The Scope of Knowledge

The Scope of Knowledge

November 1, 2018 0 By Felso

In the first section we utilised intuitions about when knowledge is had in order to
derive an account of its nature.

This might seem to beg the question against the
sceptic by guaranteeing that our criteria for knowledge are met for the most part.
But we are not in fact assuming that scepticism is false. This is because we allow
that purported cases of knowledge to which we appeal in analysing its nature can
turn out under sceptical attack to be not genuine.

Indeed, sceptics themselves must
adopt the same procedure of analysis – first using ordinary intuitions to derive
criteria – and then give us reasons for doubting that these criteria are really satisfied.
Otherwise, they risk basing their sceptical attacks on an assumed analysis that is
too demanding and so out of touch with our concept of knowledge. In that case we
would not need to take them seriously. Here we will take them seriously by
dismissing all claims that their doubts are necessarily misplaced.

Scepticism challenges us because our beliefs about the properties of real things
transcend the evidence we have for those beliefs. Such evidence consists in the ways
those things appear to us. But objective properties of real objects are what they are
independently of our beliefs about them and the ways they appear to us. Thus, our
beliefs are underdetermined by our evidence. There will be alternative possible
explanations for all the evidence we have. If everything can seem exactly as it does
to us and yet nothing be as we believe it to be, then how can we know that it is as
we believe it to be? If all our evidence is compatible with alternative explanations
of it, then how can we rule out all but one, indeed any, of those explanations? If
knowledge is belief best explained by its truth, then how can we know we have
knowledge when different explanations are compatible with all the evidence we
have for our beliefs? How can we know that the explanatory chains end in the facts
as we take them to be?

Sceptics dramatize this problem by presenting us with alternative scenarios or
sceptical worlds in which everything appears to us as it does now, i.e. our experience
remains exactly the same, and yet nothing in the world is as we take it to be. Descartes
challenged us to show that we are not dreaming all that we currently experience,
or that we are not being deceived by some powerful demon who causes us to have
the experiences we do. Or, to take the contemporary version, suppose that we are
brains in vats programmed by super scientists or computers to have exactly the
experiences we do. We believe this scenario to be possible, since we believe that our experiences are immediately caused by happenings (neuronal firings) in our brains.
How, then, could we know that it is not actual? If we cannot know that it is not
actual, that we are not brains in vats, then it seems we cannot know that we have
bodies surrounded by middle-sized objects with any of the properties we take them
to have. Thus, the sceptic concludes, since his scenario is possible, we do not have
any knowledge of real objects.

A recent trend among epistemologists who battle this sceptic is to grant that we
do not know that we are not brains in vats, but then to argue that we do nevertheless
retain ordinary knowledge of such things as the properties of middle-sized objects.
This response is held to refute the brunt of the sceptic’s argument while
simultaneously showing the source of its plausibility, a goal now endorsed by antisceptical
epistemologists as well. Both claims – that the sceptic’s first premise must
be granted, but his conclusion denied – are suggested by the counterfactual analysis
of knowledge described in the first section. According to this account, we do not
know we are not brains in vats because in the possible world in which we are, we
do not believe we are (since everything appears as now). But this sceptic’s world is
assumed to be very distant from the actual world. It therefore does not affect the
fact that in the closest possible worlds in which particular propositions now believed
about ordinary objects are false, we do not believe them. Hence ordinary knowledge
is retained despite the truth of the sceptic’s premise and resultant plausibility of
his argument.

There are nevertheless three crushing problems with this response to the sceptic.
First, its dependence on the counterfactual account is itself problematic, since we
saw earlier that this account is too strong, ruling out legitimate claims to knowledge.
And, by its own lights, the response relies on the account in just the case in which
it is most dubious, where our evaluation of a knowledge claim takes us to a distant
possible world. Only in this way is the sceptic’s premise endorsed. Second, the
analysis implies that the sceptic’s second, conditional premise (that if we do not
know we are not brains in vats, then we do not know we are surrounded by middlesized
objects) is false, and it clearly seems to be true. If we do not know that we are
not surrounded by a vat’s clear liquid, how can we know that we are surrounded
instead by tables and chairs? Third, and perhaps most important, in denying the
sceptic’s conclusion, the proponent of the counterfactual analysis simply assumes
that the sceptic’s world is a very distant one. But if, as the account admits, we cannot
know that the sceptic’s world is not actual, how could we possibly know that it is
distant from the actual world? As an answer to the sceptic, this response simply
begs the question. Even if accepted, it shows only that knowledge is possible, not
that it is actual.

