Justification and Context
MATJAŽ POTRČ AND VOJKO STRAHOVNIK, LJUBLJANA
The general drive in epistemology is to deliver necessary and sufficient conditions for
knowledge with the use of exceptionless general epistemic principles. There is another way,
however, to approach the phenomenon of knowledge – by particularistic beautiful patterns.
David Lewis in his paper „Elusive Knowledge” provides a nice contextual epistemology. We
also think that contextualism is the right way to go and that the epistemic context plays an
important role in our endeavors to gain knowledge. But, we disagree with Lewis on two points
of his account, namely that we can talk of knowledge without justification and that a set of
exceptionless rules determines relevant alternatives. We retain the overall notion of
knowledge as justified true belief and try to work out a contextualist account of knowledge
within this notion, at the same time pointing to an alternative, particularistic view on
relevance and relevant alternatives. We briefly sketch our proposal building upon the
distinction between the local and global justification and we put forward some suggestions
how this approach tackles skeptical scenarios, the lottery problem and Gettier cases.
1. The definition based account of knowledge
The general drive in epistemology is to deliver necessary and sufficient conditions for
knowledge with the use of exceptionless general epistemic principles. There is another way
however to approach the phenomenon of knowledge – by particularistic beautiful patterns.
For a long time, and in diverse ways, one has tried to deliver a story of knowledge by
definitory means. Here is the definition of knowledge:
Kap =def p & Bap & Jap
where a is for a subject, p is for a given proposition and K, B, J respectively figure for
operators of knowledge, belief and justification. Subject S will know that p in the case p is
true, S happens to believe p and there is a proper justificatory route leading to this belief.
The definition of knowledge was provided by Plato in an informal manner. But it was only
with the explicit definition of knowledge in analytic epistemology, such as it is furnished
above, that the chase for an ultimate assessment of conditions for knowledge was really open.
One has witnessed the spelling out of each imaginable proposal to capture conditions for
knowledge, the chase proceeding over all possible counterexamples to the definition of
knowledge. A case in point is the Gettier counterexample showing the possibility of justified
true belief failing to result in knowledge.
The definition-based approach to knowledge tried to deliver necessary and sufficient
conditions. It also proceeds by the usage of general exceptionless principles for determining
knowledge. This certainly has enriched epistemology and it has provided a wide and rich
range of ways to look at knowledge. But on the other hand, it seems perhaps time to conclude
that all the excellent minds preoccupied to find a definition of knowledge have failed to
deliver because they were operating on the basis of a false presupposition: that there is a
possible account of knowledge by the means of a definition providing necessary and sufficient
conditions on the basis of exceptionless general principles.
Perhaps the nature of knowledge simply resists definitions. Definitions would provide a
generalist pattern, where knowledge would be specified for a whole range of cases. If the
1
lesson of the definition-based account’s failure is taken to heart, then it may finally happen
that there is no generalist pattern around to take care of knowledge.
Some signs point in the direction that we should free ourselves of general patterns. There is
the shift from the naturalized epistemology towards an interest in a priori forms of knowledge
and in intuition. Virtue epistemology also aims at knowledge as something that escapes the
approach by usage of exceptionless generalities.
One proposal is to mellow the pressure of general requirements for justification, say, to
general principles allowing for exceptions or ceteris paribus clauses. One may also embrace
quasi-particularism as the way to account for knowledge. Quasi-particularism still retains
ceteris paribus principles all in believing that thereby it has already taken over the holistic
point of view. But the radical proposal is particularism.
Particularism was elaborated for the area of morality, claiming that one should oppose the
atomist persistence of a feature’s value over a range of cases, and that one should allow for the
change of a feature’s contribution along the rich and intractable pressures of its dynamical
holistic environment. Although moral particularism is elaborated for the moral thought, the
idea of particularism originated in the area of causality. We accordingly propose to extend
particularism to such areas as metaphysics and epistemology. An important device related to
particularism is what we call unique and holistic beautiful patterns, which we oppose to
generalist patterns. There is a generalist opinion allowing for relevance in several areas to be
achieved only upon the basis of generalist patterns. But we believe that relevance comes from
particularist beautiful patterns. This stresses the point that the particular does not need to be
arbitrary and that it is not arbitrary indeed.
