1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories...

24
1 Methodology I HD-method (S7&8, App. 7A/B, 8A) • for testing and further evaluation of theories – derive and test (general and individual) test implications (in observation terms) • examples – Einstein-Eddington: GRT light bending – Newell&Simon: physical symbol system hypothesis •a physical symbol system has the necessary and sufficient means for general intelligent action • falsification or confirmation (no verification!)

Transcript of 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories...

Page 1: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

1

Methodology IHD-method (S7&8, App. 7A/B, 8A)

• for testing and further evaluation of theories– derive and test (general and individual) test

implications (in observation terms)• examples

– Einstein-Eddington: GRT light bending – Newell&Simon: physical symbol system hypothesis

• a physical symbol system has the necessary and sufficient means for general intelligent action

• falsification or confirmation (no verification!)

Page 2: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

2

Dogmatic HD-strategies: challenge (App. 7B)

• auxiliary hypotheses (many types, see DN-model)• validity of logico-mathematical argumentation• observation presuppositions • initial test conditions • decision criterion (statistical/approximative)• inductive jump (to general success)

Page 3: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

3

Separate HD-evaluation

• HD-method applied to theory X– derive and test general test implications– per I: ito (in terms of) individual test implications,

• either falsification: by individual counter example of I, and hence of X

• or acceptance of I: general success of X• NB: success: prediction or explanation, minimal. derivable

Page 4: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

4

Evaluation report of X at t

Negative: problems

Asymmetric model:– individual counterexamples

Symmetric models:– general counter-examples

– indicvidual counterexamples

Positive: successes

– general successes

– general successes

– individual successes

Page 5: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

5

Comparative HD-evaluation• Def: Y is at t more successful than X

– no extra counter-examples– previous successes are retained– more successes and/or fewer counter-examples

• Asymmetric (S8.1), or symmetric (S.8.2) – then with (qualitative or quantitative) comparative matrix

• CSH: Comparative Success Hypothesis:– Y remains at least as successful as X– interesting hypothesis,

• even if Y is already falsified!

Page 6: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

6

Rule of Success (Instrumentalist):IRS

• IRS: If CSH has been “sufficiently” tested, choose, for the time being, the more successful theory

• Application of IRS: empirical progress

• CLAIM: IRS, and hence the HD-method, are functional for truth approximation

Page 7: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

7

General Methodological Principles• 1) Falsifiability (= confirmability/testability)• 2) Evaluation (=> evaluation report)

• specifically: aim at likely falsification (= potential strong confirmation!)

• 3) Improvement Principle (IP) (=> empirical progress)– not: elimination-principle (EP) (Popper?)– e.g. by idealization and concretization

• 4) For remaining choices: simplicity, and other aesthetic criteria

Page 8: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

8

Dogmatic behaviorKuhn/Lakatos App. 8A

• Improvement principle (IP)• Programmatic improvement principle (PIP)

– aim at a better theory with the same hard core– if necessary, adapt the hard core– if no other option, look for another program

• (P)IP functional for empirical progress and truth approxination• Types of dogmatic behavior:

– scientific: if with PIP– pseudoscientific: if without PIP

Page 9: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

9

Hiërarchie van epistemologische posities

Q0: onafhankelijke natuurwerkelijkheid? Nee ontologisch idealisme Ja: ontologisch realisme

Q1: ware claims mogelijk? Nee epistemologisch relativisme - ervarings-scepticisme

Ja: epistemologisch realisme - inductief scepticismeQ2: voorbij waarneembaar? Nee observational realisme

- instrumentalisme Ja: wetenschappelijk realisme - constructief empiricisme

Q3: voorbij referentie? Nee referentieel realisme entiteiten realisme

Ja: theorie-realismeQ4: ideale conceptualisering? Nee constructief realisme

Ja: essentialistisch realisme

Page 10: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

10

Vier perspectieven voor theorie-realisme

ModusFocus

actueel nomisch

waarheids-waarde

standaard/traditioneel

Giere

waarheids-benadering

Peirce/Niiniluoto

Popper

Page 11: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

11

Soorten actuele en nomische waarheidsbenadering

• PM: het beste afleidingsinstrument: instrumentalist• observationeel: constructive empiricist• referentieel: referentieel realist• theoretisch: constructief realist• essentialistisch: essentialistisch realist

PM: “de waarheid”: de sterkste ware theorie over een gegeven domein in een gegeven vocabulair

Page 12: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

12

Conclusies ICRvooruitblik: How to approach the truth?

• goede redenen voor overstap:instrumentalist 1 constructief empiricist 2 referentieel realist 3 constructief realist, maar niet voor 4 essentialist

• 1,2 3 tbv lange termijn dynamiek: theorieën als waar accepteren levert nieuwe observatietermen

• instrumentalistische methode efficiënter voor waarheids-benadering dan falsificationistische methode

• hiërarchie van heuristische posities, geen dogma– everything goes sometimes– reculer pour mieux sauter

Page 13: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

13

A probabilistic perspective on the hypothetical method

Theo A.F. Kuipers, Groningen, www.philos.rug.nl/personae/kuipers

• concept explication by I&C I.e. idealization and concretization (to appear A)

• tested by the approximative reduction principle: AR-test, i.e. extreme special case

Page 14: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

14

From d- to p-consequences• Idealization:

