Suppose, with Boghossian, that we distinguish epistemic rules and principles. Both are conditionals, but rules have imperatival consequents whereas principles are true or false. To each rule, there corresponds a principle: if the rule says “In C, believe p,” the principle can say either “In C, it is obligatory to believe p” or “In C, it is permissible to believe p.”
Let’s simplify a bit a say that a cognitive system operates in accord with a rule when the imperatival consequent is fulfilled and the antecedent describes the causal/explanatory factors that prompted (in the near enough past) the belief formation in question. We’ll have to say something at some point about deviant causation, but I’ll ignore that for now. Let’s then say, also simplifying, that the good rules are ones that, by operating in accord with them, a person can arrive at a doxastically justified belief. Not that this result is always achieved, but that it can be achieved, i.e., the rule specifies in its antecedent a possible basis for a doxastically justified belief.
Given this stagesetting, it is pretty obvious that the epistemic principles that correspond to the rules that we operate in accord with will be mostly false.
We see an oak tree in the back yard and believe that it is an oak tree–the causal/explanatory path goes from a certain perceptual state to a belief state. The belief is doxastically justified in virtue of being based on the perceptual state itself. But both of the following epistemic principles are false:
When you are appeared oak-tree-ly, it is obligatory to believe that an oak tree is present.
When you are appeared to oak-tree-ly, it is permissible to believe that an oak tree is present.
The primary difficulty with both principles is that they lack a defeater, or grounds for doubt, clause. Perhaps our perceptual skills develop to the point that we come to be on alert for defeaters as we mature, but that is certainly not true of early cognition.
Once we appreciate this point, the connection between rules and principles can no longer be taken to be a simple one. The hard question is this: take your stock of true epistemic principles, including the global one that specifies the conditions under which any belief is justified. What rules must cognition operate in accord with in order for the resulting beliefs to be justified?
Here’s a first step: strip true epistemic principles of their defeater clauses, and the “function in accord with” requirement for doxastic justification is met by functioning in accord with the corresponding rule. Then, if the defeater clause is actually satisfied and any other requirements for justification are met (including any purported requirement that one follow the rule in addition to operating in accord with it), then the belief is justified.
The interesting question from my perspective is whether further distancing between rules and principles is needed. Herein lies potential catastrophe. Here’s one example. Suppose you think of some domain of belief as skill-based, such as perceptual belief. Then you think of the true epistemic principles as specifying in their antecedents skill-displaying features. Finally, you notice that when people are developing their skills, they behave in a stilted fashion and are even instructed to do so: think of, for an analogy, the difference between someone following instructions about how to flyfish and someone who knows how to flyfish. And then the pressure is felt to avoid having to say that those developing their skills can’t have any justified beliefs.
In the face of this pressure, two options are available. The first is to abandon the skill-based picture of (some of) cognition, but that requires starting over on a new research program. So the temptation is to take the less revolutionary approach, and allow more distancing between rules and principles. The rules no longer need to specify in their antecedents displays of skill relevant to the kind of belief in question, but can specify only operations relevant to mastering the skill in question.
So now you have epistemic principles that tell you when it is permissible to form beliefs, and you have rules that it is, in some sense, OK to follow that are unrelated to any true epistemic principle. They are not derived from principles by stripping the antecedent of a defeater clause, and yet the motivation for presenting such rules has something to do with wanting a theory that allows justified beliefs to be obtained in the process of learning a skill.
The critic presses: “this makes no sense! The beliefs, you say, are justified even when the skill is not yet present, but no epistemic principle allows that such beliefs are permissible.” And here the temptation is to reply: “Oh, that’s OK, the skill-based principles specify conditions for objective justification and the rules specify conditions for subjective justification.”
Depending on their patience with silly philosophy, critics may just walk away shaking their heads. Or, if they stay, the proper thing to say is this. Did you mean it when you formulated principles that defined the arena of impermissible belief? You had to, or else your principles are false. So, whatever “subjectively justified” means, it carves out a subclass of beliefs that you simply should not hold. Period, no qualification allowed.
So further distancing seems to me to fail. The rules we operate in accord with often (typically?) correspond only to principles without defeater clauses in them, but a good theory should not countenance any further distancing between the two.