Concept

Epistemic Contextualism

Intro

Most people, in ordinary life, will say something like "I know the bank is open Saturday; I was just there two weeks ago." That sentence is fine. Nobody would correct you.

But imagine the stakes change. Your mortgage payment is due Monday. Your wife asks, "Are you really sure? They change hours sometimes." You hesitate. You realize you have not actually verified it this week. And now you find yourself saying, "You know what, I don't really know."

Same evidence. Same memory. Same world. In one conversation you know; in another you do not. What changed?

Epistemic contextualism is the view that the word know itself shifts what it means depending on the conversation's standards. When stakes are low and nobody is raising weird objections, the bar for "knowing" is set fairly low; ordinary good evidence counts. When stakes are high or a skeptic is raising live alternatives, the bar shoots up; the same evidence no longer clears it. Both speakers can be telling the truth, because know in the casual conversation is doing different work from know in the high-stakes one.

The position was developed by epistemologists like Keith DeRose, Stewart Cohen, and David Lewis in the late 1900s. It has a striking application to skeptical arguments. The famous brain-in-a-vat puzzle goes: if you cannot rule out that you are a brain in a vat being fed fake experiences by a mad scientist, then you cannot really know that you have hands. In ordinary conversation, the brain-in-a-vat hypothesis is not on the table, the standard is low, you know you have hands. In the philosophy seminar where the skeptic raises it, the standard rises, and you really cannot prove you have hands. Contextualism says both are true in their respective contexts.

This has a lot of bearing on Christian apologetics. When an atheist says "you don't really know Christianity is true because you can't rule out [some hyper-skeptical scenario]," the move is to raise the standard. Sometimes that is fair; sometimes it is the philosopher's standard pretending to be the universal standard. Contextualism gives the believer a principled way to say, "by the standards of ordinary historical and personal evidence, I do know. You are inviting me to play at a different and arguably impossible standard, and most of your other beliefs would also fail under that one."

The page below works through the technical machinery, the bank cases and lottery cases that anchor the position, the main objections (especially from contextualism's main rival, subject-sensitive invariantism), and the apologetic uses.

In full

The position in epistemology that the predicate "know", and so what counts as knowing, is context-sensitive. The same person, with the same evidence, in the same world, can correctly be said to "know" p in one conversational context and correctly be said not to "know" p in another. The shift is driven by changes in the conversational standards for knowledge, principally by the salience of error possibilities. Contextualism is principally a semantic thesis (about the meaning of knowledge-attributions in different contexts), not a metaphysical one (the underlying epistemic state of the knower doesn't change). Its anti-skeptical force is to render the skeptical and ordinary claims both true in their respective contexts: in everyday conversation, we know we have hands; in the philosophy seminar where the skeptic raises the BIV scenario, we don't. The resolution is not to find which is really true but to recognize that "know" picks out different relations in different settings.

The standard motivation

Contextualism gets its initial support from a pattern of intuitions about knowledge-attributions across different conversations. Examples (from the Bank Cases of DeRose 1992 and Cohen 1986; Hawthorne's "Lottery" cases):

Bank Cases

Bank Case A (low stakes): It's Friday afternoon. I am driving past the bank with my wife. I notice the long lines. My wife says: "We should deposit the check today; the lines aren't normally this long on Friday." I say: "I know the bank is open Saturday morning, I was there two Saturdays ago. Let's wait." This seems perfectly fine; I do know.

Bank Case B (high stakes): Same setup, but in this case our mortgage is due Monday and depends on the deposit. My wife: "Banks change their hours sometimes. Are you sure you know the bank is open Saturday?" I pause and reflect: I was there two Saturdays ago, but maybe their hours changed. I say: "Hmm, you're right, I don't really know." This also seems fine.

The data: I have the same evidence in both cases (the memory of being there two Saturdays ago); but I correctly say I "know" in (A) and correctly say I don't "know" in (B). Contextualism explains this by saying the standards for "know" rose in (B) because the practical stakes raised the salience of error possibilities.

