It is of course true that a belief, in order to be justified, needs to have been formed and sustained by a reliable epistemic practice. But in the case of rational inference, what is the practice supposed to be. The reader is referred, once gain, to the description of a reasoning process given a paragraph back. Is this not, in fact, a reasonably accurate description of the way we actually view and experience the practice of rational inference and assessment/ It is furthermore, a description which enables us to understand why in many cases a practice is reliable—and why the reliability varies considerably depending on the specific character of the inference drawn and also on the logical capabilities of the epistemic subject. And on the other hand, isn’t it a severe distortion of our actual inferential practice to view the process of reasoning as taking place in a “black box,” as the externalist view in effect invites us to do? Epistemological externalism has its greatest plausibility in cases where the warrant for our beliefs depends crucially on matters not accessible to reflection—for instance, on the proper functioning of our sensory faculties. Rational inference, by contrast, is the paradigmatic example of a situation in which the factors relevant to warrant are accessible to reflection; for this reason, examples based on rational insight have always formed the prime examples for internalist epistemologies.
There is also this question for the thoroughgoing externalist: How are we to satisfy ourselves as to which inferential practices are reliable? By hypothesis, we are precluded from appealing to rational insight to validate our conclusions about this. One might say that we have learned to distinguish good reasoning from bad reasoning, by noticing that good inference-patterns generally give rise to true conclusions, while bad inference-patterns often give rise to falsehood. (This of course assumes that our judgments about particular facts, especially facts revealed through sense perception, are not in question here—an assumption I will grant for the present). But this sort of “logical empiricism” is at best a very crude method for assessing the goodness of arguments. There are plenty of invalid arguments with true conclusions, and plenty of valid arguments with false conclusions. There are even good inductive arguments with all true premises in which the conclusions are false. There are just the distinctions which the science of logic exists to help us with; basing the science on the kind of ham-fisted empiricism described above is a hopeless enterprise.
William Hasker, The Emergent Self (Cornell, 1999), pp. 74-75. From the chapter "Why the Physical Isn't Closed."