's easily solved), but rather the problem of explaining why a rational thinker should prefer belief to suspension when belief does not cross the threshold or suspension to belief when belief does cross the threshold. This is not a problem for my preferred theory of rational belief or the Lockean view, but it is a problem for virtually every other theory of rational belief on the market.