PhilSci - Philosophers

What this is

Everyone worth knowing for the exam, in course order, with what each one actually claims. Names that only appear as an example or a footnote are in the short list at the end, or left out. “Mock Q” marks the people the official mock exam asks about directly.

Some names come back in more than one week, so do not file them under one topic: Laudan (under-determination in week 3, the pessimistic meta-induction in week 5), van Fraassen (explanation in week 4, realism in week 5), Kitcher (explanation in week 4, reference in week 6), Duhem (under-determination in week 3, against explanation in week 4).

The companion list, by topic, is PhilSci - Topics and Stances.

Week 1: meaning, demarcation, falsification

Carnap and the Vienna Circle (mock Q1)

Logical positivism, later logical empiricism. L01

• Meaning: a statement is meaningful only if it is analytic (logic and maths: true by form, says nothing about the world) or verifiable in principle by observation. “The meaning of a statement is its method of verification.” No synthetic a priori, against Kant.
• Demarcation: unverifiable statements are pseudo-statements, so theories made of them are pseudo-science: literally meaningless, which is worse than false. Metaphysics is the main target.
• How a word gets meaning: its elementary sentence must reduce to observation sentences (“arthropod” = animal, segmented body, jointed legs). “Teavy” has no empirical criterion, so it means nothing.
• Against Heidegger: “The Nothing itself noths” treats “nothing” as a name and a verb. Logically “nothing is P” is $\neg \exists xP(x)$, so the sentence is a type confusion and says nothing. Metaphysics expresses an attitude to life: metaphysicians are “musicians without musical ability”.
• What a theory is: the syntactic conception (Received View): axioms, derived sentences, correspondence rules linking theoretical terms to observational ones. Kinetic theory gives $PV=NkT$.
• Later Carnap: verification softened into graded confirmation, a full semantics, partial interpretation of theoretical terms. Still anti-metaphysics, but it lost its force.
• Hit by: the criterion is hard to state (strict kills universal laws, weak lets metaphysics in), seems self-undermining, and Quine’s two dogmas (week 3).

Karl Popper (mock Q3, plus one line in Q1)

L01b

• Demarcation by falsifiability: a theory is scientific if it can conflict with possible observations. This sorts science from non-science and says nothing about meaning: metaphysics can be meaningful and can grow into science.
• Why falsify: no number of white swans proves “all swans are white” (Hume’s problem of induction), but one black swan refutes it (modus tollens). Science is conjectures and refutations.
• A good theory forbids a lot and makes risky predictions. Eddington’s 1919 eclipse test could have failed, which is what made relativity scientific. Irrefutability is a vice. Corroboration only reports past survival; it does not make a theory probable.
• Scorecard: Einstein scientific; astrology fails (vague prophecies); Marxism was testable, got falsified, then was rescued by re-interpretation (the conventionalist twist); Freud and Adler were never testable (the two men and the child).
• Rescuing a theory: he grants Duhem’s logical point that a failed prediction refutes theory plus auxiliaries. Blaming the measurement is allowed only with independent reasons; otherwise it is ad hoc and lowers scientific status. Einstein’s “sorry for the dear Lord” gives no such reason, so it is dogmatic.
• Empirical basis: which observation reports to accept is settled by intersubjective agreement, in the end a decision.
• Hit by: Lakatos (scientists rightly keep refuted theories), Kuhn, Duhem-Quine.

Imre Lakatos (mock Q2)

L01b

• Against Popper’s “naive falsificationism”: it ignores the tenacity of theories. “Scientists have thick skins.” All theories are born refuted (Newton and the moon’s motion; Kaufmann against special relativity), and crucial experiments are myths written up afterwards.
• Unit of appraisal: the research programme, a series of theories over time:
  • hard core (Newton’s laws plus gravitation), irrefutable by methodological decision;
  • protective belt of auxiliary hypotheses, initial conditions, instrument theories, which takes the hits;
  • negative heuristic: never aim modus tollens at the core;
  • positive heuristic: the plan for developing the belt, which “digests anomalies” (invent an unseen planet: Neptune).
• Progress: a programme is progressive when it predicts novel facts that are then found (Halley’s comet, Einstein’s eclipse) and degenerating when it only accommodates known facts (Marxism’s theory of imperialism, cooked up after workers got richer).
• Revolutions are rational: scientists move to the progressive programme. No instant rationality, loyalty to a struggling programme is allowed, no refutation without a better theory.
• Against Kuhn: reads him as making theory change irrational, “mob psychology”.
• Weak spot: verdicts only come in hindsight; nothing tells a scientist today when to quit.

