The Plausibility of Infallibilism
There are adequate reasons for accepting epistemic infallibilism (EI), which is the view that propositional knowledge requires epistemic certainty (i.e., the impossibility of being wrong that p given one’s evidence for p).[1] In other words, if S knows that p on the basis of evidence e, then S cannot be wrong that p given e.
To start, EI explains the problem of Gettier cases. Subjects in such cases do not possess knowledge; they have true beliefs that are to some extent luckily true, which prevents them from being items of knowledge. However, if one has epistemic certainty that p, then one is not vulnerable to having a luckily true belief that p.
One might have other reasons to accept EI; hence, EI is not open to the charge of being an ad hoc position taken simply for the sake of avoiding a chat with Gettier. For example, assuming epistemic invariantism, EI provides a plausible account for the cross-context uniformity of the invariant standards for knowledge: in every case of knowledge, epistemic certainty is required. In addition, EI explains better than epistemic fallibilism (EF) does why cases of knowledge are more valuable than cases of true belief which fall short of knowledge.[2] EI also provides a non-arbitrary difference between items of knowledge and items of belief which fall short of knowledge, while versions of EF seem beleaguered to find a non-arbitrary difference. EI appears to handle problems of pragmatic encroachment better than EF does. EI provides a substantive account for why knowing that p permits one to stop rationally investigating p, while EF does not provide an equally reasonable account. And EI explains why so-called concessive knowledge attributions (e.g., “I know that p but p could be false”) seem inconsistent.
Furthermore, there are independent arguments for EI. For instance, Moti Mizrahi contends that the factivity of knowledge entails EI, and that since most epistemologists hold that knowledge is factive, they are committed to EI.[3]
Now, a contextualist could try to solve the Gettier problem by dividing all knowledge claims into two categories: claims vulnerable to the Gettier problem (i.e., susceptible to epistemic luck) and claims not vulnerable in this way. For the former category, one might say that in such epistemic contexts, epistemic certainty is required for knowledge; the standards for knowledge are highest here, and the threat of Gettier is eluded. For the latter category, one might claim that the standards are lower; it is possible to have a justified true belief such that the justification falls short of epistemic certainty – as long as such beliefs are safe from Gettier problems.
It seems to me that this contextualist move is open to the charge of being an ad hoc attempt to avoid Gettier while also sidestepping the (unsavory for some) skepticism associated with EI.[4] Is there a sufficient independent reason to accept such contextualism? The contextualist might say that this view makes sense of our language about knowledge: sometimes we speak of knowing to mean that we are certain of what we claim to know, and sometimes we mean that we are less than certain but that we still know. The contextualist can make sense of both kinds of claims without reducing the second kind to something less than knowledge. This answer might not cut much ice with the infallibilist, since EI can make sense of both kinds of claims as well.
Moreover, it should be noted that not all Gettier problems are high-stakes practical issues demanding strict standards for knowledge, though at least some contextualists hold that it is high-stakes matters that require strict standards. The case of the sheep in the field is not necessarily a high-stakes epistemic situation, nor are the cases about Jones and Smith getting the job, about Brown being in Barcelona, and about the fake barns. Therefore, it seems strained for the contextualist to divide knowledge claims into those susceptible to epistemic luck and those not susceptible, and to hold that the former category requires stricter standards.
One might instead appeal to interest-relativism to address the Gettier problem. Here, one might claim that knowledge requires epistemic certainty only for those whose personal interests include avoiding Gettier problems, but that in other cases which are not susceptible to Gettier problems, or for folks who don’t care about Gettier problems, lower standards for knowledge apply. This move seems ad hoc as well. Why should the definition of ‘knowledge’ vary according to whether or not a person is interested in the Gettier problem? This view of knowledge seems guilty of being constructed merely for the sake of escaping the problem. Notwithstanding this point, interest-relativism does not seem to explain the other issues addressed in paragraph three above as well as EI explains them.
In sum, with respect to the problems addressed and the four theories of knowledge (EI, EF, contextualism, and interest-relativism) covered in this article, EI seems the best model for the nature of propositional knowledge.
[1] Epistemic certainty differs from psychological certainty, which is a subjective sense of being sure and thus a property of the person who has it, not of the proposition the person believes.
[2] The reason is that knowledge guarantees certainty and thus prevents one from error, while cases that fall short of certainty do not provide this benefit. As Socrates might put it, epistemic certainty “tethers” true belief and keeps it from wandering away. It should be noted that there are different versions of EF and that the claim that EI is explanatorily better than EF does not take into account the different versions of EF. Given the shortness of this blog post, I cannot address each of the versions here.
[3] See You Can’t Handle the Truth: Knowledge = Epistemic Certainty at https://philarchive.org/archive/MIZYCH
[4] As Mizrahi astutely notes, the fact that some people don’t like the consequences of an assertion is not good evidence against that assertion. Thus, the fact that people don’t like the skeptical consequences of EI does not indicate that EI is false.