A yet more contemporary response, contextualism, builds upon the previous one
by agreeing that we do not know that we do not occupy sceptical worlds even though we do retain knowledge in ordinary contexts. Contextualists differ from
counterfactualists in holding that the sceptic’s second premise is true also, as is their
conclusion in the context of their argument (DeRose 1995). In the context of
sceptical doubt, sceptical worlds such as that of the vatted brains become relevant
alternatives that cannot be ruled out. And if they cannot be ruled out, then we do
not retain knowledge of mundane facts with which they compete. But in ordinary
contexts free from sceptical doubts, the sceptic’s distant worlds are irrelevant, and
our beliefs must vary with the presence or absence of the facts to which they refer
only in close possible worlds. When judges of knowledge claims raise sceptical
doubts, they raise the standards for evaluating beliefs; and when beliefs have to be
sensitive to facts in the distant worlds of the sceptic, they cannot pass this unusual
test. Recognition of such varying standards in different contexts of evaluation allows
the contextualist to say that the sceptic’s argument is sound but irrelevant to our
ordinary knowledge claims. How better to show the plausibility of the sceptical
position while defending the ordinary knower?

Despite this attraction, contextualism fares no better in the end than counterfactualism.
It may improve on the latter by allowing that if we cannot know we are
not brains in vats, then we cannot know that we are surrounded by tables and
chairs. Once more, in the context of the doubt that the antecedent expresses, this
conditional is held to be true. But the other problems facing counterfactualism
plague contextualism too, and additional ones as well. First, the position still relies
on evaluating beliefs in what it holds to be distant possible worlds, and we have
seen that this demand is too strong in any context. Some of the most mundane truths
that are easiest to know are false only in very distant possible worlds, where there
is no telling what we would believe. The counterfactual account makes these the
most difficult facts to know. Second, in defending ordinary knowledge the position
once more simply assumes that the sceptic’s worlds are distant while admitting that
we cannot know they are not actual. This does not satisfy the demand to answer
the sceptic by showing that we have knowledge. Third, there is the implausibility
of the claim that we can destroy knowledge we have by simply thinking of sceptical
alternatives. One unwelcome implication of this claim is that philosophers, who
frequently entertain sceptical hypotheses, have so much less knowledge than their
more fortunate, if more naive, counterparts in the real world. While ignorance may
be bliss in some contexts, pursuing a profession that so systematically substitutes
it for knowledge is probably not what young philosophy undergraduates have in
mind. Contextualists who may be content to know so much less than anyone else
nevertheless had better not advertise their position.

Can we then defend knowledge by rejecting the sceptic’s first premise? Can we
claim to know that we are not brains in vats? Can we show that the evidence
we have from experience is evidence for the world as we take it to be, and not for the sceptic’s worlds? How could we know or show this, when experience itself
cannot differentiate between the world as we take it to be and the phenomenal
world of a brain in a vat? One older answer favoured by some epistemologists is
that we know this a priori, that its defence does not require any inductive argument
since it could not be false. It is held that we must know it a priori precisely because
experience in itself cannot distinguish these worlds and so cannot be the source of
this knowledge. Defenders of this tradition give different but related explanations
of how we have this a priori knowledge, of how we can know that the way
something appears, for example, is necessarily evidence for how we take it to be.
Many of the arguments here begin from an account of how we learn to understand
the terms in our language, how we learn to use them correctly or to interpret
their use by others (Hamlyn 1970, ch. 3). If we learn to pick out tables, for example,
by how they appear to us, how they look and feel, then it must be correct that
whatever looks and feels to us continuously in those ways must be tables, or at least
that such looks and feels are necessarily evidence for the presence of tables. In the
language game in which we apply the term ‘table’ to tables, such ways of appearing
are criteria for the correct use of this term. We therefore cannot all be mistaken in
this use based on these experiences any more than we could all be mistaken in the
way we play chess. Tables are whatever we call tables based on correct application
of the term, and correct application is determined by the agreed upon criteria, in
this case certain ways of appearing. Thus, these ways of appearing are necessarily
evidence for tables and for the properties that define them to be tables. We can
neither use the term correctly without accepting these criteria nor interpret its use
by others without typically ascribing true beliefs about tables to them. Likewise, of
course, for other middle-sized objects and their properties.

