Standards of Rationality and Epistemic Sophistication
Franz-Peter Griesmaier

It is widely assumed that standards of rationality, or the norms of reason, are universal principles. For example, the heated debate about whether or not humans are largely irrational, which followed on the heels of experimental results about reasoning by Wason, Tversky, Kahneman, and others, can only be understood if the assumption of the universality of rationality is made. For some canons of rationality, this assumption seems plausible enough. Norms that are derived from logic, probability, and decision theory may well be universal in that they apply to every actual and possible reasoner. At least, I won’t question the universality of those norms.

However, when assessing the rationality of an epistemic agent, those norms are usually not all that are involved. There are further norms adherence to which is part of the epistemic responsibility of agents. For example, in defeasible reasoning, there are norms guiding acceptance or rejection of belief in light of the availability or absence of defeaters which are not readily expressed in probabilistic terms. One such norm might be expressed by the rule ‘innocent until proven guilty’. A specific instance of this rule can be found in the doctrine of epistemic conservatism, according to which a belief is prima facie justified in virtue of the mere fact that an agent has it, as long as no defeaters are available to the agent.

In my talk, I want to raise doubts about the universality of a norm like this one. In particular, I will argue (i) that the norm is ambiguous between two readings of prima facie justification, one of which focusses on whether an agent is epistemically blameless in continuing to believe p on the ground that she believes p and hasn’t encountered any defeaters for p, while the other emphasizes the question whether the agent has adequate grounds for believing p (the distinction is Alston’s). I will further argue (ii) that the answer to the first question depends on the epistemic sophistication of the agent under scrutiny, such as the extent to which the agent has beliefs about the comparative reliabilities of various of her belief-forming processes. While an agent who is naive in such matters may be blameless in believing p, and thus rational, another agent, who is epistemically more sophisticated, may be blameworthy in believing p, and thus irrational, even if both agents are in the same evidential situation. If this is correct, it follows that to the extent to which such norms are indeed part of the standards of rationality, to that extent those standards will have to be relativized to the degree of the epistemic sophistication of particular agents.