Causality, Reliabilism, and Mathematical Knowledge

Abstract

Benacerraf maintains that causal constraints on knowledge, such as those imposed by the causal theory of knowledge, are incompatible with knowledge of statements, such as mathematical statements, whose truth conditions involve abstract entities. Reliabilism appears more hospitable to mathematical knowledge, for it does not overtly endorse such causal constraints on knowledge. This paper argues that the appearances are deceiving. First, I distinguish two forms of reliabilism: reliable indicator theories and process reliabilism. Second, I argue that the former theories cannot accommodate a priori knowledge since they involve causal constraints similar to those required by the causal theory. Third, I maintain that although some versions of process reliabilism are compatible with mathematical knowledge, they are too weak to provide plausible constraints on knowledge. I propose and defend a stronger version of process reliabilism and argue that on this version causal considerations become relevant to knowledge in a surprising new way.