Phil 8: Introduction to Philosophy of Science
Outline 2: Problem of Induction
I. Terminology
Godfrey-Smith and I do not use “induction” and related terms in the same
way.
My use: “Inductive argument” refers to any argument that is ampliative.
Godfrey-Smith’s use: “Inductive argument” refers to any argument that
infers a generalization (e.g. All ravens are black) from particular observations
(e.g. raven 1 is black, raven 2 is black, raven 3 is black, etc.).
Both uses are common in philosophy.
I’ll reserve the term “enumerative inductive argument” for what Godfrey-
Smith calls an “inductive argument”.
II. The r value
The r value represents the strength of an inductive argument. It is a real
number between 0 and 1.
The r value is a measure of the support that the premises provide to the
conclusion.
Confidence, like the r value, comes in degrees. Example: I am 10%
confident that it will rain tomorrow, and 99% confident that my dog is
currently in my house.
If a subject is certain of the premises of an inductive argument, then the
inductive argument justifies her having a confidence equal to r in the
argument’s conclusion provided there is nothing else relevant about the
conclusion that she knows.
To solve the problem of induction is to say why the conclusions of inductive
inferences are justified to any degree by their premises.
The problem of induction is NOT that induction is fallible.
III. Hume’s Argument
The argument:
Premise 1: The only way to justify induction would be to give a deductive argument
or an inductive argument.
Premise 2: No deductive argument can do the job.
Premise 3: No inductive argument can do the job.
Conclusion: Induction is not justified.
Premise 1 seems true. All justificatory arguments appear to be either
deductive or inductive.
Premise 2 seems true. Deductive arguments are non-ampliative but inductive
arguments are ampliative. A non-ampliative argument seems unable to justify
the conclusion of an ampliative argument.
Premise 3 seems true. An inductive justification of induction would be
circular.
Could induction be self-justifying?
o Consider counter-induction: inferring that unobserved instances are
not like observed instances.
o If induction is self-justifying, so is counter-induction.
IV. Principle of Uniformity of Nature
PUN: Unobserved instances are like observed instances.
We do believe that there are certain uniformities, or laws, that guarantee that
future occurrences will be like past occurrences.
If we could justify the PUN, then it seems we could justify induction. But
we can’t simply see that the PUN holds.
We can’t justify the PUN with deduction since we can’t deduce facts about
unobserved instances from facts solely about observed instances.
We can’t justify the PUN with induction, since such an argument would
require that we already had a justification for the PUN:
So, we can’t justify the PUN.