PPT

advertisement
‘Who’s afraid of
underdetermination?’ (‘Not I’, said
the Structural Realist)
John Worrall
Philosophy, Logic & Scientific Method
London School of Economics
Part 1: Laudan and Leplin leave
things much as they were
 2 main theses
 1. What they claim: There is no “general
guarantee of the possibility of empirically
equivalent theories” (449)
 1’ What the argument – if successful - would
establish: ‘any finding of empirical equivalence
is both contextual and defeasible’
Part 1: Laudan and Leplin leave
things much as they were
 2. The empirical equivalence of T and T’ does
not entail that the choice between T and T’ is
evidentially underdetermined:
 “ One of a number of empirically equivalent
theories may be uniquely preferable on
evidentially probative grounds” (450)
 AND as a consequence
 “The thesis of underdetermination, at least in
so far as it is founded on presumptions about
the possibility of empirical equivalence for
theories .. stands refuted.” (466)
Part 1: Laudan and Leplin leave
things much as they were
 Argument for thesis 1 (1’) uses 3 premises
 (i) VRO: the ‘variability of the range of the
observable’
 (ii) NAP: the ‘need for auxiliaries in prediction’
 (iii) IAA: the ‘instability of auxiliary
assumptions’
Part 1: Laudan and Leplin leave
things much as they were
 First note an important ambiguity in the term
‘empirically equivalent’ (and in their definition:
“theories are empirically equivalent just in case
they have the same class of empirical, viz
observational, consequences.” (451)
Part 1: Laudan and Leplin leave
things much as they were
 What (i)-(iii) entail is allegedly that:
 T and T’ may be empirically equivalent at one
‘stage of science’ but not at some later stage.
 VRO is a ‘mistake’ – but …
Part 1: Laudan and Leplin leave
things much as they were





Use of auxiliaries just seems sloppy
IF T and T’ are ‘single’ ‘isolated’ theories then
IF Duhem is right
(AND he is!)
THEN T and T’ are automatically empirically
equivalent since both have set of empirical
consequences = Ø
Part 1: Laudan and Leplin leave
things much as they were
 IF we include all the necessary specific,
auxiliary, instrumental assumptions to create
two theoretical frameworks F and F’ (built
around the ‘core’ T and T’)
 THEN problem simply re-arises with F and F’
 Of course this tie may be broken later – by
addition of a further auxiliary to one or other,
but the other can then catch up.
 Exactly the Duhem way into
underdetermination
Part 1: Laudan and Leplin leave
things much as they were
So there is a general proof of empirical
equivalence
 Given any finite set of data D and any rival
core theories T and T’ we can always engineer
the auxiliaries so that both F and F’ entail D.

Part 1: Laudan and Leplin leave
things much as they were
 So in fact everything rests on thesis 2- viz
 2. The empirical equivalence of T and T’ does
not entail that the choice between T and T’ is
evidentially underdetermined:
 “ One of a number of empirically equivalent
theories may be uniquely preferable on
evidentially probative grounds” (450)
Part 1: Laudan and Leplin leave
things much as they were
 This has to be correct of course
 (as those of us who think that there is a
solution to the ‘Duhem problem’ have long
banged on about)
 Prediction/Accommodation
Part 1: Laudan and Leplin leave
things much as they were
 L & L’s own arguments somewhat obfuscatory
 A. ‘significant evidential support may be
provided for a theory by results that are not
empirical consequences of the theory.’ (460)
 B. ‘true empirical consequences need lend no
evidential support to a theory.’ (ibid)
Part 1: Laudan and Leplin leave
things much as they were
 B just rests on a confusion between ‘no
support’ and ‘not much support’
 A: argument 1 – ‘next instance’
 argument 2 – ‘spread of support’ to rest of a
theory
 BUT only when theory is unified
 My prediction/ accommodation distinction
similarly dependent on this sort of judgment
 So back to where we always were
Part 2: Underdetermination and
Structural Realism
Underdetermination and Structural
Realism
 “But these are merely names of the images
which we substituted for the real objects which
Nature will hide for ever from our eyes. The
true relations between these real objects are
the only reality we can attain ….”
 “If the relations are known to us, what does it
matter if we think it convenient to replace one
image by another’
 In fact anything you said about those ‘images’
would be ‘illusory’/’meaningless’
 (Or at best of some heuristic value)
Underdetermination and Structural
Realism
 Hence (E)SR: although our current theories
may not be true, their basic theoretical terms
may have no real referent – and we cannot,
in principle, know whether or not they
have -, nonetheless it is reasonable to believe
that they are – at least approximately –
structurally correct.