Skeptical Cases (parallel structure)

Ordinary context: "I know I have hands." Philosophy seminar context: "Hmm, can I really know I have hands? I might be a brain in a vat. So I don't know I have hands."

Both attributions are correct in their contexts, on the contextualist view. The skeptical argument works because in the philosophy-seminar context, the standards for "know" have been raised by the salience of skeptical hypotheses; the ordinary attribution works because in the ordinary context they have not been.

Three flagship contextualist proposals

David Lewis, "Elusive Knowledge" (1996)

Lewis offers a "rule of attention" formulation. S knows p iff S's evidence eliminates every possibility-of-not-p, except those possibilities we are properly ignoring. In ordinary contexts we properly ignore exotic skeptical scenarios; once they are mentioned, we no longer properly ignore them, and the standard for "know" rises to require their elimination. Knowledge "vanishes" when we look at it directly and try to specify what we know, hence "elusive."

Lewis's slogan: "Do not put forward any proposition you do not wish to assert. Once you have asserted that S knows p, every alternative-to-p that is now relevant must be one S can rule out."

Keith DeRose, "Solving the Skeptical Problem" (1995); The Case for Contextualism (2009)

DeRose's contextualism turns on the rule of sensitivity: in any context, one of the contextually relevant standards is whether the subject's belief tracks the truth. When skeptical hypotheses are raised, sensitivity-conditions rise to require tracking across more remote possible worlds (including BIV-worlds), which the subject's belief can't do. So "know" becomes more demanding; the subject may not satisfy it; the skeptic's claim is correct in that context. But in ordinary contexts the sensitivity standards are lower; the subject does satisfy them; the ordinary attribution is correct in that context.

Stewart Cohen, "How to Be a Fallibilist" (1988); "Contextualism, Skepticism, and the Structure of Reasons" (1999)

Cohen develops contextualism as a way to be a fallibilist (knowledge does not require certainty) without being trapped by skepticism. The idea: knowledge requires the subject's evidence to make p sufficiently probable, where "sufficiently" varies with context. In ordinary contexts the threshold is low enough that we know things; in philosophical contexts the threshold rises and we don't.

What contextualism is not

Several positions are similar to contextualism but distinct:

Subject-Sensitive Invariantism (SSI) (Hawthorne, Knowledge and Lotteries, 2004; Stanley, Knowledge and Practical Interests, 2005): "know" picks out one relation invariantly, but knowledge depends on the subject's practical stakes (not the attributor's conversational context). Bank Case B looks the same as Bank Case A from the attributor's standpoint, but in Case B the subject has higher stakes, and knowledge requires more evidence in higher-stakes situations. SSI accommodates the same data but locates the variation differently.
Pragmatic Encroachment: a closely related family of views holding that practical stakes encroach on what counts as knowledge. SSI is one form of pragmatic encroachment.
Relativism (MacFarlane, Assessment Sensitivity, 2014): "know" is sensitive to the assessor's context, not the speaker's. This is a distinct semantic thesis with its own technical machinery.
Skeptical invariantism: the claim that the skeptic is right and we never know things. This is consistent with the form of contextualism but rejects the contextualist's preservation of ordinary attributions.

Contextualism and the skeptical argument

Contextualism's anti-skeptical structure:

The skeptical argument:

(P1) I do not know I am not a BIV.
(P2) If I know I have hands, then I know I am not a BIV. [Closure]
(C) Therefore I do not know I have hands.

The contextualist accepts that in the philosophical context where BIV-possibilities are salient, the standards for "know" are high; (P1) is true; (P2) holds; (C) follows. So in that context, "I do not know I have hands" is correct.

But in the ordinary context where BIV-possibilities are not salient, the standards are low; "I know I have hands" is correct, "I know I am not a BIV" is correct (in the ordinary context, this just means: I'm not in a relevant alternative world where I'm a BIV), and the skeptical argument's (P1) is false in this context.

