Induction
1 The problem of Induction
An argument has premises and a conclusion. The premises are meant to
support
the conclusion. We evaluate arguments by asking two
questions.
(1) Are the premises true? (2) What support do
the premises lend to the conclusion?
1.1 Deductive arguments
Logicians are interested solely in the question of support (the second
question) not the question concerning the truth of premises. In a
deductively valid argument the support which the premises lend to the
conclusion
is complete. That is, it is logically impossible for the premises
to be true and the conclusion false.
Example: All university students are clever,
because
all university students are insane and all insane people are clever.
The conclusion of the argument is stated first: all university
students
are clever. The premises which are meant to support this come
after:
all university students are insane and all insane people are
clever).
The argument is not particularly good, because the premises are almost
certainly false. The argument is not sound. Still, the
truth
of the premises would guarantee the truth of the conclusion, and that
is
all that is required for the argument to be deductively valid. The
essential
feature of a deductive argument is that the information in the
conclusion
never really goes beyond the information which is packed into the
premises,
taken as a whole.
1.2 Inductive arguments
Often we give reasons for our beliefs about the world, but the reasons
we have do not always yield deductively valid
arguments.
In an inductive argument the premises do not deductively entail the
conclusion.
At best they support the conclusion. There are a wide variety of
inductive arguments, but one kind which is very important (and which
have
tended to dominate the debate) is that of arguments in which the
premises
tell us about the past, and the conclusion tells us about the
future.
For example:
Examples: The sun has risen
every day in the past. Therefore, the sun will rise tomorrow.
Bread has always nourished me in the past.
Therefore, bread will nourish me again tomorrow.
In both cases we daily rely on the premises supporting the
conclusion.
But in both cases the premises do not support the conclusion
deductively.
It is logically possible for the premises to be true and the conclusion
false.
2 Our reliance on induction
What is clear is that all of our expectations and beliefs about the
future,
and about those things we have not directly experienced, must be based
on some kind of inductive argument, either directly (past-to-future) or
via some general principles (e.g. Newton's laws of motion) which are
supported
inductively (particular-to-general). All of our knowledge, if we
have any at all, about what has not yet been directly experienced, is
based
on induction. So we face a dilemma: either induction can be
justified or we have no knowledge about what is not directly
experienced.
The second horn of this dilemma would be disastrous for
empiricism.
Most of what we think we know is knowledge about what we are not
currently
experiencing. If genuine knowledge is confined to particular
experienced
facts of the present moment then our knowledge would be very thin
indeed:
certainly nothing to write home (or a study guide) about.
3 Can induction be justified?
David Hume, the famous Scottish sceptic, raised the problem of the
justification
of induction over two hundred years ago. What follows is derived
from Hume's argument against the possibility of a justification of
induction.
What would a justification of induction have to show?
Would
it have to show that inductive arguments always lead from true premises
to true conclusions? No: that would be far too stringent.
Strong
inductive arguments don't purport to lead from truths to truths without
exception. They purport to lead from truths to highly probably
propositions,
and highly probable propositions may be false. Still, highly
probable
propositions must have a high probability of being true. If Tom
keeps
offering us what he tells us are highly probable propositions, and
those
propositions keep on turning out to be false, then we would not take
Tom's
original claim very seriously. A method which is meant to yield
highly
probable propositions ought to be right most of the time (though not
necessarily
all of the time). So a justification of induction should yield
the
conclusion that inductively strong arguments give us truths most of the
time.
Now we can present a dilemma. There are only two sorts
of
argument: deductive and inductive. So any argument for induction
would have to be either deductive or inductive. We will
present
each horn of the dilemma in turn.
3.1 A deductive justification of induction?
What we know are particular facts about the present, and whatever we
still
remember about the past. We know that inductively valid arguments
have been pretty good in the past, yielding truths from truths, most of
the time (in the past). However, deductively valid arguments
contain
no more information in the conclusion than in the premises. So
any
deductively valid argument with premises known to be true must
not
contain any information about the future. In order to justify induction
we have to show that inductively strong arguments yield truths from
truths
most of the time. But most of the time entails most
of the time in the future, and all we have is evidence about how
well
inductive arguments have done in the past. So any
argument
which tried to establish the future reliability of inductive arguments
on the basis of their past performance must be deductively
invalid.
And so any argument for the reliability of induction which is
deductively
valid must have at least one premise which contains information about
the
future, and hence it contains at least one premise that could only be
known
to be true on the basis of induction itself. The argument
would
only be of use if we could know the premisies to be true, but we
need induction to undergird our confidence in the truth of the premises.
Any argument, with premises known to be true, for the future
reliability of induction would have to have the following form, or else
something like it:
Inductive arguments have been
reliable
in the past;
therefore:
Inductive arguments will
continue
to be reliable in the future.
And this argument is clearly not deductively valid. Since the
truth
of the conclusion of an inductive argument is not guaranteed by the
truth
of the premises, for all we know in the future inductive arguments
might
well start leading us consistently into falsehood.
3.2. An inductive justification of induction?
Still, one might argue that such a justification of induction, by
appeal
to induction's reliability in the past, is not completely
unsound.
It is, after all inductively sound. The argument is
inductively
sound, but, of course, to justify induction by appeal to induction is
to
commit a fallacy with which you should by now be familiar: begging the
question. We are trying to justify induction and yet to do so we
are having to rely on induction.
What is wrong with
begging
the question in this way? Consider an analogy. Suppose that
Tom believes a certain religious system, a system expounded in detail
in
some holy book. Tom tells us we ought to subscribe to this
religious
system as well. We ask him to give us some argument for so
believing
the doctrines expounded in the book. Tom replies: "The book can
be
relied on to yield truths". We ask him how he knows this.
He
replies: "Look here, it says on page seventeen of the book that
this
very book contains only truths." Tom may win some converts by
this
argument, but it has to be admitted that the argument is
extraordinarily
weak. We want some independent reason for accepting the contents
of the book. The fact that the book assures us it is true is not
an independent reason. It presupposes the very conclusion for
which
we want an argument.
4 Summary of the dilemma:
Any justification of induction would have to be either (i) deductive or
(ii) inductive. But (i), any argument with known premises and
with
the conclusion that induction will work in the future will be
deductively
invalid. This is because the premises must contain information
only
about the past, whereas the conclusion must contain information about
the
future. And (ii), any inductive argument will beg the question at
issue: it relies on the reliability of induction in order to establish
the reliability of induction.
Exercise: I have presented an argument to show that
induction
is not justifiable: Is deduction any better off?
Can
you construct an argument against the justifiability of deduction which
would parallel the above argument against the justifiability of
induction?
If so, what, if anything, does this show?