The standard account of knowledge states that one of the requirements for a belief to be knowledge is that it must be true. Because of this requirement it is an externalist account of knowledge, meaning that whether a belief is knowledge or not depends, at least in part, on factors that are not consciously available to the subject. And the standard account of knowledge is not the only externalist theory about knowledge, reliabilism is another such theory, which states that a belief can only be considered knowledge if it is generated by a process that generally results in true beliefs, again something that is not consciously available to the believer*.
But does it make sense to give an externalist account of knowledge? We can certainly give an externalist account for many other objects of philosophic study, for example truth, without difficulty, so why not knowledge too? Problems arise, however, if we expect our theory of knowledge to be normative. By this I don’t mean that it should make knowledge normative, that is a given; I mean that the theory should be able to aid us in acquiring knowledge, or at least in separating our mere beliefs from those that are knowledge. An externalist account can’t perform this function, since such accounts define knowledge in terms of factors that are not consciously accessible to the subject. And this is a serious problem for externalist accounts of knowledge, because without the ability to be normative such theories seem like philosophical curiosities, and not interesting contenders for a definition of what knowledge is. Sure, they define a category of beliefs, but since we can’t tell which beliefs fall in that category why should we care?
From such considerations it might seem as if we should simply turn our backs on externalist theories. Even if externalists about knowledge demonstrate one difficulty after another facing internalist accounts of knowledge there would still be no good reason to turn to externalism as the solution, since externalism simply can’t provide the kind of theory that we are searching for. Of course nothing is ever so clear-cut; one potential problem for the internalist program, pointed out by Hilary Kornblith, is cognitive sanity requirements.
A cognitive sanity requirement, in general, is a requirement that the mind of the person who is to have knowledge is reporting with some accuracy upon its own operations to consciousness. If our internalist theory defines knowledge as a belief that is reflected upon appropriately (of course, this is not an actual internalist theory about knowledge, it is far too vague) then we may want to include a cognitive sanity requirement to ensure that the person who is to have knowledge isn’t mistaken about reflecting upon their beliefs. We could imagine situations in which a person always believes that they have reflected appropriately upon their beliefs, mistakenly; are all of their beliefs to be considered knowledge then? Surely not, and thus we add a cognitive sanity requirement.
The problem with cognitive sanity requirements is that by adding them to a theory we make that theory externalist, since in principle it is possible for a person to be unaware of the fact that their mind is operating inaccurately. Is this the end for internalism? No, the way out of this apparent dilemma is to reject cognitive sanity requirements. The intuitions leading us to believe that we need a cognitive sanity requirement are the same ones that lead some people to think that knowledge is necessarily true; they are requiring too much of knowledge. Internalism about knowledge means that we should view knowledge as the best possible beliefs that a person can have given their limitations (for example, what experiences they have had), with best being defined differently in different internalist theories, of course. Thus what is knowledge may vary from person to person, given their cognitive resources, and there is no guarantee that the beliefs picked out as knowledge are true.
* Some might object, saying that we can know whether a belief is true or formed reliably. Unfortunately this is quite circular, since our knowledge of that fact relies on factors that are themselves not consciously available to us, and knowledge about those factors would rely on different external factors, ad infinitum.