Summary/Analysis of "Epistemic Luck" Chapters 1 and 2

(Note: What follows in this post is mostly my own summary of "Epistemic Luck" (Chs. 1 and 2), although I've added some critical analysis at some points; mostly, though, my attempt was to get clear on what the main claims were. I am open to suggestions as to whether I have some points misconstrued. --AC)

Chapter 1: Scepticism in Contemporary Debate

The task of this chapter seems to be twofold; first, to explain what will be a key term in Part two: the epistemic luck platitude (i.e. the pervasive supposition that knowledge excludes luck) and pose as a question (which will not be addressed until Part II) how it is that the epistemic luck platitude is a motivation for skepticism; secondly, Pritchard presents two separate motivations for skepticism that drive the contemporary debate, the infallibilism-based radical skeptical argument, and the closure-based radical skeptical argument.
His argument is that the closure-based argument is logically weaker, and because it is also capable of generating the radical skeptical conclusion generated by the infallibilsm-based argument, it is the most appropriate for the antiskeptic to address. A correlative point Pritchard makes is that, because the closure-based argument is what should be dealt with by the antiskeptic (and not the logically stronger infallibilism-based skeptical argument), we can conclude something important about the relationship between skepticism and epistemic luck, namely, that we are prevented from “identifying the skeptical challenge as simply arising out of an unduly demanding—indeed, an unqualified—reading of the claim that knowledge excludes luck. We would only do this if tackling the skeptic amounted merely to tackling the infallibilism-based argument. However, because the task is more difficult (i.e. to combat the closure-based argument), we can see that the skeptical challenge arises out of a “qualified” reading of the epistemic luck platitude. However, it does not follow from the fact that skepticism is motivated by a qualified reading of the epistemic luck platitude that that qualification requires adopting fallibilism. Closure provides an alternative to fallibilism, and among motivators for skepticism, appears to be the strongest contender.
Two principles are introduced in this chapter, which serve for the basis of the two arguments he presents that motivate skepticism.

(1) Infallibility Principle: For all agents, ϕ, if an agent knows a proposition ϕ, then that agent knows that all error-possibilities associated with ϕ are false.

(2) Closure Principle for Knowledge: For all agents, ϕ, ψ, if an agent knows that ϕ, and knows that ϕ entails ψ, then that agent knows that ψ.

Let’s see why closure is “logically weaker”: to do this we must identify that both principles can be used to generate the radical skeptical conclusion, and secondly, that between two arguments which generate the same conclusion, the logically weaker (i.e. closure) is the “stronger” argument to rebut. Here are the two arguments which are formed from these respective principles.

The infallibilism-based skeptical argument
1. If one is to have knowledge of a wide range of everyday propositions, then one must know the denials of all error-possibilities that are associated with these propositions, including radical skeptical hypothesis.
2. One cannot know the denials of radical skeptical hypotheses.
3. Therefore, one cannot have knowledge of a wide range of everyday propositions.

The closure-based radical skeptical argument
1. If one is to have knowledge of a wide range of everyday propositions, then one must know the denials of all radical skeptical hypotheses that one knows to be incompatible with the relevant everyday propositions.
2. One cannot know the denials of radical skeptical hypotheses.
3. Therefore, one cannot have knowledge of a wide ranger of everyday propositions.

∀x [(Ks(Ixt ) → EAxt]

(For all x, if x is incompatible with a target proposition and s knows this, then x is an error-possibility associated with that target proposition).

However,

~∀x [EAxt → Ks(Ixt)]

This is because not all error-possibilities associated with a proposition are known by the agent to be incompatible with that proposition. The entailment which takes place in the closure principle involves only those propositions that the agent knows are incompatible with the target proposition. And so, the set of those error-possibilities that must be ruled out according the infallibility principle could be, in principle, larger than the set that the agent knows to be incompatible with the target proposition, however, the inverse could not be the case. And hence, the closure-based argument requires that we rule out (in principle) less error-possibilities than the infallibility-based argument requires. Because the closure-based argument can reach the same conclusion (radical skepticism) as the infallibility-based argument, and with a smaller set of error-possibilities we must rule out, it is a more dangerous threat for the anti-skeptic.

