Natural Kinds, Naturalistic Epistemology and Philosophical Method Richard Boyd October 2007.

Natural Kinds, Naturalistic Epistemology and

Philosophical MethodRichard Boyd

October 2007

Big Questions

• Goodman, “The New Riddle of Induction” +• Quine, “Natural Kinds” and “Epistemology

Naturalized”• → Renewed interest in philosophical

naturalism• Naturalism? About kind? About

epistemology? About metaphysics?• Philosophical naturalism? • Relationship between philosophy and (other)

sciences?

BIG CONCLUSIONS: METAPHYSICS OF KINDS

• KINDS ARE DISCIPLINE SPECIFIC SOCIAL ARTIFACTS

• “REALITY” OF KINDS ≈ CONTRIBUTION TO ACCOMMODATION OF DISCIPLINARY PRACTICES TO CAUSAL STRUCTURES

• REALISM STILL OK• NEED TO REFORMULATE “MIND

INDEPENDENCE”: SOCIAL PRACTICES ARE “METAPHYSICALLY INNOCENT”

BIG CONCLUSIONS: EPISTEMOLOGY

• NO FOUNDATIONS: RELIABILITY OF METHODS IS RADICALLY CONTINGENT

• NO CONTEXT OF DISCOVERY VS CONTEXT OF CONFIRMATION

• NO INTERNAL VS EXTERNAL FACTORS• NO A PRIORI “CONCEPTUAL ANALYSIS”• NO TRANSPARENCY RE: SOURCES OF SCIENTIFIC

JUDGMENTS• NEEDED: AN EPISTEMOLOGY OF INEXPLICIT

INSIGHT, ERROR RE: ACCOMMODATION• (ALMOST) ALL EPISTEMOLOGY IS (NATURALISTIC)

SOCIAL EPISTEMOLOGY

BIG CONCLUSIONS: SEMANTICS AND REPRESENTATION

• REFERENCE ≈ LANGUAGE MEDIATED ACCOMMODATION

• REFERENCE → SOME ≈LY TRUE CONCEPTIONS, ≈LY RELIABLE METHODS

• →=ANY A PRIORI (EVEN ≈LY) TRUE/RELIABLE CONCEPTIONS, METHODS. INDEED,

• CONCEPTUALLY CENTRAL BELIEFS/INFERENTIAL PRACTICES SOMETIMES IRREDEEMABLY FLAWED!

• TRUTH, REFERENCE ARE EPISTEMIC NOTIONS• NEEDED: SEMANTICS OF INSIGHT/ERROR TO GO

WITH THE EPISTEMOLOGY OF INSIGHT • TRUTH/FALSITY SPECIAL CASES OF

SUCCESSFUL/FAILED ACCOMMODATION

Background Philosophy of Science, I: Projectibility and Confirmation

• To a good first approximation:• At time, t, in research community, C, T is highly

confirmed by evidence, E, ↔

T is projectible by standards current in C at t, AND

E favors T over all relevant alternatives ( = rival theories also projectible @ C, t)

ANDE was gathered so as to control for artifacts

suggested by theories projectible @ C, t.

• WHERE: Projectibility = plausibility given the theories best established @ C, t.

Background Philosophy of Science, II: Projectibility is Evidential

• Projectibility, unprojectibility are evidential @ C, t:

Theory, T, is projectible → already some “indirect” evidence @ C, t for T

T is unprojectible → already some “indirect” evidence @ C, t against T

• Projectibility comes in degrees

T is highly projectible + no highly projectible rival → minimal additional evidence required

T is highly unprojectible → Ignore T altogether

Example: Neutrinos• Puzzle: β-decay. Neutron gives off electron to become proton

E.g,: 3H → 3He + e- (Tritium to helium 3 + an electron)1913: discovered that total energy 3H > energy 3He + energy e-

Energy not conserved??• 1930: Wolfgang Pauli: new particle; 1933: F. Perrin: (almost) massless!

Projectible just because preserves conservation of energy

• Experimental confirmation:1956: Faint light emissions 15 microseconds apart in tank of

CdCl2 solution near nuclear reactor; predicted by theory of neutrinos (Reines and Cowan). Neutrinos detected! Nobel Prize to Reines 1995.

• Role of projectibility: Suppose theory of neutrinos predicts flashing lights but no

crisis re conservation of energyThen, theory of massless particle not projectibleExperiment: no evidence at all for neutrinos!But T highly projectible, so experiment counts as neutrino

detection

Objectivity: Reliability of Theory Dependent Methods

• Basic methodological rule: Carefully choose from among relevant alternatives (= theories recommended by best current theories), controlling for effects suggested by best current theories.

• Why is this rule reliable? In particular, what makes relying on current theories reliable?