If one takes beautiful patterns as one’s departure, then such features as phenomenology
become an important constitutive ingredient of the subjective account of justification or
knowledge. However, this should not point in the direction of limiting our approach to the
first-person perspective which would perhaps lead us away from the contextualist features that
we endorse.(FOOTNOTE 1) Contextualism in epistemology should be compatible with truthtracking. Moreover, beautiful patterns featuring knowledge should not be arbitrary to any
lesser extent than are the moral decisions taken on the basis of considerations in the particular
holistic circumstances. The phenomenology of knowing, so to say, was just sorely
underestimated because of the blind subscription to knowledge as resulting only from
generalist patterns.(FOOTNOTE 2)
2. Elusive knowledge
In his paper “Elusive Knowledge” David Lewis provides a well-taken contextual
epistemology. We also think that contextualism is the right way to go and that the epistemic
context plays an important role in our endeavors to gain knowledge.
David Lewis is well-known as a contextualist. In his paper “Scorekeeping in the Language
Game” he provided a pragmatic contextual variation-based approach, whereby he accounted
for multiple normative pressures determining the position and meaning of terms in a
conversational sequence. Expressions such as “flat” or “tall” may have different truth-values
attached to them depending on the variation in context. Lewis tries to capture the forces that
determine the semantic contributions of elements in a context by a series of general
accommodation rules.
The paper “Elusive Knowledge” is basically an extension of the scorekeeping approach to the
area of epistemology. Lewis appropriates epistemic contextualism and he also formulates a set
of rules that account for conditions of knowledge.
2
As a proponent of common sense, Lewis’ first claim is that we know a lot. But, this claim of
knowledge gets shattered once we try to assess it in the framework of epistemological
investigation. At this point we meet the skeptic. The tension is eventually resolved by taking
into account the difference between strong norms in the context of epistemology and much
weaker norm requirements in the context of ordinary circumstances. Of the two evils of
contextualism and fallibilism we may somehow embrace the latter.
Here is the definition of knowledge provided by Lewis:
“S knows that p iff
(i) S’s evidence eliminates every possibility in which not-P
(ii) – Psst! – except for those possibilities that we are properly ignoring.”(FOOTNOTE
3)
The evidence in (i) refers to the reasons for accepting knowledge, and p refers to the
situations where p does not hold. As for (ii), it sets the scope of possibilities that one can
properly ignore according to the context in which one finds oneself. An example figures
skeptic situations that may be properly ignored in the ongoing activity of the daily life. We
may also properly ignore matrix possibilities because these are worlds far away from our usual
situation.
Part (i) of Lewis’ definition of knowledge aims at the evidence concerning the case in
question and at its elimination. Part (ii) of this same definition aims at what may be relevant
for proper ignoring in a situation. We use both (i) and (ii) in order to confront cases though,
and the relation between (i) and (ii) has its own dynamics that determine what to count as
knowledge.
It is crucial for such an account of knowledge that we, in some way, delimit what is relevant
and should be eliminated and again what is irrelevant and can be properly ignored. This
setting of limits is crucially context-dependent. We think that such contextualism offers an
important insight into the nature of knowledge. Questions pertaining to knowledge could not
be answered irrespective of the context in which they appear.
We also think that the contextualist insight may be pushed a little bit further by the following
question: “Is there any general set of conditions that would be appropriate for an account of
contextuality?”
We suppose that there is the following presupposition at work in definitions of knowledge:
“Knowledge has to be provided by generalist conditions, i.e. features that determine
knowledge are statable by the help of general patterns.” General patterns are those where the
given features retain the same valence over a number of cases.
There is also the possibility of particularist patterns determining knowledge though, as we
have claimed. Knowledge may be dependent upon rich holistic, but relevant circumstances,
similarly as it happens with the moral particularism claims as the basis for moral value
ascriptions.