Ch. 2: Closure and Context

By the beginning of Ch. 2, “Closure and Context”, Pritchard hopes to have already have taken us aboard in adopting a template closure-based skeptical argument over the logically stronger infallibility-based template skeptical argument; secondly, he hopes that, in making this move, we will come to reject the idea that the skeptical problem merely arises from a robust reading of the claim that knowledge excludes luck.
How can the skeptic, viz. motivating skepticism with the closure-based skeptical argument be rebutted? This purpose of Chapter two is to offer two attempts, which ultimately fail. These attempts are (1) to deny the closure principle, a move made most notably by Fred Dretske, and (2) to embrace the position of attributer contextualism. Pritchard shows how each of these two strategies attempts to counter the skeptic with the closure-based argument in mind; he argues, though, that each fails, and for a common reason. Both denying closure and embracing attributer contextualism are examples of externalist positions which are, as he says, “at certain key junctures, implicitly motivated by internalist epistemological intuitions. ” Ultimately, Pritchard’s goal here isn’t so much as to prop up two straw men, as to instructively show why particular responses can’t work in the face the closure-based motivation for skepticism, and only given such instruction, does it become as blatantly clear (as Pritchard seems to want it) why a third response, Neo-Mooreanism, is the most desirable strategy. Merely arguing for neo-Mooreanism would not have been so persuasive if it were not made obvious why other attempts run into dead ends.

Denying Closure: Dretske

Dretske proposes that we can know a given proposition Φ while at the same time lacking knowledge of the known entailment Ψ, because “the truth of Ψ is already presupposed in the agent’s knowledge of Φ.” What is significant here is that, when a knowledge operator is applied to Φ, it doesn’t “penetrate the (known) entailment Ψ because, qua being a presupposition of the agent’s knowing Φ, it is not part of what is being operated on when the knowledge operator is applied to the known proposition.

Pritchard captures the force of what Dretske’s up to here in saying that: “most knowledge quite legitimately presupposes the truth of certain propositions that one does not know. ” Denying closure allows us to preserve a motivation for fallibilism that was advanced by J.L. Austin’s “relevant alternatives” proposal: the proposal that to know an everyday proposition, one need only rule out error-possibilities that one has a reason to take into account. We must be careful, though, not to conflate the two. What is preserved of the this fallibist motivation by the closure-principle is the possibility that one can have knowledge of an everyday proposition without knowing the denial of skeptical hypotheses (which would be incompatible with knowledge of that proposition). It is in this sense that the closure principle can preserve a fallibilist motive. However, there is more to denying closure than merely rejecting the knowledge operator entailment. Dretske also advocates a sensitivity-based theory of knowledge, which is externalist (i.e. it rejects that facts needed to justify a proposition must be reflectively accessible), and which preserves his anti-closure position previously mentioned. The sensitivity principle Dretske advocates is as follows:

Sensitivity Principle: For all agents, Φ, if an agent knows a contingent proposition Φ, then the agent does not believe that Φ in the nearest possible world or worlds in which ~Φ.

Interestingly, accepting this principle (as Dretske does) as essentially the necessary and sufficient condition for knowing a proposition, leads us to the fallibilist-intuition-preserving-friendly result that we can know everyday propositions (which do meet the sensitivity condition) whilst not knowing the denial of skeptical hypotheses which we know to be entailed by the everyday propositions we know, because the denial of skeptical hypotheses will fail to meet the sensitivity condition. They will fail for the reason that, even if we believe the denial of a skeptical hypothesis and it is true, we would still believe it was true in the nearest possible world in which the denial was false (i.e. in worlds in which we really are BIVs).
One problem of denying closure Dretske-style is that there emerges an asymmetry between the relevant-alternatives motive for fallibilism, and Dretske’s position. The asymmetry can be presented as follows: on the relevant-alternatives view, we need not know the denials of skeptical error-possibilities because (given that they are presumed to be in such far-worlds) they are not relevant to our knowledge of everyday propositions. However, what is interesting is that in our discussion of the denials of skeptical hypotheses, the denial of the skeptical hypotheses is relevant to knowledge. And so, in principle, it seems that the core relevant-alternatives thesis would allow that we would need to know the denial of skeptical hypotheses in at least some contexts; however, because Dretske’s position would never allow us to know such a denial (given that knowledge of it fails to meet the sensitivity condition), we should, Pritchard claims, “be wary of any view which proclaims itself to be the true heir to the relevant alternatives tradition.” I suspect Pritchard means this both in the sense that the relevant alternatives tradition is problematic in that it is somewhat ambiguous (in that it has a double standard, arguably) and in the sense that Dretske’s view claims itself to be an heir to a position from which it appears to relevantly stray.
It seems, though, that Pritchard’s central problem with Dretske’s closure-denying-sensitivity approach lies not in its relation to the relevant-alternatives thesis from which it is a supposed heir, but rather, with a problem which involves an ambivalence in motivation. The problem is that: Dretske’s position endorses an externalist theory of knowledge. Given this, Dretske’s view must be committed to not requiring internalist justification as necessary for knowledge. Pritchard thinks, though, that internalist intuitions are just what Dretske his in mind when explaining how it is that we lack knowledge of the denials of skeptical hypotheses. The passage of Dretske’s which Pritchard quotes to support this claim is:

If you are tempted to say [that the agent does know (Q)…[, think for a moment about the reasons that you have, what evedence you can produce in favour of this claim. The evidence you had for thinking them zebras has been effectively neutralized, since it does not count toward their not being mules disguised. Have you checked with the zoo authorities? Did you examine? Did you examine the animals closely enough to detect such a fraud? (Dretske 1970: 1016)

Pritchard claims here that, if we substitute a radical skeptical hypothesis in for the local one in this passage, then Dretske would be claiming that what prevents an agent from knowing the denial of a skeptical hypothesis “is that she is unable to adduce good empirical evidence in favour of thinking that this position is true. Since there is, intuitively at least, a close connection between the evidence that one is able to explicitly adduce and the evidence that is reflectively available to one, it follows that a natural way of reading this passage is as saying that agents are unable to know the denials of skeptical hypotheses because they lack internalist justification for their beliefs in these propositions.” (53) Pritchard’s problem with this is that, if Dretske’s position, qua being externalist, denies the necessity of internalist justification for knowledge, then it is a problem that he motivates the denial of closure (which I take it is what he was up to in his passage regarding the local hypotheses) with what appears to be an explanation for a lack of knowledge resulting from a lack of internalist justification.
I’d like to comment on this problem briefly before moving on to the second anti-skeptical strategy, which is attributer contextualism. I think that Pritchard’s motivation-based criticism of Dretske’s position is intuitively right-on. Although some more reflection led me asking questions down this avenue:
Is his position properly externalist? If it is, then that doesn’t seem completely evident by the definition given of the sensitivity principle. Recall that the sensitivity principle is claiming that:

Sensitivity Principle: For all agents, Φ, if an agent knows a contingent proposition Φ, then the agent does not believe that Φ in the nearest possible world or worlds in which ~Φ.