Underdetermination and Structural
Realism
 That is, the structure of our successful theories
reflects to some extent the real structure of the
universe.
 And it’s reasonable to believe this, despite
scientific “revolutions”, precisely because
structure is preserved (or quasi-preserved)
through “revolutionary” change.
 And of course because each successive theory
has enjoyed striking and independent
predictive success (NMA)
Underdetermination and Structural
Realism
 SR is clearly committed to the view that the
full cognitive content of any theory T is
captured by the Ramsey sentence of T
 And a good job too!
Why the only sensible view is the
‘Ramsey view’
 First, what is the RT of any given T?
 Ramsey’s whole approach – just like the whole
underdetermination issue - is based on making
a distinction between theoretical and
observational terms.
Why the only sensible view is the
‘Ramsey view’
 Any scientific theory S will therefore involve a
number of theoretical predicates T1,…, Tn and a
number of observational predicates O1,…,Om
 So we can write S as S (T1,…,Tn; O1,…,Om)
Why the only sensible view is the
‘Ramsey view’
 Then the Ramsey sentence RT of T is formed
by
 (i) replacing all occurrences of any theoretical
predicate Ti by a variable Θi to form
 S (Θ1,…,Θn; O1,…,Om); and
 (ii) existentially quantifying on all those
variables to form:
 VΘ1, …,VΘn S (Θ1,…,Θn; O1,…,Om)
Why the only sensible view is the
‘Ramsey view’
 So why is it obvious that the full cognitive
content of any theory T is captured by its RT?
 First and foremost: everybody agrees (though
then forgets it) that standard referential
semantics is entirely misleading
 At any rate when it comes to theoretical ‘terms’
in science
 You – obviously – can’t stand outside of all
theory and compare what the theory says with
an independently given reality
Why the only sensible view is the
‘Ramsey view’
 ‘Global descriptivism’ (at least so far as
theoretical terms are concerned)
 Causal referential talk would be ‘just more
theory’
 Unless you appeal to something mystical
(clearly unobservable but nonetheless
somehow apprehendable ‘semantic glue’)
Why the only sensible view is the
‘Ramsey view’
 If asked ‘what does ‘gluon’ mean’? I can only
rehearse the latest theories involving gluons
 N.B. All of those theories
 (Lots of mistakes made by leaving some theory
‘outside’ (i.e. unRamseyfied))
 That is, a gluon, so far as we can tell, is the
whatever it is that does this that and the other
(as stated by those theories)
 But then if you have accepted this, you have
accepted the ‘Ramsey view’
Ramseyfication and
Underdetermination
 Lots of ways of convincing yourself there is a
problem with underdetermination that are
immediately obviated once you accept Ramsey.
 For example:
 1. The ‘natural’ argument
 2. Tacking paradox
 3. Schmelectron theory
 4. As if
Ramseyfication and
Underdetermination
 1. The ‘natural’ argument:
 Take any theory T and take the set O of the
theory’s consequences that are expressible
entirely in the observation language
 Take any conservative extension of O back into
the full language
 There are indefinitely many of those
 BUT R(T) is itself expressible in the observation
language
Ramseyfication and
Underdetermination
 If we restrict initially to the data set, then
things are different –
 But that takes us back to Duhem and unity
Ramseyfication and
Underdetermination
 2. The Tacking paradox
 T’: GTR & God exists has the same set of
observational consequences as T: GTR
 Moreover so does T’’: GTR & ¬ God exists
 T’ and T’’ are mutually inconsistent
 You want to say that T’ and T’’ are only
nominally different – they have no ‘cognitive
content’ not already held by T
Ramseyfication and
Underdetermination




R(T’) = R(T) & a tautology* = R(T)
R(T’’) = R(T) & a tautology** = R(T)
* VΦVx (Φ x) (take Φ = x=x)
** V Φ Λx(¬ Φ x) (take Φ = ¬(x=x))
Ramseyfication and
Underdetermination
 Take T:current theory of electrons
 T’ says the same except that you consistently
substitute ‘schmelectron’ for ‘electron’
 R(T) = R(T’)
Ramseyfication and
Underdetermination
 4. T: theory of electrons
 T’: Everything is as if electrons (as described
by T) exist
 Always mystified by this – T’ just a
reformulation of T
 Just a reexpression of the Ramsey sentence
 If you add ‘ but actually they don’t exist’?
 Only sense that can be made is surely forwardlooking
 But this theory of the future will be a different
theory with a different Ramsey sentence
Ramseyfication and
Underdetermination
 Summary:
 1. The ‘Duhem problem’ can be solved
satisfactorily – hence the main way into alleged
undetermination is pre-blocked
 (Laudan & Leplin only appear to add something
to this.)