Before going on with the specification of rules we should remark that Lewis’ definition of
knowledge is revisionist in respect to the standard definition of knowledge in that it does not
contain any condition for justification. Therefore, knowledge comes without justification, at
least in an explicit sense. We argue though, that justification is implicitly contained in the
seven rules of proper ignorance. We think that what is wrong with these rules is that they try
to capture relevance (whatever is relevant to be ignored in order that knowledge is attained),
which may not even be in principle feasible by tractable generalist means.
The bulk of Lewis’ knowledge proposal rests on proper ignoring. He supplies a list of rules
that should be obeyed in order that one would attain knowledge. These rules that are seven in
3
number determine what is relevant for proper ignoring in our search for knowledge (Lewis
1996, 554-560):
The rule of actuality says that whatever is actual cannot be properly ignored in our search for
knowledge.
The rule of belief says that we should not ignore something in the case in which we have good
reasons to believe it. The rule of actuality matches the truth condition in determining
knowledge and the rule of belief somehow matches the belief condition. Together, they offer
true belief conditions for relevant knowledge considerations.
The rule of resemblance says that of two very similar possibilities both should be either
rejected or considered. Lottery ticket situations where a ticket wins or where a ticket looses
resemble each other. And so it is with the Gettier case where somebody looks at the clock that
unbeknownst to him stopped at 5pm, and with the resembling situation in which he would
take a glance at the same clock at 4:40pm.
The rule of reliability is a positive rule determining what to trust, in opposition to the former
rules of actuality, belief and resemblance that determine what is to be ignored. Lewis thinks
that such processes and institutions as perception, memory and reliable testimony are reliable
in ordinary contexts.
The rules of method pertaining to non-deductive inference figure the representativity of
samples and the reliability of the best explanation, i.e. induction and abduction.
The rule of conservatism appreciates the common knowledge of the community to which we
happen to belong.
The rule of attention says that once we become attentive to some alternative, we cannot
properly ignore it anymore. This is a controversial rule for the reason that once we really
become attentive to scepticism, no evidence can be available to demonstrate the truth of the
contrary position.
3. No justification and rules
We disagree with Lewis on two points of his account, namely that:
(a) We can talk of knowledge without justification.
(b) There is a set of exceptionless rules determining relevant alternatives.
“Between the rock of fallibilism and the whirlpool of skepticism, the former represents the
less intrusive madness”, says David Lewis at the beginning of his “Elusive Knowledge”. We
agree with this common sense pull-away from skepticism in respect to our ordinary
knowledge. But we disagree with some features of Lewis’ contextualist solution to problems
in epistemology. The points in question relate to Lewis’ giving up justification as an important
aspect of knowledge(FOOTNOTE4) as well as his list of rules that determine relevant
alternatives for his conception of knowledge. We think that relevance and salience cannot be
captured by general rules.
When Lewis finishes with his description of the “rules of relevance”, he points out something
that we think is very important in his account – degrees of knowledge. The more possibilities
that we have eliminated by our evidence and the less that we have ignored, the better this will
be for our knowledge. A better knowledge gives us more stability while we change the context
(Lewis 1996, 562-563). And, if our ignoring was correct (even if we have done lots of it) our
knowledge will still be real knowledge in some epistemically safe contexts(FOOTNOTE 5).
We think that Lewis gives up justification much too easily and that we can still make sense of
knowledge as justified true belief. The overall argument for this may only become clear once
we expose our proposed account of knowledge and justification. But here are some
preliminary thoughts. Lewis says that justification is neither a sufficient nor a necessary
4
condition for having knowledge. In the lottery problem, one has all the justification that one
may wish, but one still lacks real knowledge. And in the case of getting knowledge through
perception and memory, there is no non-circular argument that would justify our usage of
them. Still we think that Lewis needs some notion of justification, even if the justification
comes back into his account at the second level or at the meta-level. At least one needs some
kind of justification to justify the placement of a borderline between those relevant
alternatives which one has to eliminate and between the non-relevant alternatives which one
can properly ignore.
And as far as the second point of disagreement is concerned, we repeat that notions of
relevance and salience cannot be suitably included into the contextualist approach to
epistemology in the form of general rules. Both notions are essentially context-dependent and
their normativity is particularistic.(FOOTNOTE 6) We do not put them into the general rulesguided patterns, but make use of particular beautiful patterns as the only candidates for
capturing the normative force of the context.
