Ways to Categorize Theories of Knowledge
There are many ways to classify theories of propositional knowledge. Philosophers often organize such theories into dichotomies: infallibilist and fallibilist theories; invariantist and variantist positions; skeptical and anti-skeptical views. But there are other ways of classification, which I will address below. First, however, a terminological summary is in order.
Epistemic infallibilism is the view that propositional knowledge requires epistemic certainty, which is the highest degree of justification. If S knows that p, then p is epistemically certain for S given S’s justification for p. If p is epistemically certain for S, then p has a probability of 1 for S, given S’s evidence. In other words, S cannot be wrong that p given S’s evidence for p. According to some infallibilists, one’s belief cannot be both justified and false. (I recently published a paper which suggests a distinction between precise justification (PJ) and loose justification (LJ), enabling the infallibilist to hold that knowledge requires PJ but that belief with LJ can be both justified and false.) In common language, if Smith knows that p, then p is 100% probable for Smith. Does Smith know that p? Yes, Smith is 100% sure. (He is epistemically sure, but not necessarily psychologically sure, though he ought to be that, too, given his evidence.) For example, Smith’s knowledge that 1 + 2 = 3 is epistemically certain.
Epistemic fallibilism is the theory that propositional knowledge is consistent with having a degree of justification which is less than certain, i.e., lower than a probability of 1. On this view, if S knows that p, then S has a justified, true belief that p, but it is epistemically possible for that belief to be false because the degree of justification could be under 100%. On this view, it is possible for Smith to know that p and yet he could be wrong because his evidence does not guarantee the truth of p. For the fallibilist, justification is not sufficient to guarantee truth.
Epistemic invariantism is the claim that the standards for propositional knowledge do not change according to epistemic context. Whatever those standards are, they remain uniform across all epistemic situations. In other words, the truth value of claims such as “Jones knows that the pencil is on the table” does not change according to epistemic circumstances. Infallibilism is generally considered a version of invariantism, but fallibilist positions can be invariant as well.
Epistemic variantism holds that the standards for propositional knowledge vary according to circumstances. Hence, the truth value of claims such as “Jones knows that the pencil is on the table” can change relative to pertinent factors.
There are at least two versions of epistemic variantism: contextualism and interest-relativism. According to contextualism, the truth value of claims such as the one about Jones varies according to the epistemic situation: epistemically demanding situations require higher standards for knowledge and epistemically relaxed situations have lower standards. Thus, the claim about Jones might be true in epistemically relaxed circumstances, but false in epistemically demanding ones. According to interest-relativism, the truth or falsity of such claims is relative to Jones’ pragmatic interests. In other words, whether Jones knows that p depends on Jones’ pragmatic interests with respect to believing that p. If the practical cost of Jones’ being wrong about p is high (say, if Jones is wrong about p, then he will lose his car), then the standards for knowing p are higher. But if the cost is low (say, he will lose his pencil), then the standards are lower.
Epistemic skepticism is the position that many or most of the items that human beings take themselves to know are not genuine items of knowledge. One might be a skeptic because one is an infallibilist: since knowledge requires certainty and we are certain about very few kinds of propositions, most of our knowledge claims are, precisely speaking, false. For instance, we know that 1 + 2 = 3, but we do not know that Julius Caesar crossed the Rubicon in 49 B.C. Anti-skepticism rejects skepticism, holding that we know much or most of what we commonly take ourselves to know. An anti-skeptic might say we know that 1 + 2 = 3 and we know that Caesar crossed the Rubicon in 49 B.C.
Now, although theories of knowledge are categorized according to the descriptions above, it seems to me that such theories can be usefully organized in other ways. For example, one might organize them into the categories of being vulnerable to epistemic luck or not. Those theories which avoid Gettier problems (e.g., infallibilist theories) go into the latter category; those which are exposed to Gettier problems sit in the former category.
Alternatively, theories which are sensitive to pragmatic encroachment go into the encroachment category, and ones which aren’t susceptible go into the non-encroachment category. According to the view of pragmatic encroachment, whether S has knowledge depends on practical factors for S. For instance, if little of Smith’s skin is riding on the claim “The berry is safe to eat” then the standards for Smith knowing that claim are low. However, if Smith is severely allergic to berries and might die if he eats raspberries, then the standards are higher. (Encroachment issues are similar to factors concerning contextualism and interest-relativism, discussed above.)
Another way to divide theories of knowledge is to place them into the dichotomy of error-prone vs. not error-prone. In the first class, one would place fallibilist theories; in the second class, infallibilist ones. Or, one might divide theories into the following categories: those which require or permit a rational agent to engage in further rational inquiry concerning p, and those which permit one to stop rational inquiry concerning p. In the latter class, one would put infallibilist theories. If Smith infallibly knows that p, he need not continue inquiring into p. For example, if Smith knows that there are no square circles for sale at the local department store, he need not conduct empirical observation to confirm what he knows. In the former class, one would place fallibilist theories which are prone to misleading evidence and thus require a rational agent to continue investigation. For instance, Detective Jones’ current evidence indicates that it is somewhat reasonable to conclude that Brown is guilty of the crime. But the evidence is fallible and incomplete. Jones thus recognizes that he ought to continue the investigation.
It seems to me that each of these alternatives for typology might be helpful depending on what epistemic factor one seeks to emphasize.