Date post: | 03-Jan-2016 |
Category: |
Documents |
Upload: | phoebe-jefferson |
View: | 220 times |
Download: | 0 times |
1
MustEvidence
UnderdetermineTheory?
John D. NortonCenter for Philosophy of Science and
Department of History and Philosophy of Science
University of Pittsburgh
2
The Underdetermination
Thesis
3
The Underdetermination Thesis
No body of data or evidence or observation can determine a scientific theory or general hypothesis.
“Underdetermined” means assured possibility of rival theories equally well supported by the evidence.
“Support”means evidence bears on theory by induction or confirmation relations.
This sense essential to science.Assured deductive compatibility of many theories with any body of evidence is trivial.
A truism in science studies and one of philosophy of science’s most successful exports.
The Gap Argument
4
1. No body of data or evidence, no matter how extensive, can determine the content of a scientific theory (underdetermination thesis).
There is universal agreement on the content of mature scientific theories.
Therefore, there is a gap: at least a portion of the agreement cannot be explained by the import of evidence.
My favorite social, cultural, political, ideological or other factor is able to account for what fills the gap.
Therefore, my favorite factor accounts for a portion of the content of our mature scientific theories.
2.
Philosophy of Science Gap
5
"…is at the very best a highly speculative, unsubstantiated conjecture. Even if the thesis can be expressed intelligibly in an interesting form, there are no good reasons for thinking that it is true.”
William Newton-Smith, "Underdetermination of Theory by Data," pp. 532-36 in W. H. Newton-Smith, ed., A Companion to the Philosophy of Science. Blackwell. p.553, emphasis in original.
6
The Underdetermination Thesis is NOT…
(Merely) de facto underdetermination
Known evidence happens to underdetermine some particular theory… and in some cases it can be very hard to procure requisite evidence.
Extreme or contrived cases of theoretical content that transcends the reach of all possible evidence.
(Merely) sporadic underdetermination
Humean underdetermination
No warrant to extrapolate any pattern. Humean skepticism denies the viability of induction itself.
Grue Special case of artificial pearl (below).
7
1 variant adjusted
Justifications
Local Duhemian adjustment of auxiliaries.
HYPOTHESIS and AUXILIARY entails EVIDENCE.
2Global Metaphors of Quinean holism
"The totality of our so-called knowledge or beliefs, from the most casual matters of geography and history to the profoundest laws of atomic physics or even of pure mathematics and logic, is a man-made fabric which impinges on experience only at the edges…”
The resulting underdetermination extends to the “abstract entities of mathematics”!
3Inductive We are to generalize from the display of a few
instances of observationally equivalent theories.
Natural Cultured Artificial
8
No basis in developed theories
of induction
9
1 and 2 depend on
Bare Hypothetico-Deductive Confirmation
If theory T entails evidence Eand theory T’ entails evidence E,then T and T’ are supported equally by E..
1 LocalEVIDENCE
supports• HYPOTHESIS (and AUXILIARY)
• variant HYPOTHESIS (and variant AUXILIARY)
equally.
2 GlobalQuine: "These observable consequences of the hypotheses do not, conversely, imply the hypotheses. Surely there are alternative hypothetical substructures that would surface in the same observable ways."
10
Most developed accounts of induction…
DO NOT admit any simple argument that assures evidence must underdetermine theory.
DO NOT restrict evidence to deductive consequences of hypotheses or theories.
DO NOT equally confirm hypotheses with same observational consequences.
DO accord evidence local import and power to discern between competing hypotheses.
11
Three basic ideasdrive all accounts of inductive inference
Family 1. Inductive Generalization
2. Hypothetical Induction
3. Probabilistic Induction
Principle An instance confirms the generalization
Ability to entail the evidence is a mark of truth
Degrees of belief governed by a calculus
Archetype Enumerative induction
Saving the phenomena in astronomy.
Probabilistic analysis of games of chance
Weakness Limited reach of evidence
Indiscriminate confirmation(the weakness of
the “bare” theory)
Applicable
to non-stochastic systems?