Knowledge and justification. Lewis argues that justification is not the mark of knowledge
because it is neither a sufficient nor a necessary condition. The former is exposed by the
lottery problem, where an epistemic agent has practically more justification for believing that
his ticket will lose as he does for almost every other ordinary belief. But it still intuitively
seems to us that he cannot claim to know that his ticket will loose. Justification as the
necessary condition fails in the case of our reliance upon perception, memory or testimony,
where no non-circular argument can be given for our forming of beliefs and knowledge by
those means.(FOOTNOTE 7) As far as the first part of the argument is concerned, we believe
that we can propose a contextualist solution to the lottery problem which still relies on
justification. Furthermore, we think that Lewis’ own account could not manage without
justification, and that justification is captured by his rules from the third up to the seventh
rule. In our view, justification (which appears in the standard definition of knowledge as
justified true belief) amounts to having good reasons for belief together with the additional
condition that we form a belief because of these good normative reasons.(FOOTNOTE 8) If
both of these conditions are met, we can say that we are being justified in holding some belief
that p.
Rules of relevance. Lewis tries to give some general account about which alternatives are
relevant and consequently, about which possibilities must be eliminated or properly ignored.
These rules do not come in a strict order of priority and conflicts between them are well
possible. Furthermore, as Cohen has shown, it is far from clear exactly how to understand
those rules, especially in the light of the speaker-sensitive/subject-sensitive distinction (Cohen
1998, 294). Cohen also demonstrates that the list of rules provided by Lewis does not lead to a
straight solution for the skeptic, lottery, and Gettier cases. We agree with this criticism. We
think that here we meet the consequences of treating relevance by means of exceptionless
general rules that delimit the scope of relevant alternatives in Lewis’ conception of
knowledge. We observe relevance and salience as altogether contextual, and therefore as basic
for contextualism. A really contextualist epistemology should show full respect to the
normative power of the context in determining which epistemic possibilities are relevant or
how wide our justification should be for claiming that we possess knowledge.
4. Particularist justification
We retain the overall notion of knowledge as justified true belief and try to work out a
contextualist account of knowledge within this notion, at the same time pointing to an
alternative, particularistic view on relevance and relevant alternatives. We briefly sketch our
5
proposal building upon the distinction between local and global justification and we also put
forward some proposals how this approach tackles skeptical scenarios, the lottery problem,
and Gettier cases.
Our proposal. We propose to keep the original conception of knowledge as justified true
belief and to give a real contextual account of it. We make a distinction between local and
global justification. Lewis revealed an important aspect with his eliminating/ignoring partition
of epistemic possibilities. In this light, we propose the following definition.
Local justification: S is locally justified in believing p in context C iff S has good
epistemic reasons for believing p in this context(FOOTNOTE 9) (and in all relevantly
similar contexts).
Global justification: S is globally justified in believing that p iff S has good epistemic
reasons for believing p in all possible contexts.
The additional condition for the case of local justification is learning to practice safe
epistemology and trying to expand justification to similar contexts – for one does not know
the ultimate nature of the epistemic context in which one finds oneself.(FOOTNOTE 10)
Global justification is appropriate only for the strictest context – the context of epistemology.
All knowledge that could not be globally justified is, in Lewis’ terms, elusive because it slips
out of our hands when we shift to the strictest context of epistemology. Local justification is
obviously a matter of degree that should be regulated according to the epistemic threats
residing in our epistemic situations.
Hard problems. Now that we have presented a brief outline of our proposal of contextualist
epistemology we will give some further clues about how this proposal handles a choice of
hard problems in epistemology. Our solutions will not be far removed from Lewis’ proposals,
the difference is just that we use our local/global justification distinction as the basic tool, and
not rules of relevance as Lewis does. Let us tackle in turn the skeptic, the lottery problem, and
Gettier cases.
The skeptic and the closure of knowledge. The traditional argument could be put in the
following way:
(1) S knows that she has hands.
(2) Having hands├─ <BiV>(FOOTNOTE 11).
 (3) S knows that <BiV>.