The consequent is something that could hold independent of an agent’s reflective accessibility to it (i.e. one might have no idea whether something holds in nearest possible worlds), and so the consequent represents an externalist condition. Also, the sensitivity principle presents the externalist condition as a necessary condition. Epistemological internalism with regard to knowledge has, itself, one necessary condition: i.e. that the belief must be internally justified (i.e. all facts needed to justify the proposition must be reflectively accessible). I’m probably going wrong somewhere here, but it seems to me that: a position can require that all facts needed to justify a proposition must be reflectively accessible (i.e. internalist with regard to knowledge) and stipulate that, for a proposition to be known, not only must this internalist justificatory condition be met, but also, an externalist (non-justifying) condition must be met (i.e. the consequent in the sensitivity principle). Notice that the consequent in the sensitivity principle need not be understood as a justificatory condition (maybe I’m wrong here). And so it seems as though an externalist with regard to knowledge, in so far as she stipulates an externalist condition for knowing (i.e. as Dretske has done in his consequent of the sensitivity principle), and still require internalist justification for a belief. (For how is requiring internalist justification for knowledge incompatible with requiring an externalist condition for knowing?)
I think there are two points that are at the source of what is probably just confusion on my part. First off, my point hangs on whether the consequent of the sensitivity principle is to be read as a justificatory condition. Secondly, I don’t think I am fully clear as to the extent to which the sensitivity principle is externalist by virtue of its requiring an externalist condition for knowing. If Dretske’s position is externalist for this reason (i.e. that it requires an externalist condition for knowing), then I’m not clear as to why this externalist-making feature of the sensitivity principle would require a rejection of internalism with regard to knowledge (which is what Dretske’s position must do if it is properly externalist).
The next anti-skeptic alternative presented is attributer contextualism, a position that combats skepticism with the added benefit of remaining true to the intuitive closure principle. Attributer contextualism amounts to the position that, for a given proposition, its truth value will be in part determined by the conversational context of the proposition circumscribed by the attributer of that proposition. This is the position advanced most notably by DeRose (2005), Lewis (1996) and Cohen (2000). A consequence of attributer contextualism is that a given proposition, “S knows that p” can be uttered in one circumstance and be true, and in another circumstance and be false. Obviously, a motivation for adopting such a view would be to preserve that we can have quite a bit of the knowledge we think we have. Why? Because, given that (we grant that) skeptical hypotheses (and the task of ruling them out) are not relevant to most everyday circumstances in which we think we have knowledge, it is not necessary to rule them out in order to say that we know (for example) that it is raining today. Ruling out the denials of skeptical hypotheses is relevant only in the context of decisively skeptical conversations. Attributer contextualism seems on the surface to be capable of preserving two generally incompatible intuitions: (1) The infallibilist intuition that skeptical conversational contexts lead us to discover that we lack knowledge of the denial of them, and a fortiori, we lack most of the knowledge we have; (2) In everyday contexts, we “know” a great deal (p. 55) of what we think we know.
An instinctive reaction would be to think that, surely, these apparently contradictory claims cannot be reconciled without denying closure, and so how can attributer contextualism preserve closure (whilst in any sense remaining true to these antithetical intuitions?) The solution for attributer contextualism is to claim that closure is preserved only within the same conversational context. And so, John’s knowing that the bottle is on the table whilst not knowing that he is not a BIV doesn’t deny closure because his lack of knowledge of the latter is appropriate only within a skeptical conversational context in which he wouldn’t have even known the bottle was on the table.
Pritchard gives two objections to attributer contextualism. The first is somewhat comical; he says, “The most obvious difficulty with attributer contextualism is its commitment to the counter-intuitive thesis that ‘knowledge’ is a context-sensitive term sin the manner described, with the epistemic standards relevant to whether or not knowledge can be truly ascribed to an agent being determined by conversational factors” (58). This is another way of saying, “One problem for attributer contextualism is that its thesis amounts to a reductio against it.” And, indeed, it seems that way. If we grant the platitude that knowledge entails truth, and also grant (as A.C. wishes to) that knowledge is context sensitive, then truth is context sensitive; further, if truth is understood (loosely) as having something to do with how things are, then how things are would be relative to presuppositions that are held within discourse. How odd that would be.
The other problem that is presented is in tune with the second problem presented to Dretske’s anti-skeptical strategy of denying closure: and the shared problem here is a problem with motivation—specifically, the problem of offering externalist positions which are motivated in part by internalist commitments. The way the argument goes (contra attributer contextualism) here seems to be this: A.C. commits itself to the claim that we can know the denial of skeptical hypothesis. This commitment arises once closure is preserved within everyday contexts in which we claim knowledge of any everyday proposition. However, it has been argued that no internalist account can account for knowledge of the denials of skeptical hypothesis (given that no internalist justification is available to support such a claim). Resorting to the available externalist option, though, will lead one to wonder why it is necessary to import all of the “sophisticated theoretical machinery” (60) of contextualism. “Why not just simply argue that agents know everyday propositions and thus, given closure, that agents can know the denials of skeptical hypotheses also” (60). The point the author appears to be making is that: once the denials of skeptical hypotheses are claimed to be “known”, then the skeptical problem seems to disappear in such a way that one ought not resort to the apparently counterintuitive contextualist assumptions in order to explain it. (My aside: ironically, it was importing just this machinery that got us to the position of having externalist knowledge of the denials of skeptical hypotheses, knowledge of which would render the machinery cumbersome). And, on another note, the externalist knowledge of denials of skeptical hypotheses which are closurelly-entailed by everyday knowledge in normal contexts, is not obviously explained. Says Pritchard: “Of course, a great deal needs to be said to explain how one could know the denials of skeptical hypotheses even on an externalist theory of knowledge” (60). A.C. doesn’t have a very clear answer to this, even though it is committed to it.
The moral of the story seems to be this: we’ve looked at two anti-skeptical strategies, and independent of other problems of these respective strategies, each is inconsistently motivated. On this note, the dialectical transition will be to investigate a third anti-skeptical strategy that is not inconsistently motivated: neo-Mooreanism. Unfortunately, though, we’ll see that whilst neo-Mooreanism looks quite good compared to the Dretskean and attributer contextualist anti-skeptical strategies, there is another problem on the table, which has to do with any externalist response to the skeptical dilemma which does not recognize that the tenor of the skeptical strategy is essentially internalist.