Week 2: paradigms, revolutions, values

Thomas Kuhn (mock Q4)

L02

• Descriptive turn: asks what scientists actually do, against the logical empiricists’ justification project. Science is not steady accumulation.
• Paradigm: what a community shares: theory, methods, tools, standards, exemplars (model problems), metaphysical assumptions. Newtonian example: laws, differential equations, harmonic oscillator, absolute space and time.
• Normal science is puzzle-solving inside an unquestioned paradigm; a failed puzzle blames the scientist. Stubborn failures become anomalies, then crisis, extraordinary science, revolution. Phlogiston to oxygen; Ptolemy to Copernicus (“planet” changes meaning).
• Revolutions are necessary: only anomalies produce new theories, and a new theory must contradict the old somewhere. Newton is no special case of Einstein: Newtonian mass is conserved, Einsteinian mass converts to energy.
• Paradigm debates are circular: each side argues from its own paradigm’s standards, so argument becomes persuasion.
• Gestalt switch (duck-rabbit), for two reasons: a revolution is a sudden, holistic shift like a conversion; and observation is theory-laden (priming experiment, Bruner and Postman’s playing cards). So there is no neutral observation language.
• Incommensurability: semantic (meanings change) and methodological (standards change). Later softened to local untranslatability.
• Reply to the irrationalism charge: five shared values, accuracy, consistency, broad scope, simplicity, fruitfulness. They are imprecise and can conflict, so choice is rational without being an algorithm, and rational disagreement is good for progress.
• His demarcation criterion: a puzzle-solving tradition. Astrology never had one.

Helen Longino (mock Q5)

L02

• Contextual empiricism: data support a hypothesis only against background assumptions, and those can carry contextual (non-epistemic) values.
• Objectivity is social: no individual is free of background assumptions, so objectivity comes from transformative criticism by a community. Four conditions: recognised avenues for criticism, shared standards, responsiveness to criticism, equality of intellectual authority. It comes in degrees.
• “Social character” of knowledge: knowledge becomes scientific through two or more individuals, by critical emendation and modification. What survives is free of any one person’s subjective preferences.
• Mechanisms: replication, peer review, treatment after publication, absorption into the body of knowledge.
• It is a theory of objectivity and method, not an ethical theory.

Heather Douglas

L02

• Values play a direct role (deciding what to research or fund) and an indirect role: under uncertainty, the cost of being wrong legitimately sets how much evidence is enough. That is inductive risk (the carcinogen case).
• Values never become evidence; they set the bar. Since this happens inside science, it is an internality position. Calls the value-free triptych the “externality model”.
• Contrast with Longino: Douglas’s values set the threshold, Longino’s help decide what counts as evidence at all.

Week 3: under-determination

Pierre Duhem (mock Q6)

L03

• Duhem’s thesis: if premises $P_1\dots P_n$ entail an observation $O$, and experiment gives $O^{\prime }$ incompatible with $O$, then the conjunction of the premises is false. It does not follow that any particular premise is. Falsification is ambiguous. Telescope example: astronomy, optics or initial data?
• No crucial experiment: hypotheses are only tested with auxiliaries, and refuting $H_1$ would prove $H_2$ only if nobody could think of an $H_3$.
• Good sense decides what to drop. Choice is not settled by formal logic, which is different from saying it is arbitrary.
• On explanation (week 4): explaining “gives hostages to fortune”, tying physics to metaphysics, so it is not physics’s job.

W. V. O. Quine

L03, also L01

• “Two Dogmas of Empiricism” (1951) rejects both: reductionism (each statement has its own set of confirming experiences) and the analytic-synthetic distinction (only a difference of degree). At bottom the two dogmas are the same assumption, that statements can be assessed one at a time.
• Web of belief: our whole body of beliefs meets experience only at the edges. Logic and maths sit in the centre, revisable but expensive to change.
• Strengthens Duhem: the bundle under test is everything we believe, logic included. “Any statement can be held true come what may” with drastic enough adjustments elsewhere, and no statement is immune to revision. This is the Duhem-Quine thesis.
• Later (1975), empirical under-determination: two formulations that are empirically equivalent but theoretically inequivalent, where no reconstrual of predicates makes them the same. He calls it an open question whether real cases exist.
• Hit by: Laudan (possible is not rational) and Stein (he dismisses meanings without argument).