If we presume that the right response should be that obviously S does not know that she is not
a brain in a vat, then something has to give. But, by contextualization we can retain the
closure of knowledge and therefore we are also able to retain our intuitive judgments that we
do have ordinary knowledge and also our philosophical intuitions that we cannot know that
we are not in a skeptical scenario. We propose the following depiction of this argument in the
light of our proposal:
(1) S knows that she has hands by local justification standards.
(2) Having hands ├─ <BiV>.
 (3a) S knows that <BiV> by local justification standards.
 (3b) S does not know that <BiV> by global justification standards.
THIS SENTENCE IS UNCLEAR By the means of the original argument contextualization
[[argument of contextualisation?? contextualizing the original argument?? … ???]] we solve
6
the closure of knowledge. We also retain ordinary knowledge all in taking the skeptical case
seriously enough to grant it a strong position in the global epistemological context.
The lottery problem. The problem is how to preserve our ordinary knowledge that the
gambling-addicted Bill is condemned to stay poor without also simultaneously willing to
assert our knowledge that Bill's lottery ticket will not win. One is locally justified to make the
first knowledge assertion according to the context of ordinary life where Bill will always stay
poor because of the way he leads his life. But, at the time as we consider the matter with the
lottery ticket-case, there is no local justification available for the assertion of knowledge that
the ticket in Bill's pocket will not win because the whole problem is now appearing at the
global justification level that characterizes epistemology. In this context, the possibility for
Bill's ticket to win is relevant and we are not globally justified to make the knowledge
assertion of Bill’s ticket not winning.
Gettier cases. In this section, we briefly consider Cohen’s criticism of Lewis. Cohen points to
a case where Lewis could not succeed in solving the Gettier problem by his rules of actuality
and resemblance (Cohen 1998, 296-299). Consider the person A who claims on the basis of
his visual perceptual data that there is a sheep on the hill. While A makes his assertion he is
looking at a sheep-shaped white rock upon the hill. A is in the Gettier situation because behind
the rock there really is a sheep. Lewis solves this case by applying the rules of actuality and
resemblance since the person A should not ignore the saliently similar possibility of a sheepshaped rock. Now, we bring in another person B who ascribes knowledge to A so that the
salience of resemblance between A’s seeing a rock and A’s seeing a sheep is not available
anymore from B’s point of view. B is not aware that A is actually in a Gettier case. What
should we say about B’s ascription of knowledge to A? Cohen shows that Lewis does not have
a convincing solution for cases of this kind. Let us now look at how our local/global
justification distinction could be of use here. If we take both persons as operating under local
justification standards, then we can claim that A does not have good reasons for believing that
there is a sheep on the hill (since he did not practice safe epistemology). But on the other
hand, B has lots of good reasons and is justified in ascribing knowledge to A (though
unfortunately he is in a bad epistemic situation, since A does not have knowledge). If both are
observed from the global justification perspective, neither of them has any knowledge in the
discussed case since neither of them is globally justified in his claims.
This was a brief presentation of our proposal. It retains the overall contextualist approach of
Lewis while we also believe that our account of relevance and salience is simpler and more
genuinely contextual.
References
Cohen, S. 1998 “Contextualist Solutions to Epistemological Problems: Scepticism, Gettier
and the Lottery”, Australasian Journal of Philosophy 76, 289-306.
Dancy, J. 2002 “The Particularist’s Progress”, in B.W. Hooker and M. Little (eds.), Moral
Particularism, Oxford: Oxford University Press, 130-156.
Dancy, J. Forthcoming Ethics Without Principles.
Horgan, T. and Henderson, D. Forthcoming “Iceberg Epistemology”.
Horgan, T. and Henderson, D. Forthcoming (a) “Morphological Content and Justified Belief”.
Horgan, T. and Henderson, D. 2001 “Practicing Safe Epistemology”, Philosophical Studies
102, 227-258.
Horgan, T. and Tienson, J. 1996 Connectionism and the Philosophy of Psychology,
Cambridge: MIT.
Lewis, D. 1996 “Elusive Knowledge”, Australasian Journal of Philosophy 74, 549-567.