The skeptic and the ordinary speaker are both right in their respective contexts. There is no contradiction; "know" simply picks out different relations.

This is a satisfying solution to many because it preserves both the skeptical-argument's force in its own setting and the genuine knowledge of ordinary life, without forcing a choice between Mooreanism (deny P1) and closure-denial (deny P2).

Strengths

Saves both intuitions, the skeptic's argument feels compelling, and I clearly know I have hands. Contextualism explains why both intuitions are tracking something real.
Explains the data of conversational knowledge-talk, the bank cases, lottery cases, and other variations in knowledge-attributions across conversations are explained.
Consistency with closure, contextualism doesn't have to deny the closure principle (unlike Dretske-Nozick), avoiding that cost.
Conservative about ordinary knowledge, preserves what we ordinarily say, without dogmatism.

Weaknesses

The "speaker error" objection, when I say in ordinary context "I know I have hands" and the skeptic in the philosophy seminar says "we don't really know we have hands," it doesn't feel like we're talking past each other. We feel we are in genuine disagreement. Contextualism predicts no such genuine disagreement, only different contexts. The phenomenology of the dispute is awkward on contextualism.
Semantic blindness, speakers don't seem to be aware of the context-shift; we don't say "ah, you're using 'know' in the philosophy-seminar standard while I was using it in the ordinary standard." Yet contextualism predicts this should be apparent. Critics: Schiffer ("Contextualist Solutions to Scepticism," 1996); Hawthorne (Knowledge and Lotteries, 2004).
Heterogeneity of contextualist proposals, Lewis, DeRose, and Cohen disagree about what makes the standards rise. The lack of a unified contextualist account is uncomfortable.
The lottery problem, even in ordinary contexts, do I "know" my lottery ticket will lose? This presses on contextualism in ways that don't map cleanly to skeptical-argument cases.
Conventional implicature and pragmatic alternatives, some philosophers (Bach, "The Emperor's New Knowledge," 2005) argue the data are better explained as pragmatic effects on assertion, not semantic effects on "know."

Recent variants

Stalnakerian contextualism, context as common ground; raising salience adds propositions to the common ground that constrain what counts as "knowing" in conversation.
Hawthorne / Stanley SSI as a competitor to traditional contextualism.
Pritchard's epistemic angst account (Epistemic Angst, 2015), combines contextualism with anti-luck virtue epistemology to handle two distinct skeptical problems.

Christian-apologetic resonances

Contextualism has interesting interactions with Christian-philosophical questions:

Religious-epistemological standards: Plantinga's distinction in Reformed Epistemology between de facto and de jure objections has a contextualist flavor, the question of whether Christian belief is warranted depends on what standards of warrant are in play. In ordinary epistemic life, properly-basic-belief standards are operative; in the philosophical critique of religious belief, more demanding standards may be brought in. Contextualism could be deployed to defend ordinary religious belief against philosophical-skeptical pressure.
The reasonable-doubt threshold and the resurrection: legal-evidence apologetic arguments (McGrew, Habermas-Licona "minimal facts") implicitly invoke practical-stakes-conditioned standards. The standards for "knowing the resurrection happened" in apologetic argumentation are calibrated to a kind of beyond-reasonable-doubt threshold, not metaphysical certainty.
The "all things considered" challenge: contextualism allows that one's belief can be "knowledge" in everyday context but fall short in maximally-strict contexts. This dovetails with the Reformed-evangelical pastoral observation that doubt and belief can coexist (the man crying "Lord I believe; help thou my unbelief", Mark 9:24), perhaps "I know" in the ordinary context while "I don't know" in the maximally-skeptical one, without internal contradiction.
A contextualist defense of fallibilism in apologetics: classical Christian apologetics has had to navigate between rationalist demands for proof and existentialist suspicion of all rational argument. Contextualism's fallibilism-without-skepticism gives apologetics a tool: the apologetic case need not deliver Cartesian certainty; it can deliver knowledge by ordinary standards, which is enough for rational belief.