Knowledge and
Lotteries. By JOHN HAWTHORNE. (Oxford:
Clarendon Press, 2004, Pp viii + 205. Price £ 25.)
(to appear in Philosophical Quarterly)
<>
Hawthorne’s book shows that the issues surrounding
skepticism arise in a much greater range of cases than most of us had
thought. In
particular, considerations about lotteries, which are an abstract
representation of ubiquitous features of human life, can do much the
same work
as traditional brain-in-a-vat skeptical hypotheses. He uses this
discovery both
to make vivid the appeal of contextual approaches to knowledge, and
eventually
to conclude that they are unsatisfactory. The discussion of the variety
of
possible contextualist views is clearer than anything else in the
literature. Though
of course I shall have some criticisms of the book, I must say at the
beginning
that this is a must-read, indeed a must-study, for everyone interested
in the
concept of knowledge.
<>
The central discovery is this: although we usually deny that
a person knows that a lottery ticket with a very small chance of
winning will
not be the winning ticket, even when that ticket will in fact win, we
usually
do not deny that the person has knowledge of many facts, her belief in
which is
based on the assumption that the ticket will not win. For example
winning the
lottery may be the only way a poor person could have an expensive
vacation, and
we take ourselves to know that she is not going to be on that cruise
ship, even
though we do not take ourselves to know that that ticket she has
foolishly
spent a much-needed five pounds on is a looser. So this is one way in
which
lotteries are like skeptical scenarios: they present issues about the
closure
of knowledge under entailment, since “she will not be on the cruise
ship” (with
other basic facts) entails “she will not win the lottery”. In fact the
resemblance to skepticism goes deeper than this formal point, since a
large
proportion of what we think we know would not be the case if various
random
processes whose causal structure is much like that of a lottery
produced some
extremely unlikely outcome. I take myself to know that I will be alive
in
fifteen minutes time. But if all the oxygen molecules in the room
migrate in
the next minute to the far end of the room and stay there for fourteen
minutes,
which the laws of physics allow but make extremely improbable, then I
will be
dead at the end of that time. And I do not take myself to know that the
oxygen
molecules will not migrate: after all, we’re dealing with a random
process and
for all I know it could turn out that way. (The paradox here can be
expressed
without putting the weight on the word “know”. It would be natural for
me to
say – at least in some conversational contexts – “I can’t guarantee
that all
the oxygen molecules in this room won’t randomly drift down to the far
end in
the next five minutes”. But it would also be natural to say, almost in
the same
breath, “I’ll have no problem meeting you at 5pm, in fact I guarantee
I’ll be
there”.)
Most of chapter one is devoted to laying out this situation,
very carefully and precisely. In particular Hawthorne considers various
versions of closure of knowledge under implication, including
formulations with
multiple premises, in enough detail to make it unlikely that subtleties
about
patterns of entailment will give an asymmetry between random processes
and
skeptical scenarios. In chapter two he makes the connection with
contextualism.
The connection is very natural. When we talk about a lottery in
isolation we
focus on the smallness of the probability of winning and that makes us,
for
reasons that contextualists have never made really explicit, draw the
threshold
for knowledge above that probability. But when we consider a practical
situation in which the outcome of the lottery is one among many
relevant facts
we switch to considering what strengths of evidence it would be
reasonable for
someone to act on, and this inevitably lowers the thresholds. (Even
though the
probabilities are now usually lower than they were when the lottery
alone was
considered. The probability that I will meet you at 5 pm is less than
the
probability that all the oxygen molecules do not migrate away from me:
my
intention could be derailed by any one of many other very improbable
events.) As
Hawthorne makes clear here and later in the book, there is a crucial
unclarity
in the intuition here. Whose context? Standard contextualism tends
towards the
attributor’s context, so if we are thinking about lotteries we will
tend to say
you don’t know and if we are thinking about whether you are planning on
going
on a cruise we will say that you do know. But when we look for reasons
why
thresholds should vary we usually find factors that depend on the
situation of
the person in question: is this information you should rely on or
commit to
memory or tell others about, given your circumstances? And these
considerations
won’t show that when we change the subject to brains in vats we should
say that
you, thousands of miles away, don’t know where your head is. It is
worth noting
that the same issues arise with the standard contextualist comparison
case of
“flat”: if you and I are talking about the mirror of a reflecting
telescope we
don’t find ourselves denying that Kansas is a flat state. The fact
seems to be
that Kansas is flat enough to disappoint skiers and climbers while not
flat
enough to serve as a mirror; similarly many of your beliefs are well
enough
known to guide many of your actions, while not well enough known to
banish all
doubts; and these facts about Kansas and your epistemic state are not
suspended
by the topic of any conversation in which we might mention Kansas or
you.