7
Lewis D. 1979 “Scorekeeping in a Language Game”, Journal of Philosophical Logic 8, 339359.
Potrč, M. Forthcomig “Particularly Lewis”.
Potrč, M. and Strahovnik, V. Forthcoming Beautiful Patterns.
FUSSNOTEN
(1) We are indebted to Timothy Williamson for this remark (at the occasion of Potrč’s
presentation of this paper at Kirchberg 2003).
(2) We think though, that the definition of knowledge should not be rejected. Only the
normative authority should not be kept with it and this normative authority should be deferred
to the particularist beautiful patterns.
(3) Lewis 1996, 554. The two ingredients of definition are numbered by (i) and (ii) so that we
may refer to them later.
(4) We think though, that the question concerning justification should be carefully
approached. One important distinction to be made is the distinction between having a
justification and being justified that Terry Horgan and David Henderson introduce in their
Horgan & Henderson 2000.
(5) For the idea of safe epistemology see Horgan & Henderson 2001, where the improved
reliabilistic account of justification is presented.
(6) We took this from moral particularism, according to which reasons are context-dependent
and function holistically. The same could be said about reasons for having belief and for good
normative reasons related to justification. See Dancy 2000.
(7) Lewis further illustrates the point that sometimes we even do not know how we know
some things (e.g. in the case where we forgot our primary reasons for believing or in the case
of the ability to recognize some visual patterns), but that we could still claim that we possess
knowledge (see Lewis 1996, 551). This is a very wide conception of knowledge so that some
may have reservations to accept such a view.
(8) For the role of this proper causal etiology and the holistic aspects of belief forming
processes, see Horgan & Henderson, forthcoming.
(9) The additional condition to this is that of being justified because of those good reasons that
one has and not just as based upon some arbitrary beliefs. We will presume this further
condition from now on and will not mention it every time as we talk about justification.
(10) We took this lesson from Horgan & Henderson 2001. In this paper, they deal with
reliabilism and they try to improve it in the direction that besides reliability we should
acknowledge the robustness of our cognitive processes for gaining beliefs as an additional
epistemic value. Reliability is always reliability in the agent’s world, but robustness (of
reliability) may be characterized as truth-conductivity in a very wide set of epistemically
relevant possible worlds. According to their view, such an account of reliability and
robustness makes it possible to talk about two kinds of safety: local and general safety. We
accept the lesson of this account, but try to formulate it on a more general level that is not
strictly connected to reliabilistic epistemological theories and such that it is compatible with
other possible epistemological starting points, while still remaining contextual by its very
nature.
Our rendering of local and global justification presents one possible approach of how to
formalize our version of contextualism. Some modified characterizations would thus be
possible that would still encompass the power of context.
(11) <BiV> stands for “being in a brain-in-a-vat scenario”.
8
References
Cohen, S. 1998 “Contextualist Solutions to Epistemological Problems: Scepticism, Gettier
and the Lottery”, Australasian Journal of Philosophy 76, 289-306.
Dancy, J. 2000 “The Particularist’s Progress”, in B.W. Hooker and M. Little (eds.), Moral
Particularism, Oxford: Oxford University Press, 130-156.
Dancy, J. Forthcoming Ethics Without Principles, New York: Oxford University Press.
Horgan, T. and Henderson, D. 2000 “Iceberg Epistemology”, Philosophy and
Phenomenological Research 61, 497-535.
Horgan, T. and Henderson, D. 2001 “Practicing Safe Epistemology”, Philosophical Studies
102, 227-258.
Horgan, T. and Henderson, D. Forthcoming “Morphological Content and Justified Belief”.
Horgan, T. and Tienson, J. 1996 Connectionism and the Philosophy of Psychology,
Cambridge: MIT.
Lewis, D. 1996 “Elusive Knowledge”, Australasian Journal of Philosophy 74, 549-567.
Lewis D. 1979 “Scorekeeping in a Language Game”, Journal of Philosophical Logic 8, 339359.
Potrč, M. Forthcoming “Particularly Lewis”.
Potrč, M. and Strahovnik, V. Forthcoming Beautiful Patterns.
9