Lewis rationalises his contextualist position by talking in terms of quantification. He says that an
idiom of quantification like ‘every’ is normally restricted to some limited domain (506).
That is, all instances
outside the domain…are irrelevant to the truth of what is said.
So when I make a knowledge-claim, only certain possibilities will be inside the domain of my utterances. When I say that
Everyone is Asian.
I may mean that
Everyone is Asian at this disco.
Of course I don’t mean that everyone in the world is Asian. Crossing over to epistemological contexts, certain possibilities will therefore be excluded from my domain. I may exclude the possibility of an evil demon, or the possibility that we are brains in a vat, or that everything everyone says is a lie, or that there is a further possibility of not-P. And so on. To take on board the universal domain, as it were, it to take on board more than is possible. The epistemological enterprise will not even get going if the universal domain of elimination is tackled. As Wittgenstein said
Certain things are not doubted because the game of doubt presupposes that certain things are not doubted.
To doubt everything is effectively to doubt nothing because no tools will be left with which to doubt even a single thing. So there are certain foundations to the very act of doubting. However, unlike traditional foundationalism, these foundations need not be the same for everyone in every context. They need not be self-evident truths, necessary truths, or axiomatic truths, either. All they need be is a support for the very act of doubting. For example, I cannot doubt that Hitler existed if I also doubt that Hitler was a mass killer. I cannot doubt that the universe has existed for more than five minutes if I am studying rock structure. I cannot doubt that there are other people if I am in the process of writing a book on solipsism I wish to have it published. I cannot doubt that water will come out of the tap rather than poison if I want to quickly quench my first. But these contexts are outside the realm of epistemology. Lewis’s important point is that even within epistemology there will be things that we cannot doubt. For example, that some of my arguments are the same today as they were yesterday. I must not doubt the possibility of identity. Also, I cannot think that no one will understand my arguments if I am to actually formulate them in order to convince other people. More importantly, I cannot doubt the validity of inductive inference if I am to assume that valid arguments today will probably be taken as valid arguments tomorrow.
Lewis calls these foundations of doubt “presuppositions”. He talks about them in terms of “proper ignoring”. That is,
we presuppose proposition Q iff we ignore all possibilities in which not-Q (506).
Such presuppositions are needed to get the epistemic ball rolling. If we have no presuppositions, then, of course, the epistemic ball will not get rolling. Presuppositions are the ground on which we doubt or carry out our epistemic tasks. Again, Q will be different for different people or it will be different in different contexts. And even the context of epistemology must have its presuppositions otherwise
epistemology destroys its own subject matter.
It is just not possible to eliminate all possibilities of not-P. Perhaps it is an a priori impossibility that we can’t eliminate all possibilities of not-P. For a start, there will always be possibilities of not-P that we are simply not aware of. Perhaps the elimination of all possibilities of not-P demands a level of omniscience of which human beings are simply not capable. Our inability to eliminate all possibilities of not-P may simply be a time constraint. Given an indefinite, or even infinite, amount of time then we could perhaps eliminate all possibilities of not-P. So we neither have the infinite lives nor the omniscient minds to eliminate all possibilities of not-P.
Of course, if we are told of or become aware of a possibility of not-P, then it is our epistemic duty to consider that possibility. But what of the possibilities of not-P that we are not aware of? The demand for infallible knowledge even outside Lewis’s scheme may be unwarranted or even simply unworkable. What would qualify as infallible knowledge? And why would P qualify as infallible knowledge?
Of course we can now ask: How can knowledge be context-dependent? Either something is an example of knowledge or it isn’t. But if epistemology is a context that makes all (or nearly all) our knowledge-claims turn out false, then perhaps we ought to revise our view of what makes knowledge, knowledge. However, context-dependence does appear to raise the spectre of relativism. After all, isn’t that what relativism emphasises – context-dependence? We are faced with a situation in which P is knowledge in one specific situation but is not knowledge in another situation. If we substitute the term “knowledge” with the term “truth”, we can see the parallel with relativism. In relativism we are faced with a situation in which P is true in a particular situation, but it is false in another situation. Is knowledge, therefore, any less or more problematic than truth? Perhaps metaphysics or semantics is a context that makes all truth-claims come out false. And yet if there are truths that cross all contexts, as well as examples of knowledge that do the same, then have we the right to emphasise context-dependence when it comes to knowledge-claims and truth-claims? Perhaps we should simply make do with those invariant knowledge-claims that Lewis specifies (e.g., the “axiomatic necessary truths”).