• NOT: Their currency• IN FACT: ≈ly their approximate truth• Basic rule reliable ≈ly to the extent that

current theories are relevantly ≈ly true!• So, for any C, t, “Methods in C at t are reliable” is• highly contingent and• anything but a priori• So, there are no foundational inductive methods!

The Naturalistic Epistemology of Objectivity, I: Context of Discovery and Context of Confirmation?

• Rule for objective testing: Test T under circumstances where it’s most likely to go wrong if it’s not (relevantly, approximately) true.

• Rule for identifying such circumstances: Refer to alternative projectible theories.

• When is rule reliable?• Simple answer: @ (C, t,) just when often enough

≈te truth is among projectible alternatives.• Naturalistic answer: @ (C, t), just when, often

enough, ≈te truth is among alternatives actually proposed and publicized as projectible.

• What’s implied?• Appropriately reliable cognitive/conceptual

structures• Appropriate social structures of scientific

communication• Appropriate patterns of funding, encouragement

of research directions,…• → No distinction: contexts of discovery and of confirmation

The Naturalistic Epistemology of Objectivity, II: Characterizing Appropriateness for

Cognitive/Conceptual Structures• How do background theories @ (C, t ) inform projectibility

judgments?Simple answer: Theories explicitly accepted @ (C, t)

provide reasons to identify some alternatives as plausible; others as implausible.

Naturalistic answer: Indirectly: @ (C, t) explicit theories interact with, inform, and are informed by, tacit commitments, conceptual structures, trained intuitions, etc. All these determine judgments of relative plausibility.

• What’s the relevant analogue to approximate truth for tacit commitments, conceptual structures, trained intuitions, etc.?

• Accommodation thesis: The relevant analogue is accommodation of conceptual, linguistic and methodological practices to causal structures.

• Natural kinds provide an example of accommodation.

An Accommodationist Theory of Reference and Natural Kinds

• Key idea (Locke, Goodman, Quine, Putnam, …): Successful inductive/explanatory practice requires accommodation between linguistic, conceptual, taxonomic resources and relevant causal factors.

• Referential hypotheses: Hypotheses about how language-world relations contribute to accommodation in particular disciplinary matrices.

• Natural kinds are social/linguistic artifacts• The kind natural kind is itself a natural kind

in the theory of how accommodation is achieved.

Reference

For terms ti in a matrix M and families, Fi , of properties:

t1 → F1

.

ti → Fi

.

tn → Fn

Satisfying Epistemic Access and Accommodation Conditions

Epistemic Access and Accommodation

• Epistemic access condition: There is a systematic causally sustained tendency for what is predicated of ti within M to be approximately true of things which satisfy Fi, i=1,..,n.

• Accommodation condition: This fact, and the causal powers of things possessing F1,..,Fn, explain how the use of t1,..,tn in M contributes to whatever tendency there is for the accommodation demands of M to be satisfied

• Remark: Accommodation sometimes requires that Fi’s be homeostatic property clusters.

Metaphysical Lesson: The “Reality” of Kinds

• Don’t ask “Is X a natural kind?” • Ask instead “To what extent/how does the

deployment of ‘X’ contribute to accommodation in discipline D?”

• Realism?Kinds are social constructionsThey’re “real” if we are and if they correspond

to relevant causal structures• Mind independence?

Beaver dams are social constructions but they’re real

Real issue: Do minds/social practices make causal reality?

• Realist answer: (2N2C) Human social practices make no non-causal contribution to causal structures. No one here but us animals!

Kinds, Modality and Modal Intuitions • Received view (term t; refers to K in actual world, w; w’ other

possible world)• Actual world uses of t refer in w’ to K• K’s definition in w’ is the same as its definition in w• Intuitions = reliable guide to K’s definition, modal

properties• Problems• For HPC K: definition is a clustering: an actual

causal process• It’s historically and situationally individuated

→members of the cluster vary→ Same variation across possible worlds

• Trans-world individuation: context dependent, depends on trans-world individuation of disciplines → individuation often ambiguous; often breaks down altogether for (even modestly) distant possible worlds

• BUT, intuitions deliver determinate (hence false) judgments

• Philosophically interesting phenomena are HPC: knowledge, rationality, reference, justification, goodness, representation, mentality, personhood,…

Appropriateness for Cognitive/Conceptual Structures: Truth and Reliability as Special

Cases of Accommodation.• Approximate truth = special case of accommodation• ≈ly governed by Tarski-style compositional

semantics• via accommodation accomplished through use of

natural kind terms• Reliability of particular methods ≈ reliability with respect to

approximate truth• Representational function of projectibility judgments, intuitions,

research styles, hunches, etc.:= causal contributions (+/-) to reliability of practices via

alignment with causal structures, where+/- contributions are determined by overall

inferential/practical architecture (cf. natural kinds)• Representational function (≈ computational role) may be

distributed not localized (e.g., projectibility judgments distributed over colleagues, referees)

• Needed: A better representational theory of insight.