 2. Several of the other ways in which you
might think there is a problem of
underdetermination evaporate once you take
the Ramsey view.
The ‘Ramsey view’ and the
Newman ‘objection’
 Since so many people seem to believe that the
result shows that SR collapses into antirealism, let me start to try to erode that belief:
 1. Neither Ketland nor Putnam nor Newman
has proved that SR entails that any two
‘empirically equivalent’ theories, in the normal
sense, are equally true (or approximately true
or, rather, approximately reflect an external
reality)
The ‘Ramsey view’ and the
Newman ‘objection’
 2. Not just a question of getting the data right.
First the two theories share all the
consequences expressible in the observational
language
 3. And clearly some of these may be
theoretical in anyone’s book – e.g. ‘there are
unobservables’ (i.e objects with no (directly)
observable properties) is (a) purely in the
observation language and (b) clearly
theoretical.
The ‘Ramsey view’ and the
Newman ‘objection’
 4. Moreover, the theories that SR takes to
reflect external reality are not only
(predictively) empirically successful but also
unified.
 (So SR would claim that Copernican theory
reflects reality better than Ptolemaic, despite
their ‘data’ equivalence (not empirical
equivalence), because of the greater unity of
Copernican theory.)
The ‘Ramsey view’ and the
Newman ‘objection’
 5. Surely the RT says that the theoretical terms
exist – it characterises them in observational
terms, or – better - in how they interrelate
with one another and how they structure the
phenomena – but this is patently not to
eliminate them.
 (6. Cp Quine’s famous account of ontological
commitment)
The ‘Ramsey view’ and the
Newman ‘objection’
 (7. David Lewis ‘My proposal could be called an
elimination of theoretical terms, if you must;
for to define them is show how to do without
them. But it is better called a vindication of
theoretical terms; for to define them is to show
that there is no good reason to want to do
without them.’)
Is Structural Realism really
realism?
 Many will be unconvinced: if this is really the
SR that you defend, then for all you say,it isn’t
really realism.
 Danger of quibbling about words – call it antirealism if you like
 (there certainly are ‘realisms’ that it is anti!)
 Doesn’t count on either of Putnam’s definitions
 A. A realist needs to assert that a theory T may
be false even though its RT is true
Is Structural Realism really
realism?
 B. A realist asserts that the theoretical terms in
our current successful theories in mature
science refer (though she can then allow that
what those theories say about the entities to
which they refer is only ‘approximately’ rather
than outright true)
 (In fact SR asserts that the mode of reference
of its terms is just as problematic,
uncharacterisable as the approximate truth of a
theory T.)
Is Structural Realism really
realism?
 So why do I hold that the right judgement is
that SR is a version of realism?
 What is realism?
 1. “Metaphysical” realism: the assertion of the
existence of a reality independent of the
human mind???
 NO: 1’. There exists a structured reality of
which the mind is a part; and, far from
imposing their own order on things, our mental
operations are simply governed by the fixed
laws which describe the workings of Nature.
Is Structural Realism really
realism?
 2. “Methodological realism” = 1’ + the
structure of reality is at least in part intelligible
to the human mind
 3. Scientific realism = 1’ + 2 + successful
theories – the unified theories that explain the
phenomena without ad hoc assumptions – are
‘approximately true’.
Is Structural Realism really
realism?
 IF you are wedded to a correspondence or
semantic view of truth as your account of
corresponding with reality, then SR does not
count.
 But there is no reason why the way in which a
theory mirrors reality should be the usual
term-by-term mapping described by traditional
semantics
Is Structural Realism really
realism?
 SR takes it that the mathematical structure of
a theory may globally reflect reality without
each of its components referring (or being
known to refer) to a separate item of that
reality.
 If you insist on reference for truth (or “near
reference” for approximate truth) then SR is
not realism
 BUT there’s no sustainable version of realism
attainable if you thus insist
Is Structural Realism really
realism?
 SR takes it that the mathematical structure of
a theory may globally reflect reality without
each of its components referring to a separate
item of that reality.
 If you insist on reference for truth (or “near
reference” for approximate truth) then SR is
not realism
 BUT there’s no sustainable version of realism
attainable if you thus insist
 Structural realism is the only viable position
(my confusion, following Poincaré)
Is Structural Realism really
realism?
 Duhem([1906], 28):
 The highest test, therefore, of our holding a
classification as a natural one is to ask it to
indicate in advance things which the future
alone will reveal. And when the experiment is
made and confirms the predictions obtained
from our theory, we feel strengthened in our
conviction that the relations established by our
reason among abstract notions truly
correspond to relations among things.
 That’s all it is! Accept it!
Download