Underdetermination and Scientific Realism

[N Tasker <pia03nt@sheffield.ac.uk>]

To reply to this message or start a new topic please email: BUPS-DIS@bups.org




Much attention has been given to defending scientific realism from the
underdetermination thesis (UT), while the truth of the thesis itself seems to
have been accepted as obviously true. I believe that UT is often expressed
ambiguously, and that shoring up the thesis in such a way that it poses a
genuine challenge to realism renders it beyond rational defence.

     Scientific realism: the doctrine that there exists a world beyond our
observation which it is the aim of science to correctly describe, and that it
is generally successful in doing so. Moreover, we are capable of discerning
when a theory is correct in describing the unobservable world.

     The underdetermination thesis: for any given scientific theory, t, there
will be an indefinite number of theories which are incompatible with t, and yet
which are equally well supported by the evidence as t. UT is supposed to
threaten realism since we can never be in a position to know that we have
thought of the correct theory.

The problem with UT is that the phrase ?equally well supported by the evidence?
is ambiguous. One popular suggestion is that a theory is supported by observed
instances of its empirical consequences. So, for example, the theory that water
boils at 100 degrees implies that this pot of water should boil when it reaches
100 degrees. When this event is observed, then we have a confirmatory instance
of the theory. The assumption here is that ?positive instance? = ?confirmatory
instance?. Given this assumption, I think UT does pose a challenge to realism.
Unfortunately for UT, this just isn?t how science works.

I assert (a) that not all positive instances of a theory are confirmatory, and
(b) that not all confirmations of a theory come from observations of its
positive instances.

Argument for (a): imagine you and I are sitting by a duck pond. I notice that
all the ducks in the pond are male, and I hypothesise that all ducks in the
world are male. Naturally you ask me to support my hypothesis. I reply that all
positive instances of a theory are confirmatory, and therefore, the very same
ducks which initially prompted my hypothesis count as good evidence for it.

Argument for (b): scientists take into account other virtues when assessing
theories, e.g. coherence with other theories, consilience, predictive power, no
ad hockery.

So just because two theories have the same empirical consequences, that doesn?t
mean that they are equally well supported. If the argument against realism is
to have any force it must assert that there will be an indefinite number of
equally well supported theories, when ?support? is defined in the way I have
gestured towards. But the arguments one tends to encounter for UT entitle it to
a much cruder definition of ?support?, a definition which assumes that all and
only positive instances of a theory are confirmatory. 

Take for example the Kripkensteinian argument about rule following. For any
sequence of numbers there will be many different functions which describe
different ways of continuing the sequence, and which are equally well supported
by the original sequence. However persuasive this reasoning may be in
mathematical examples, this way of justifying UT seems to assume the fallacious
principle of confirmation which I diagnosed above.

Does anyone know a better way of arguing for UT ? one which better approaches
the true nature of confirmation in science?  


Browse or search the BUPS-DIS archives, or unsubscribe from the mailing list at: http://www.bups.org/mailinglist.shtml