Credences and Degrees of Belief

A credence is a number assigned to a belief to say how confident you are in it — for example, being 80% sure an idea is true. It is one of a family of degree-based measures (probability, plausibility, weight, strength, support) that purport to rank how good an idea is on a spectrum. Bayesian epistemology treats credences as central: rational belief just is a probability that updates with evidence.

CF rejects credences as the wrong tool for epistemology. The objection is narrow and deliberate. Probabilities are entirely appropriate for dice rolls, random sampling, measurement error, and other genuinely statistical situations. The mistake is transferring probability of physical events onto the truth of ideas. Evaluating, debating, and choosing between ideas should proceed by criticisms — explanations of error — and responses, a kind of reasoning that does not correspond to any quantity.

Temple gives two specific arguments. First, a credence below 100% is usually a way of not dealing with a criticism: saying “90% confident” parks the problem the criticism raises instead of refuting it or fixing the idea. Second, applying certainties to all ideas generates an infinite regress, because each certainty judgment is itself an idea needing its own certainty.

CF’s alternative is binary: an idea is either refuted by a decisive criticism or it is not. You act on a non-refuted idea — the best available — while staying open to new criticism. Confidence differences between accepted ideas are handled by context and goals, not by attaching numbers. See also gradations of certainty and Yes or No Philosophy.