Larry Laudan (week 3 side)

L03. His week 5 side is further down.

• Quine argues in an “evaluative vacuum”: showing a theory can be kept is a claim about logical possibility, whereas the threat to methodology would be about rational warrant. Quine never examines the ampliative rules of theory choice.
• Non-uniqueness thesis: at least one rival is as well supported. Harmless. Egalitarian thesis: every theory is as well supported as every rival. Only this would make theory choice arbitrary, and nothing shows it.
• With Leplin: the contrived cases of empirical equivalence all lean on the relativity of motion.

Henri Poincaré

L03

• Geometry is conventional: neither a priori (against Kant) nor a posteriori.
• Sphere world: temperature $T=R^2-r^2$, bodies shrink with temperature. From outside: a finite Euclidean disk. From inside: infinite non-Euclidean space. No experiment decides, because measuring instruments obey whatever physics you assume.
• Geometry faces experience together with physics: Duhem’s point again.

Week 4: explanation and understanding

Carl Hempel (mock Q7)

L04

• Deductive-nomological model: an explanation is a deductive argument; the explanandum follows from general laws plus particular conditions (the explanans). Conditions: logical consequence, a law used essentially, empirical content, true premises. The inductive-statistical version uses probabilistic laws and high probability.
• To explain is to show the phenomenon was to be expected. Explanation and prediction have the same structure.
• Objections: flagpole and shadow (asymmetry, the strongest), hexed salt (irrelevance), barometer and storm (common cause: correlation is not explanation), the ball and the window (no law needed), syphilis (low probability still explains). He bit the bullet on the flagpole.
• On understanding: eliminativist. It is pragmatic in the sense of subjective, so epistemically irrelevant, a topic for psychology.

Wesley Salmon

L04

• Causal-mechanical account: to explain is to show the causes, placing the explanandum in the causal nexus. Solves the flagpole: the pole causes the shadow.
• Costs: needs an account of causation (Hume’s problem of telling it from correlation) and may be too narrow (quantum theory, explaining one law by a more general one).

Philip Kitcher (and Michael Friedman)

L04, L06

• Unification: to explain is to fit phenomena into a broader pattern, deriving many facts with the same argument patterns and reducing the number of brute facts. Handles non-causal explanation. Darwin won people over through explanatory promise.
• Flagpole without causes: heights already follow from one pattern (how objects are made); deriving them from shadows adds a pattern that covers only lit objects, so the best-unified system runs from pole to shadow.
• On reference (week 6): terms do not automatically refer; reference is fixed use by use. “Aether” sometimes refers to the electromagnetic field, sometimes to empty space, sometimes to nothing. Some of Priestley’s uses of “dephlogisticated air” refer to oxygen. Stanford replies that the theories were still radically wrong.

Peter Godfrey-Smith

L04

• Causation and unification are not rivals (Salmon and Kitcher each came round to the other’s view). Pluralism is a step forward, but the real mistake is thinking there is one explanatory relation.
• Contextualism: standards of good explanation partly depend on the field and the period (from Kuhn: Newton’s gravity came to count as explaining). Still not “anything goes”: a standard can rest on a factual error.
• Explanation is internal to science, the opposite of van Fraassen.

Henk de Regt and Dennis Dieks

L04

• Understanding is more than explanation: it is the skill to use a theory to build models that explain phenomena ($T\sim M\sim P$). Scientist S, in context C, understands P on the basis of T.
• Pragmatic does not mean subjective: Hempel and van Fraassen slide between “subjective” and “useable for aims”. The second is inter-subjective, so understanding is epistemically relevant.
• Intelligibility: a context-dependent value scientists place on the qualities of a theory that make it usable (visualisability, for example).
• CUP: P is understood iff there is an explanation of P based on an intelligible theory that is empirically adequate and internally consistent. CIT: a theory is intelligible if scientists can recognise its qualitatively characteristic consequences without exact calculation.
• Against the subjectivity charge: three levels. Macro: understanding is a universal aim. Meso: communities set the standards. Micro: individuals have the skills or not.
• Standards change: Huygens found Newton’s action at a distance unintelligible; in 1926 Schrödinger’s visualisable wave mechanics against Heisenberg’s abstract matrix mechanics. Pauli: new theories become anschaulich once people can use them.