Is this argument sound? What it really establishes is only the way we must initially
conceive of things. Once we develop the notion of objects whose properties are
independent of our experiences and beliefs, once we develop theories of how these
properties cause our experiences, and once we see that our experiences can mislead
us as to the real properties that cause them, the possibility of wholesale error
becomes intelligible. In fact this possibility is entailed by the notion of independence
that defines the concept of realism about objects and their properties. That real
properties are independent of the ways they appear and the beliefs they cause means
that these appearances and beliefs can be misleading and false. Once we recognise
the possibility of wholesale error on our part, we need not necessarily ascribe mostly
true beliefs to others (although we will ordinarily do so). If, for example, we were
to see some brains in vats and understood their situation, we would not ascribe to
them mostly true beliefs about the objects around them. To interpret the language
of others, we need to explain their utterances, but truth of the beliefs expressed need
not necessarily enter into the majority of these explanations. Nor will we explain the brains’ utterances as true of phenomenal objects instead of false of real ones,
since they will have the same concept of real objects as we do and will intend to
refer to them and their properties. What we and the brains take to be evidence for
the objective properties of real objects cannot dictate what those properties are.
Our shared concept of chess may determine the nature of that game, but this is what
distinguishes games from reality.

Thus, premises about how we learn and interpret our language do not show that
the evidence we have for our beliefs about real objects must necessarily be evidence
for their objective properties as we take them to be. Is this notion of criteria as
necessary evidence short of entailment even coherent? If we do not do away with
real objects and their properties by reducing them to experiences or appearances,
can the latter nevertheless necessarily be evidence for the former? To say that
appearances are necessarily evidence for real properties is to say that they are evidence
in all possible worlds. But in a world of brains in vats in which the brains were
informed or knew of their own situation, their experiences would not be evidence
of objects as we take them to be. This would be true of any sceptical world believed
to be such by its victims. Such worlds are possible. The sceptic’s descriptions of them
do not involve logical contradictions. We could even grant that we could not all be
brains in vats, but that would leave open the possibility that any one of us is and
could possibly be informed of this by our programmers. Thus, there is no necessary
connection between experiential evidence and the real properties of objects.
Do we have any a priori knowledge of reality, as opposed to that which reflects
only definitions of terms, including logical connectives and operators? Is there any
a priori insight into the necessary structure of reality, knowledge of what is real but
not contingent, that needs no inductive confirmation? Well-worn examples that
seem to express such knowledge include the claims that nothing can be red and
green all over and that, if A is taller than B, and B is taller than C, then A is taller
than C (Bonjour 1998, pp. 100–3). It turns out, however, that such examples express
lack of experience or imagination, instead of a priori insight into necessary truth.
When I was much younger and clothing styles were much different, I owned an
iridescent raincoat that looked red and green (as well as tan) all over. Whether the
effect was achieved with discrete red and green threads is irrelevant here, since on
most accounts of colour, whatever looks a certain colour to normal observers in
normal conditions is that colour.

I must admit that I have never seen three people or objects that disconfirm the
example regarding tallness, but imagine the following possible world. Imagine a
world with only three rods and one observer (you). When rod A is visually compared
to B, A appears to be taller; when B is visually paired with C, B appears taller; but
when C is compared to A, C appears taller. When you attempt to place all three in
the same visual field, you cannot take them all in by sight. One always disappears out of the visual field. This world is possible: in fact, it is not so unlike the world
of quantum mechanics, with its indeterminacy and measurement problem. And the
world described seems not to be a world in which the supposed necessary truth
about tallness holds true. Hence it is not a necessary truth (unless we simply stipulate
a definition of ‘tallness’ in which only a transitive relation will bear that term).
Even if we are not so sceptical of necessary truths about reality, the relation
between appearing and being a real property of objects seems not to be necessary. If
we are to answer sceptical challenges to our beliefs about that relation, we therefore
require an inductive argument, broadly construed. We have looked at the implications
for the sceptical argument of the counterfactual analysis of knowledge. It is time to
use the analysis we endorsed, that knowledge is belief best explained (in significant
part) by truth. Here the question becomes whether the ways we are appeared to in
experience are better explained by objects and their properties as we take them to
be than by sceptical hypotheses such as the programming of brains in vats.
It will be objected to this approach right off that we cannot construe our beliefs
about physical objects and their properties as a theory that best explains the patterns
within appearances because we cannot even formulate such patterns without appeal
to physical space and objects. We learn the language of appearing only after learning
to hedge our claims about real properties, which we immediately and naturally
ascribe without inference. More important, it is claimed, we know what experiences
we have had and will have only by reference to locations in physical space among
physical objects. Thus, to even formulate the supposed data we must presuppose
the ‘theory’ that is supposed to explain the data, which ‘theory’ therefore does not
require this explanatory justification.