In the last two chapters of the book Hawthorne
considers
“invariantist” alternatives to contextualism, in which the thresholds
for
knowledge are independent of conversational context. Skepticism is an
invariantism. For Hawthorne the greatest objection to it is the
consequence
that nearly all our assertions are improper – since to assert something
is to
claim to know it - and that all our actions are inappropriate – since
one
should not act on what one does not know. Hawthorne’s preferred view,
though
the book should not be taken as an argument for this as much as an
assessment
of the costs of many positions, is “sensitive moderate invariantism” in
which
the standards of knowledge are allowed to be sensitive to details of
the
person’s situation (and perhaps the person’s purposes), but are
independent of
conversational context. The advantages of this position lie mostly in
what it
avoids. In particular it avoids strange consequences of the form “It is
not
true that I know that p but in five minutes it will be true”.
Some forms of contextualism can, it seems to me,
avoid such
consequences. For the way we talk about truth often has an adjustment
to the
circumstances of utterance built into it. Suppose that I am now, in
Edmonton,
pointing to the sky, and just about to get on a plane for Sydney. I
say, truly,
“this direction is up”. But relative to tomorrow’s situation that
direction
will be down. But the assertion “it is true that this direction is up,
but
tomorrow it will be false” is still bizarre. Our normal talk of truth
seems to
set some parameters to the moment of utterance even when we are
discussing
things said at other times. (Defeasibly: it would not be bizarre to say
“… but
tomorrow I will be able to say truly of that direction that it is
down”. And it
seems to me at any rate less bizarre to say “…and in five minutes I
will be
able to say truly of this belief that it is knowledge.”)
<>This example also shows how we lurch between ascriber-relativity
and object-relativity. For at the north pole we could discuss a rocket
being
launched at the south pole and say “it is going that direction –
straight down
from here – towards the constellation C” , and we could also say “the
boosters
are now falling away, and are coming down to earth” even though the
direction
they are coming is up from our perspective. Perhaps if we fuse the
semantics of
“up” and the semantics of “flat” we will get something like the
semantics of “know”.
<>
Focusing on the situation of the knower, it is plausible to
think that some of the differences between knowledge and ignorance
concern the
ways information can be a basis for action. I think Hawthorne misses
something
here. He tends to write as if action is normally based on what we take
to be
knowledge. But in fact we often act on things we don’t take ourselves
to know,
because we must act and have no better information. In fact, one
central
function of attributions of knowledge is to signal such cases. It is
clearest
with shared activities. You are asked whether p and you say “Yes - I
think so -
but I don’t know?”, which means “for the purposes of our shared
enterprise we
may have to act on this basis, but it would be better if we could get
some
better information.” (Skeptics should say that we only make such
qualified
assertions, though for simplicity we go along with the conversational
myth.) One
situation in which one is especially prone to label assertions as
non-knowledge
is when much better evidence will soon become available, so “don’t
know” has
the force of “if we can afford to, let’s wait and see”. So, given a
lottery
where the winning ticket will soon be announced, we are somewhat more
inclined
to withhold the K-label because instead of making up our minds now we
can wait.
According to my suspect intuitions we are somewhat less reluctant to
say that
someone does not know that a ticket will not win a lottery when either
the
outcome will never be announced or the agent is acting out of necessity
on the
assumption that a ticket will not win, making the best choice that she
can. (Somewhat,
only: this is not meant to be a solution to the puzzle. But note also
that we
should keep separate our reluctance to believe that a ticket will win
and our
reluctance to call it knowledge when someone does truly so believe.)
The main effect of this book will be to make clear
the
relevance of lottery-like situations to a wide variety of questions in
epistemology. So the number of observations of the form “but wait,
here’s
something else to consider” is enormous. That’s a sign of how
stimulating it
is. Read it.
<>
ADAM MORTON
University of Alberta