Knowledge and Justification: The Lottery Paradox
This scenario displays the violent disjuncture between knowledge and justification. Just think about it. Say that there are a billion tickets. It is still possible, both logically and epistemically, that you could win the jackpot. However, you would be profoundly justified in believing that you would loose. But you would not have knowledge that you will loose because you may, after all, win. But why not say that you will probably loose, rather than that you will loose? This introduction of probability still involves the notion of justification, that is, if it is highly probable that you will loose then you are justified in that belief. However it doesn’t make the absolute claim that you will loose knowledge. Despite that qualification, introducing probability into the equation still does not introduce an example of knowledge.
It is not just a case that justification in this particular example about tickets is not good enough; it is the case that
no justification [would be] good enough.
There is no limit condition where justification suddenly turns into knowledge. Knowledge and justification, in this scenario, and perhaps all scenarios, are torn asunder. In this example, knowledge is not adequately justified opinion or belief. Knowledge and justification are two different kinds.
Instead of thinking that adequate justification will give us knowledge, perhaps we should, in some cases defined by context, simply throw justification overboard. Or, as Lewis puts it
justification is not always necessary (504).
Lewis gives his own examples of knowledge that does not require justification. For example, he asks:
that (non-circular) argument supports our reliance on perception, on memory, and on testimony? (504)
That is, our use of perception in the past justifies our use of perception in the present. Yet, of course, this is a circular argument. We are using past instances of perception to justify our use of perception in the present. But if we hadn’t justified past uses of perception, then present uses wouldn’t be justified either (by intransitivity). It could have been the case, of course, that the past uses of perception were indeed justified, but only by past pasts uses of perception. And if that’s the case, we have a non-infinite regress on our hands. A regress that perhaps goes back to our initial reliance on perceptions which were not, of course, justified. Exactly the same remarks hold for our reliance on memory and on testimony. Past uses of memory have proved to either correctly represent past actualities or to have helped us cope with present actualities and exigencies. But how did we know that those past uses of memory were in fact correct representations of past actualities? Was it because we thought the same about past past uses of memory?
Lewis gives another interesting example of unjustified knowledge (putting it that way makes such knowledge sound odd). He argues that we know things even though “we don’t even know how we know” (504). It could be the case that we don’t actually have “supporting arguments” to justify our knowledge-claim – we have forgotten them. This means that at one time we had supporting arguments for P, but now we’ve forgotten them. P was justified at time t, but it is not justified at time t1 (the present). But if P was justified at time t (that’s if it was adequately justified), it need not be justified again at t1. This may be a question of time constraints or epistemic common sense. If we had to re-justify all our bits of knowledge again and again, then we wouldn’t have time to acquire new bits of knowledge or even function in the world. For example, I firmly believe that Adolph Hitler was the dictator of Germany in the 1930s. Indeed I believe that this belief is in fact a knowledge-claim. However, I haven’t justified this particular belief for some considerable amount of time (perhaps for years). Perhaps other beliefs of mine are dependent upon or related to this belief about Hitler. I have derived other beliefs or bits of knowledge from this initial belief or bit of knowledge. However, if I need to continually re-justify my initial belief or beliefs, then perhaps I wouldn’t get around to justifying the beliefs that are dependent upon or related to or derived from the initial set of beliefs (which were, after all, justified at one time). Perhaps this is either an argument for some kind of foundationalism or some kind of coherentism. It is foundationalist in the loose sense that I rely on certain beliefs not being continually re-justified in order to make way for new beliefs. However, the initial set of beliefs were in fact fully justified at one point, so they will not come under the rubric of self-evident or unjustified foundations.
However, the argument can also be deemed coherentist because new beliefs also depend upon or are related to or are derived from initial sets of beliefs, which may, in turn, depend on new beliefs and other sets of initial or older beliefs. Indeed the initial set of beliefs may depend upon other sets of initial beliefs and also on new beliefs just as much as they depend on these older sets of beliefs. This is a clear coherentist picture of the inter-relations between the whole set of beliefs, which itself includes sets of beliefs and individual beliefs. In this whole there are no genuinely foundational beliefs that take the weight of all the non-foundational beliefs above them. It is a coherent system of mutual support without a pyramidical structure.