The former point about learning, however, is again irrelevant to the question of
justification, or demonstrating knowledge. Even if we must conceive of physical
properties first, that does not show that there are such properties or that they are as
we take them to be. Our evidence for our beliefs about them still consists in the ways
they appear, although we don’t initially conceive it that way. The latter point about
formulating the data is first of all debatable. While we cannot translate statements
about physical properties into statements about appearings, it is not so clear that
we could not learn to weaken all claims about objects to claims about appearances,
although doing so would be very cumbersome and awkward. Instead of talking of
seeing an unsupported object fall, we could talk of a visual experience as of an
unsupported object followed by a visual experience as of an object falling. But even
if such reductions would not be universally possible for a skillful language user, this
still only indicates a conceptual necessity, the way we must express ourselves. It still
does not imply that the ways we conceive objects are the ways they are independent
of our conceptions, and it still does not preclude the attempt to show that real
properties as we conceive them best explain the ways they appear to us.

That physical objects and their properties do provide the best explanation we
have for the ways we are appeared to seems easily established. Appeal to real
physical objects first of all explains more deeply than do explanations for particular
appearances in terms of regularities within experience itself. The former can explain
the regularities themselves that are otherwise taken to be ultimate. And explanations
in terms of physical objects and their properties are also superior to sceptical
explanatory hypotheses such as programmes for brains in vats or deceptive Cartesian
demons. The agreement of the different sense modalities on the qualities and
dimensions of objects, for example, strongly suggests a realist explanation, and of
course we also use the realist model successfully to predict future experiences. The
physical realist picture explains both how physical objects interact with each other
and, at least in part, how they cause experiences (by reflecting light, emitting sound
waves, and so on). No such predictions or explanations are forthcoming from the
brain in vat or evil demon hypotheses, which are therefore ad hoc and useless
additions. The only way they have any explanatory power is by being parasitic on
physical object explanations: the demon or programmer of brains must make it
seem as if we are surrounded by interacting physical objects (Vogel 1990). Not only
does this add nothing to the explanations available without these additions; it also
raises natural but unanswerable questions, such as why the programmers or demon
would deceive us in this way.

If the commonsense and scientific explanations for our experience are superior
to the sceptical hypotheses, does this show that we know that we are not brains
in vats? When a hypothesis is put forth only to explain certain data, and when a
superior explanation is later offered, we have some reason to disbelieve the former
hypothesis and the entities it posits solely for explanatory purposes. When the
demon theory of disease was replaced by the germ theory, rational people ceased
to believe in disease-causing demons. In that case, however, there was additional
evidence against the existence of demons – the fact that no one has ever seen one
(except perhaps in a highly irrational frame of mind). The best explanation for the
latter fact is that there are none. In the brain in the vats case, we would not see
ourselves as such if we were, and so we lack that additional evidence. Unlike the
usual case of knowledge of negative existential propositions (propositions that
certain things do not exist), we have no evidence against the existence of the brains
(while also lacking evidence for their existence). It is this, we can claim, that explains
the plausibility of the sceptical argument, while leaving it open to us to defend our
knowledge claims against that argument via inference to the best explanation.

Two problems remain for this inference as an answer to the sceptic. The more
tractable one is that an inference to the best available explanation at a given time
is not necessarily an inference to the best explanation tout court, so that even if
we can accept the latter as producing the true explanation, we cannot so easily accept the former as doing so. We need additional reason to think that the available
explanations exhaust the field of plausible ones. Here that demand seems easily met,
however. There does not seem to be the remotest possibility of an explanation for
our experience being developed that rivals that which appeals to physical objects
and the scientific theories of physical reality.