12
Families develop to remedy weaknesses
Inductive Generalization Hypothetical Induction Probabilistic Induction
Enumerative Induction
Hempel's Satisfaction Criterion
Mill's Methods
Glymour's Bootstrap
Demonstrative induction
Exclusionary accounts (error statistics, common cause)
Inference to the simplest
Inference to the best explanation
Reliabilism
Probabilistic accounts (especially Bayesian)
Interval valued beliefs
Non-probabilistic calculi
Demonstrative induction converts:
H saves phenomena E
toE and auxiliaries
ENTAILH.
Controlled studies:
Only treatment can account for difference in test and
control group.
Very fine grained distribution of weight of
evidence.
Limit theorems specify determination of theory.
…and more.
EVIDence refutes (HYPothesis and AUXiliary), but …
13
HYP .045 AUX .045
HYP & AUX .01
EVID .09
P(HYP & AUX) = 0.01
P(HYP) = 0.045
P(AUX) = 0.045
P(HYP & AUX | EVID) = 0 HYP & AUXrefuted
P(HYP | EVID) = 0.5 HYP confirmed
P(AUX | EVID) = 0.5 AUX confirmed
14
Observationally equivalent
theoriesare self-defeating
15
Observationally equivalent theoriesclassified
Natural Newtonian mechanics with different states of rest.
Special relativity and Lorentz ether theories.
Flat and curved spacetime forms of Newtonian gravitation theory.
Matrix and wave mechanics.Standard and Bohmian mechanics.
Cultured Poincaré's disk.
Reichenbach's universal forces.
Continua without reals.
Artificial Deceiving demons.
Recently created worlds with memories, fossil records.
Variant T' of theory T with same observables as T but negation of T's theoretical claims.
mimic natural cases
gratuitous impoverishment
or mutilation
16
They are Self-Defeating
For natural and cultured pairs:
IF the observational equivalence can be demonstrated within a philosophy paper,
THEN we cannot preclude the possibility that they are notational variants of one another, perhaps with some superfluous structure.
I do NOT claim that they MUST be notational variants.
e.g. Lorentz’s ether theory and Einstein’s special relativity. The ether state of rest is dismissed by Einstein as superfluous.
I DO claim the possibility makes them inadmissible as the inductive base for the underdetermination thesis.
Should not ask evidence to separate theories that may not be factually distinct.
17
Conclusion
18
Must evidence underdetermine theory?
Our theories of inductive inference are too rudimentary to support a decision either way on a thesis of such strength.
The underdetermination thesis is a strong claim about the logic of inductive inference.
The thesis survives largely through a mix of wishful thinking and inattention to theories of inductive inference.
19
20
Appendices
21
This Talk
The underdetermination thesis and the arguments for it are reviewed.
The thesis has no basis in developed theories of induction.
Attempts to justify the thesis by displayingobservationally equivalent theories are self-defeating.
Underdetermination by grue fails to add novelty for the same reason.
"Must Evidence Underdetermine Theory?" in The Challenge of the Social and the Pressure of Practice: Science and Values Revisited, M. Carrier, D. Howard and J. Kourany, eds., Pittsburgh: University of Pittsburgh Press, 2008, pp. 17-44.
"The Formal Equivalence of Grue and Green and How It Undoes the New Riddle of Induction." Synthese, (2006) 150: 185-207.
Claims:
22
More on Pairs
23
The Argument
The
observational consequences
must be compactly describable
…else the equivalence
cannot be proved.
When two theories are readily shown to have the same observational consequences:
Compact description must be given in
terms of the
theories’ structures
… else both theories are most likely
superfluous.
The two theories structures must be
easily intertranslatable
… else observational equivalence cannot be
shown.
Anything not carried over in the transformation is
superfluous to the observational
consequences
…hence dismiss as superfluous structure.
Thus they are good candidates
for notational variants of one
theory.
24
Illustration: Toy Ptolemaic and Copernican observational
astronomy.