Week 5: scientific realism

Bas van Fraassen (his side of mock Q8)

L04, L05

• Explanation is pragmatic: a three-term relation between theory, fact and context, an answer to a why-question. It is a use of science, “not (pure) science but an application of science”, and adds nothing to what we believe. So explanation is not an epistemic aim of science.
• Constructive empiricism: “Science aims to give us theories which are empirically adequate; and acceptance of a theory involves as belief only that it is empirically adequate.” Empirically adequate means what it says about the observable is true, future observations included.
• Where he stands: semantic realist (read theories literally), epistemic and metaphysical anti-realist, agnostic about unobservables. He does not say electrons do not exist.
• Replies to the realist:
  • to Grover Maxwell: the theory-observation distinction mixes up terms and entities (a category mistake); “observable” is vague but usable, like “bald”, and only belief hangs on it;
  • to IBE: a rival hypothesis fits our practice as well, that we believe the best explanation is empirically adequate;
  • to the demand for explanation: it can never be met, so explanation is not an aim of science (Musgrave answers this);
  • to the no-miracles argument: selection. Only successful theories survive, as in Darwin, so success needs no truth behind it.

Alan Musgrave (mock Q8)

L05

• Van Fraassen “tacitly conflates realism with essentialism, the demand for explanation with the demand for ultimate explanation”.
• Ultimate explanations remove puzzlement. Non-ultimate ones relocate and even enhance it, and are still real explanations.
• So there can always be better or deeper explanations, science does aim to explain, and there is no unlimited demand. Van Fraassen’s attack hits essentialism and misses realism.

Hilary Putnam

L05

• No-miracles argument: “Realism is the only philosophy that does not make the success of science a miracle.” Novel predictions like the Higgs boson and the top quark would be absurd coincidences if the theories were false. It is an inference to the best explanation.
• Also a causal theory of reference (with Kripke, week 6) and an early version of the pessimistic meta-induction (1978).

Grover Maxwell

L05

• The theory-observation distinction cannot be sharply drawn: glasses, window, microscope, telescope form a continuum. Fatal for the positivists, who hung meaning on the line. Van Fraassen escapes it.

Larry Laudan (week 5 side)

L05

• Pessimistic meta-induction against convergent realism: many once-successful theories had central terms that do not refer (aether, phlogiston, caloric, crystalline spheres), so they were not even approximately true. By induction over theories, ours may well be false too.
• Three problems for the realist: no definition of approximate truth; no justified link success, then approximate truth, then reference; successful theories with non-referring terms break the link outright.

Juha Saatsi

L05

• The PMI reads two ways: an induction predicting current theories are false, or a timeless argument that success is not a reliable sign of truth, which undermines the no-miracles argument. The second is more modest and harder to resist.

Week 6: answering the pessimistic meta-induction

Stathis Psillos

L06

• Counts only success with novel predictions, and only mature theories.
• Divide et impera: success confirms the causal core (what really fuels the derivation, typically retained), not the idle constituents (typically dropped). In Maxwell’s electromagnetism the equations are core and the aether as a material medium is idle.
• “Luminiferous ether” plausibly refers to the electromagnetic field, on a mixed theory of reference (causal contact plus a causal description).
• Answer to Stanford: scientists drew the line themselves at the time (Lavoisier, Laplace, Carnot believed the caloric laws and doubted that heat is a fluid). Against the PMI’s analogy: methods improve, and posits like gene, atom and electromagnetic field survived several revolutions.

Kyle Stanford

L06

• Maturity is arbitrary and post hoc: phlogiston and caloric were well confirmed.
• Against Kitcher: “little comfort” if the terms referred while the theories were wrong about virtually everything else.
• Against Psillos, a Pyrrhic victory: the core/idle split is anachronistic (Maxwell held the aether essential); convergence is virtually guaranteed, since whatever survives gets called core; it needs criteria that work forwards, or it is Whig history; and a dilemma: either the split matches what past scientists believed, which we cannot know, or they kept misidentifying what made them succeed. Either way it cannot say which parts of today’s theories to believe.
• Later: “‘Atoms exist’ is probably true” is no comfort; the worry is repeated, profound, unpredictable changes in theory. Also the problem of unconceived alternatives.

John Worrall

L06

• Structural realism: believe the mathematical structure, which is retained (Fresnel’s equations survive into Maxwell); the content about what entities are gets replaced. Objection: structure may not fix an ontology.

Sebastian De Haro

L06. Course coordinator, sets the exam, and this is his own research.