The Naturalistic Epistemology of Objectivity, III: Sources of Background Theories, Concepts,…

• Which theories, doctrines, etc. determine projectibility judgments @ (C, t)?

Simple answer: The most fundamental findings in C at t.Naturalistic Answer: Complicated story involving ≥

two important factors.• Plurality of sources

Disciplinary diversityBorrowed expertise (auxiliary hypotheses,

instrumentation,…)Unrefereed common knowledge (or error)

• Each of these → limited individual critical expertise• Social certification of expertise

By universities, journals, academies, etcRelative prestige of research institutions, groups

• Each factor socially orchestrated → No viable internal/external distinction

Naturalistic Epistemology, III, cont’d: Objectivity and Social Embedding

• When are scientific methods objective?Simplistic (but common answer): When presupposition freeSimple naturalistic answer: When background theories,

methods, etc are ≈ly true, reliable, …More sophisticated naturalistic answer: Depends also on the

political economy of science @ (C, t)• When are scientists confident about a finding?

1. Published in major journal2. From respected author or research group3. Results relatively unsurprising

• When do scientists subject a finding to special scrutiny?Failure of one or more of above.

• When are such practices reliable?When 1-3 reliably indicate approximate truth

• When not? Social ideology in science.E.g., literature on race, intelligence, IQ

• When corrected?Science and “external” political struggles.

The Naturalistic Epistemology of Objectivity, IV: “Conceptual Analysis” and Critiques of Scientific

Judgment

• How to understand, criticize the conceptual/inferential structures of science?

• Simple (logical empiricist) answer: Formally rationally reconstruct concepts scientists actually use.

• Simple naturalistic answer: Analyze the concepts scientists actually use as approximations to real definitions (cf. “causal descriptive” theories of reference)

• More complicated naturalistic answer: Acknowledge malignant conceptual meanings = massively mistaken concepts central to scientific communication & practice.

• Conceptual meaning of term q at (C, t) = cognitive/inferential commitments regarding q with which one must engage in order to understand literature, practice.

• Two sorts of engagement: Uncritical: Acceptance. Think: ordinary sophisticated

practitioner Critical: Special stance for historian or philosopher of

science

Malignant Conceptual Meanings: Human Sociobiology

• PREMISE:: Early human behaviors of kind B had evolutionary function F.

• CONCLUSION: The underlying, innate and relatively nonmalleable (perhaps unconscious) motive of contemporary behaviors of kind B is ≈ly to accomplish F.

• Such inferential patterns define the human sociobiological literature, are presupposed in the writing/reading of articles.

• Their inappropriateness follows from principles all their authors acknowledge.

• Daly, M. and M. Wilson. 1997. "Child Abuse and Other Risks of not Living with Both Parents," in L. Betzig ed. 1997. Human Nature: A Critical Reader. New York: Oxford University Press:

• PREMISE: In early humans evolutionary function of child rearing was to enhance survival of own/relatives’ children

• CONCLUSION: “…we should expect parental feeling to vary as a function of the prospective fitness value of the child in question to the parent.” NB: In contemporary societies.

• Malignant meanings!

Metaphilosophical Conclusion: Nonreductionist Philosophical Naturalism

• Quine: Philosophy is empirical and continuous with the empirical sciences

• Simple (REDUCTIONIST) reading: metaphysics ≈physics; epistemology ≈ individual perceptual, cognitive psychology

• Correct (NONREDUCTIONIST!) reading: Philosophy continuous with empirical sciences including social, cognitive and linguistic architecture and political economy of science.

• No distinctly philosophical methods• Philosophical/linguistic/modal intuitions = trained

judgments ≠ sources of a priori knowledge• FINAL CONCLUSION: ALL THIS → PHILOSOPHY IS

NOT JUST CONTINUOUS WITH, BUT AN ESSENTIAL COMPONENT OF, SOUND SCIENTIFIC PRACTICE.

Postscript: Clarifying Foundationalism (An Approximation)

• Let F = some set of “foundational” beliefs.• MF = {methods justifiable either a priori or by reference

to members of F}• Foundationalism in domain D with respect to F =

Whenever p (in D) is known it’s justifiable ultimately from premises in F using methods in MF.

• Foundationalism is more or less radical/modest depending on choice of F

• Modest foundationalism: Take F = {q| q is a priori or q is (true and) commonsensical and could be doubted only as a philosophical exercise}

NOTE: Modest foundationalism for D ≈ Research in D is (pre-D) common sense iterated.

• Theory-dependence of scientific methods → For scientific D, no version of even modest foundationalism is true. Science is not pre-scientific common sense iterated!