Image from http://faculty.fullerton.edu/cmcconnell/Planets.html
Observations:the retrograde motion of Mars.
25
Illustration: Toy Ptolemaic and Copernican observational
astronomy.
Animations from http://faculty.fullerton.edu/cmcconnell/Planets.html
Ptolemy’s construction Copernicus’ construction
are inter-translatable by
switching Mars’ deferent (Ptolemy) with
Earth’s orbit (Copernicus)
Which of Earth or Sun is truly at rest is superfluous to saving the astronomical observations.
26
Artificial pairs require a different response
Artificial Deceiving demons.
Recently created worlds with memories, fossil records.
Variant T' of theory T with same observables as T but negation of T's theoretical claims.
Artificial pairs are NOT equally confirmed by the observations.
Observations of the fossil record
better confirman ancient
Earththan
Earth created on Wednesdaywith spurious fossil record.
27
Glymour, Malament, Geroch, Manchak
Observationally indistinguishable spacetimes
This is a hard case that has bothered me a lot!
The equivalence involves only deductive compatibility of many spacetime models to one set of observations.… but Manchak has recently extended the equivalence to some inductive discriminations.
1Underdetermination of general facts about many possible worlds (=theory) by facts in this world.
Underdetermination of facts in this world by other facts in this world.
versus2
Is past light cone determinism better suited to relativity? We cannot observe a complete time slice. Is time slice determinism an artifact of classical physics? Or of a metaphysics of presentism that does not belong in relativity?
NOT this. Only one theory at issue, general relativity. No observationally equivalent rival theories.
THIS. It is a form of indeterminism. Fixing past light cone fails to fix the rest.
28
Underdetermination by
Grue
29
Incompatible hypotheses
confirmsequally
Same evidence if emerald
observed prior to t.
Grue, again
A greenemerald
confirms All emeralds are green.
A grueemerald
confirms All emeralds are grue.
grue = green and examined before t
or blue
30
What is NOT new
Patterns may not be projectible.
Hume (18thc.) That fire has always burned and bread always nourished gives us no assurance they will continue to do so.Jevons (19thc.) Multiple continuation of any numeric sequence. “An apparent law never once failing up to a certain point many then suddenly break down…”
Enumerative induction is
troubled.
Denounced since antiquity
Francis Bacon (17th c.) “The induction which proceeds by simple enumeration is puerile, leads to uncertain conclusions, and is exposed to danger from one contradictory instance, deciding generally from too small a number of facts, and those only the most obvious.”
31
What IS new: Symmetry of grue/green
Grue = observed before t and green or blue.
Bleen = observed before t and blue or green.
Green = observed before t and grue or bleen.
Blue = observed before t and bleen or grue.
…hence cannot dismiss “grue” as derived or contrived.Take “grue” as fundamental and declare “green” derived.
32
The symmetry suggest a more general inductive result
Observationalevidence
Ordinary theory
Grue-ified version of ordinary theory
bears symmetrically in any inductive
logic on
Factually distinct butobservationally equivalent theories equally supported by the evidence...
To avoid this escape
grue-ify our total science. Grue is natural grue-kind.Green in not a natural grue-kind.
… but only if the logic of induction
cannot step outside the symmetry.
Natural kinds only are projectible.Green is a natural kind.Grue is not a natural kind.
33
Isomorphism of ordinary and grue-ified total science
Ordinary science
This emerald is green.All emeralds are green
All emeralds have the same color.Green is a natural kind.
…
Grue-ified total science
This emerald is grue.All emeralds are grue.All emeralds have the same g-color.Green is a natural g-kind.…
The language of ordinary gemology text books.
Gemology text books would convey the same information if every ordinary term were replaced by its grue-fied counterpart.
We would have created a notational variant of ordinary gemology.
That the two are not notational variants is inexpressible in science!!
Ordinary science is the real science.
Grue-ified science is the g-real science.
More: analog in the problem of the nocturnal expansion.