• Diagnosis: the PMI debate assumes a single kind of meaning. Separate intension (sense: what the words describe) from extension (reference: what they pick out). Morning star and evening star: two intensions, one extension, Venus. Thermodynamic and statistical-mechanical temperature: same.
• Slogan: we are justified in being realists about extensions, not necessarily about intensions. An extension is fixed by the intension plus the circumstances, so belief is limited to the domain of application where the theory was tested.
• Successive theories are extensionally equivalent on that domain, shown by predictive (same predictions under approximations), material (same system studied) and conceptual (concepts play the same role) correspondence.
• The aether exists: extensionally it is the electromagnetic field plus a fixed frame of reference, at speeds far below light. Maxwell’s energy-conservation argument for it was very good.
• No Pyrrhic victory: he sorts no parts of theories; all terms scientists claim refer, do refer. Against Stanford: convergence was not guaranteed, and Stanford reads too much into Maxwell.
• Approximate truth: extensional truth on a domain. $T$ is closer to the truth than $T^{\prime }$ iff it is extensionally true on a larger, more significant domain.
• Parallel with van Fraassen: both read theories literally and restrict belief. His line falls between extension and intension, so electrons and fields stay in. Still a realism.
• Also (week 3): dualities in physics are semantic under-determination and do not threaten a cautious realism.

Week 7: laws of nature

David Lewis

L07

• Best Systems Account (Humean): laws are the axioms of the system with the best balance of strength and simplicity in describing the Humean mosaic of local, non-modal matters of fact, using only natural properties. Maxwell’s equations plus the Lorentz force beat “list every fact” and “things exist”.
• The world determines the laws, the reverse of the non-Humean picture. Gold spheres under a mile stay out: too little strength for the cost in simplicity.
• Counterfactuals (with Stalnaker): “if A, B” is true when the closest A-worlds are B-worlds; laws fix the similarity that makes worlds close. Causation as counterfactual dependence.
• Objected to Armstrong: calling $N(F,G)$ “necessitation” does not make it necessitate.
• Problems for the BSA: entanglement does not fit a mosaic of local facts; an empty spacetime governed by the Einstein equations comes out impossible; the patterns stay unexplained.

Eddy Chen

L07. Author of the week 7 reading.

• Minimal primitivism (non-Humean): laws are primitive facts, with no restriction on their form, that govern by constraining the possibilities, as axioms govern mathematics.
• Laws explain by a simple constraint behind a rich pattern (the Mandelbrot world). Simplicity and informativeness are clues for finding laws, where for Humeans they are what makes something a law.
• To “who imposes the laws?”: $2+3=5$ constrains every world and nobody imposes it.
• Also: the Past Hypothesis gives time its direction; strong determinism (with Penrose: only one possible world); more careful than Cinti about chance.

Tim Maudlin

L07

• Laws are primitive and are FLOTEs: they produce later states from earlier ones, with a built-in direction of time ($F=ma$, the Schrödinger equation).
• Objections: most laws are time-symmetric, and some are constraints on whole spacetimes (Einstein field equations), not evolution.

Dretske, Tooley and Armstrong

L07

• Laws are necessitation relations between universals, $N(F,G)$: F-ness necessitates G-ness, so all Fs are Gs. Non-Humean, and reductive about laws.
• Gain: says what laws are. Problems: constraint laws and path integrals do not fit the schema, and the necessitation relation is put in by hand.

Smaller names, one line each

• Kant: synthetic a priori knowledge (Euclidean geometry); we know phenomena, never things-in-themselves.
• Hume: problem of induction; “no necessary connections between distinct entities”; causation as constant conjunction.
• Mach: radical empiricism; theories are economical descriptions of experience (instrumentalism).
• Weber: value relevance (values pick problems) and value freedom (not in the conclusions).
• Stein: Quine rejects meanings (Frege’s sense and reference, Carnap’s intensions) without argument.
• Khalifa, Lipton, Trout: reductivists about understanding. Lipton: “simply more knowledge: knowledge of causes”. Trout: the eureka feeling is cognitive bias.
• Halvorson, Lutz, Frigg: the syntactic and semantic views of theories collapse into each other or complement each other.
• Hesse: “principle of no privilege”: we are as liable to be wrong as past scientists.
• Bird, Demarest: dispositionalism. Laws come from dispositions (fragility), Demarest’s laws summarise causal patterns; Bird admits conservation laws fit badly.

Links

• PhilSci - Topics and Stances
• PhilSci - Overview
• PhilSci - Exam Questions, PhilSci - Flashcards