This post is by Will McNeill, currently a teaching fellow at King's College London. He researches the epistemology of perception and recognition.
The main focus of PERFECT is on finding the epistemically valuable within mental pathologies of varying seriousness; of identifying pathological beliefs which are at least epistemically innocent. But in this post I want to bring our attention back to the cognitive processes of those with no serious cognitive impairments. It would not surprise me if many of our cognitive processes turn out to be epistemically innocent in a particularly direct way. They may turn out to be reliable, and to produce justified beliefs, while not being capable of explaining why the beliefs they produce are justified. Justified beliefs which are the direct products of such epistemically innocent processes are – I believe – foundational.
Suppose that we had good reason to think that someone was good at spotting emeralds. They tend to know, of emeralds, that they are emeralds. And they tend not to be fooled by the fakes. But suppose too that on other grounds we had good reason to think that the cognitive processes which generate their emerald beliefs relied in part on the thought that emeralds are grue.
On the one hand, we would have discovered something about the nature of their reliable capacity to spot emeralds. But on the other hand, we would not have produced any satisfying answer to the question of why they were warranted in their belief. At least, discovering that these deviant cognitive processes were at work would provide no more of an explanation of their warrant than the simple observation that they were de facto reliable at spotting emeralds.
The grue example is very obviously contrived. It is hard to imagine why our cognitive theory would have us positing grue generalizations. But think of those early ocean navigators. We have good reason to suppose that their heads were littered with false beliefs about the relationship between stars, sun, moon, and position. With imperfect tools they fashioned surprising reliability. Even if they had good warrant for holding some of these beliefs, and even if some of them were true, they are not in terms of their content sufficient to explain how the navigators knew where they were. Given how much falsity resided in the relevant cognitive processes their products could hardly be thought warranted on theoretical grounds. If the navigational beliefs such cognitive processes secured were good it was only because they were reliable. If there is knowledge here it is more intuitive than inferential.
Hopefully the thought is becoming clear. In both of the above cases, the actual cognitive processes at work appear epistemically innocent. They are not good in their own right. They do not manifest such theoretical virtues as parsimony, accuracy, or explanatory power. If they are good this is just because in the contexts in which they are employed they reliably secure the right results.
I see no reason to think that these kinds of epistemically innocent processes would be isolated or unusual. When we are engaged in modelling actual cognitive processes we want to be sensitive to the kinds of pressure that they are under. They need to be fast and frugal. Our models need to be evolutionarily plausible. But these criteria lend themselves to positing cognitive shortcuts, heuristics and other quick fixes. Positing such tricks might best describe the methods by which we reliably form true beliefs about our environment. But to the extent that such descriptions genuinely involve tricks, shortcuts, false generalizations, and biases they are surely epistemically unsatisfying. Or to put it another way, they are in a very direct way epistemically innocent. They may be reliable inferences. But they secure no distinctively inferential warrant.