H E: E deductive (d-)consequence of H • Concretization:

p(E/H) > p(E): E probabilistic (p-)consequence of H

1 > p(E/H) > p(E): E pp(proper p)-consequence of H

• AR-test: let HE, then p(E/H)=1 > p(E)hence, assuming p(E)<1, a d-consequence is a p-consequence

• To be studied: PCn(A) =def {B/p(B/A)>p(B)}

Page 15: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

15

ComparisonsNot: probable consequence/validity, e.g.• J.L.Cohen: The provable and the probable

– mainly about Baconian vs Pascalian probabilities• F. Jackson: assertability of a conditional (AB) iff p(B/A) high

Perhaps: probabilistic conditional/validity, but not so e.g.• R. Bradley and N. Schwartz: “B follows probably from A” =

– iff most models of A are models of B• E. Adams: “probability conditional” = p(B/A), and “p-validity” =

uncertainty conclusion sum uncertainty premises

To be checked: D. Lewis, R.C. Jeffrey, F.P. Ramsey, R. Stalnaker

Page 16: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

16

From the HD- to the HP-method of testing

Definition E is a d-/p-test implication of H iffE is an ‘observational’ d-/p-consequence of H

• Idealization: Hypothetico-Deductive (HD-)Methodaims at d-confirmation or falsification of d-test

implications

• Concretization: Hypothetico-Probabilistic (HP-)Method aims at p-confirmation or p-disconfirmation of p-test

implications

Page 17: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

17

From d- to p-confirmation: Conclusions Analysis ICR Part I (SiS Ch. 7.1.2)

• There is a coherent landscape of confirmation notions, allowing different languages of (degrees of) confirmation

• Idealization: Deductive confirmation• Concretization: Probabilistic confirmation

– basic definition: p(E/H)>p(E) – instead of standard: p(H/E)>p(H)– ‘p’ may be

Popperian: no inductive meansCarnapian: only inductive likelihoodsBayesian: only inductive priorsHintikkian: both

– which one, no fact of the matter

Page 18: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

18

Challenge

• A coherent I&C explication of deductive and probabilistic methods of testing and of separate and comparative evaluation, taking counterexamples into account

• testing sep. eval. comp. eval.deductive ICR ICR ICR

probabilistic ICR/SiS to be done to be done

Page 19: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

19

HD- and HP-testing and -evaluationConsequences Test results Separate

evaluationresults* &reports

Comparativeevaluationresults &reports

Deductive

H E

d-confirmationorfalsification

positiveornegatived-results

**d-results (not)favoring H2 to H1

(9 log.pos.kinds)

Probabilistic

p(E/H)>p(E)

p-confirmationorp-disconfirmation

positiveornegativep-results

**p-results (not)favoring H2 to H1

(25 lp kinds)

* positive/negative results: à la Laudan’s emp. successes/problems** E favors H2 (relative) to H1 iff p(E/H2) >p(E/H1)(d- and p-reading)

Page 20: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

20

D-/P-Evaluation Matrix (Bx: deductive boxes, ICR/SiS)

+: d-/p-results favorable for H2 relative to H1

H1negatived-result

negativepp-result

neutralp-result

positivepp-result

positived-result

negatived-result

B4 0

B2

B1

negativepp-result 0 neutralp-result

B8

B5 0

B3

positivepp-result 0

H2

positived-result

B9

B7

B6 0

Page 21: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

21

DN-/PN-predictions and -explanations: “H predicts/explains E, assuming C (=A&B&C)” • DN-idealization

H&C E

• PN-concretizationp(E/H&C) > p(E/C)

• AR-test: OK• assuming C, E2 more risky prediction of H than E1, iff

p(E2/C) < p(E1/C) and p(E2/H&C) p(E1/H&C)

• assuming C, H2 explains E better than H1, iff p(E/H2&C) > p(E/H1&C)

Page 22: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

22

Comparative Evaluation and Truth Approximation• ICR-story in terms of positive and negative HD-results:

– Definition “more successfulness” – Comparative Success Hypothesis– Instrumentalist Rule of Success (IRS)– Inference to the Best Theory (IBT, to appear B)– Inductive Jump to the Best Theory (to appear B)

• Extendable to HP-results!?

Page 23: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

23

Conclusion• There is a HP-method as a straightforward

concretization of the HD-method: AR-tests: – all transitions from p- to d-notions: p (E/H)=1/0– from separate evaluation to testing: not yet falsified– from comparative to separate evaluation: one tautology

• Depending on p: Popperian, Carnapian, Bayesian, Hintikkian

• Enabling: testing; separate and comparative evaluation; explanation and prediction

• To be further studied– similar perspectives on truth approximation– PCn(A) =def {B/p(B/A)>p(B)}

Page 24: 1 Methodology I HD-method ( S7&8, App. 7A/B, 8A) for testing and further evaluation of theories –derive and test (general and individual) test implications.

24

ReferencesKuipers, T. (ICR/2000), From Instrumentalism to Constructive Realism, Synthese Library 287, Kluwer AP, Dordrecht,

Kuipers, T. (SiS/2001), Structures in Science, Synthese Library 301, Kluwer AP, Dordrecht,

To appear A: ” Empirical and conceptual idealization and concretization. The case of truth approximation", to appear in Liber Amicorum for Leszek Nowak, homepage

To appear B: ” Inference to the best theory, rather than inference to the best explanation”, to appear in Proceedings ESF-workshop Induction and Deduction in the Sciences, Vienna, 2002, homepage.