The far more difficult problem is that an inference to the best explanation defeats
the sceptic only if we can defend the principle that underlies such inference against
sceptical challenge. We must be able to show that what appears to us to be a best
explanation is likely to be true. In general, the most difficult part of any anti-sceptical
epistemology will be to defend the fundamental principles of reasoning or the basic
sources through which we seek knowledge. We have seen that we can defend
perception as a source of knowledge through an inference to the best explanation.
Similar arguments will be available for memory, knowledge of other minds, and
ordinary induction (for example, the best explanation for many coherent memory
impressions will appeal to earlier veridical perception or testimony; the best
explanation for an observed ratio of instances in a class may well be a deeper or
universal regularity, and so on). But this leaves the formidable task of defending the
principle of inference on the basis of which these other sources of knowledge can
be defended.

In indicating the nature of this defence, we must note first that we cannot know
a priori that such fundamental cognitive principles as inference to the best explanation
lead to truth. This is not a necessary truth, since sometimes such inference fails;
and whether it is generally reliable depends on the type of brains we have and on
our relation to our environment. Nor can it be necessary that inference to the best
explanation is likely to be reliable. It is hard to imagine how it could be necessary
that anything is likely (contrast Bonjour 1998, p. 214). In terms of the possible
worlds model of probability, this would mean that in all galaxies or groups of
possible worlds, the number of worlds in which the proposition in question is true
exceeds the number in which it is false. But how could that be necessary and how
could we know that it was: what is to keep the worlds in which the proposition is
false from clustering?

In the absence of an a priori defence of inference to the best explanation and
other fundamental cognitive principles and practices, we would require an inductive
or empirical argument. One promising approach might be to argue that for creatures
with such limited physical capacities and instincts as humans, basic cognitive
capacities would have been naturally selected for their capacity to provide information
or truth necessary to survival. It is plausible that humans would have had
to infer correctly the proximity of predators from tracks or predator noises, for
example, in order to survive, since they could not outrun them or otherwise protect
themselves. There are three major obstacles that a generalised version of such an argument would have to overcome, however. First, there is the question whether,
in the environment in which our brains evolved or were selected for their cognitive
capacities, truth was in general the key determinant of fitness or utility. Second, a
natural next question is whether an affirmative answer to the first one suggests that
inferences to explanations far removed from the environment in which natural
selection took place continue to be truth preserving. And third, there is the problem
of circularity in the argument.

We can only very briefly indicate answers to the first two questions here, since we
will pay more attention to the third, which will introduce the topic of the structure
of knowledge. In regard to the first question, it must be admitted that systematic
distortion can be utile and even fitness enhancing, as when the exaggeration of
colour contrasts enables us to see object boundaries more easily (not to mention
that the perception of colour in itself may be a systematic distortion of objective
reality). But such systematic distortions seem to occur precisely to enable us to
obtain more vital veridical information about the environment, for example about
the locations of various objects. It remains hard to see how creatures like us could
have survived if our basic cognitive capacities did not generate true information
about our environments. We must know the means necessary to our ends, including
survival, as well as the consequences of our actions, and all this is a matter of
inferences as well as perception and memory.

As for inferences far removed from contexts in which survival is at stake, we can
ask generally whether there is reason to suspect that a cognitive capacity that
produces truth in one context will cease doing so in another. More specifically, we
can separate everyday inferences that can be later verified directly in perception
from those which produce the more remote products of scientific theory. The former
inferences are demonstrated to be true if perception is accepted as a source of
knowledge, but the charge of circularity will be raised again, to be addressed shortly.
Scientific inferences not only take explanations to deeper levels, but seek to correct
for distorting subjective inputs into earlier, commonsense explanations. This gives
us more instead of less reason to believe in the truth of such explanations, although
there might be less reason here to believe that the best of available explanations at
a given time is the best overall, hence the true explanation.

The problem of circularity can be pressed at every stage of this suggested defence
of inference to the best explanation as a basic cognitive principle. First, it was
suggested that brains with such cognitive capacities to provide truths necessary to
survival were probably products of natural selection. But theory of natural selection,
indeed appeal to the physical environment, is itself legitimated via inference to the
best explanation. The attempt to legitimate a principle by appeal to the products
of its own use is circular. Second, we noted that many such inferences can be
confirmed by later perceptions, giving direct evidence of their truth-preserving nature. But once more perception itself is certified as a source of knowledge only
via inference to the best explanation for the ways things perceptually appear, again
using the very cognitive principle that perception is supposed to help legitimate.

Addressing this problem of circularity takes us to the question of the structure
of knowledge, to which we now turn.