Philosophy of Science
Philosophy of science studies the assumptions, methods, structure, and implications of scientific inquiry.
It addresses questions such as: what distinguishes science from non-science?
How are theories confirmed or refuted?
What is a scientific explanation?
Are theoretical entities real?
What are laws of nature?
How do scientific revolutions occur?
Pre-history
The methodological tradition begins in earnest with Francis Bacon and develops through 19th-century empiricism.
Francis Bacon’s Novum Organum (1620) developed the inductive method.
It contains tables of presence, absence, and degree.
The four idols obstructing knowledge are the idols of the tribe, cave, marketplace, and theater.
John Stuart Mill’s A System of Logic (1843) gave five methods of experimental inquiry.
These are agreement, difference, joint method, residues, and concomitant variation.
William Whewell’s The Philosophy of the Inductive Sciences (1840) coined “scientist” (1834).
He defended consilience of inductions as confirmation.
Auguste Comte’s Cours de philosophie positive (1830-1842) developed positivism.
The three stages (theological, metaphysical, positive) of intellectual development frame his philosophy of history.
Ernst Mach’s The Science of Mechanics (1883) emphasized economy of thought.
Concepts must reduce to sensations.
His influence on relativity (Einstein) and logical positivism was substantial.
Pierre Duhem’s The Aim and Structure of Physical Theory (1906) held that physics aims at saving phenomena, not metaphysical truth.
Hypotheses are tested in bundles (Duhem-Quine thesis).
Logical empiricism
The Vienna Circle (1924-1936), the Berlin School (Reichenbach), and allies developed logical empiricism (later logical positivism).
Members and associates included Moritz Schlick (founder), Rudolf Carnap, Otto Neurath, Hans Hahn, Philipp Frank, Friedrich Waismann, and Herbert Feigl.
In Berlin, the school included Hans Reichenbach, Carl Hempel, Walter Dubislav.
Visitors and influences included Wittgenstein, Quine, and Tarski.
Core doctrines
Verificationism holds that a statement is cognitively meaningful only if analytic or empirically verifiable.
Anti-metaphysics holds that metaphysical claims (about substance, the absolute, free will) are meaningless pseudo-statements.
Unity of science holds that all sciences share a method and can be reformulated in physicalist language (Neurath, Carnap).
Logical analysis treats philosophy as the analysis of the logic of science.
Key works
Rudolf Carnap’s Der logische Aufbau der Welt (The Logical Construction of the World, 1928) sought constitution of all concepts from elementary experiences.
Carnap’s “Empiricism, Semantics, and Ontology” (Revue Internationale de Philosophie 1950) distinguished internal vs external questions about linguistic frameworks.
A. J. Ayer’s Language, Truth and Logic (1936) popularized logical positivism in English.
The Vienna Circle dispersed after the rise of Nazism.
Carnap, Reichenbach, Feigl, and Hempel emigrated to the US and shaped American philosophy of science.
Falsificationism
Karl Popper’s Logik der Forschung (1934), translated as The Logic of Scientific Discovery (1959), and Conjectures and Refutations (1963) rejected verificationism.
Science cannot verify universal claims (all swans are white) by finite evidence.
But it can falsify them (one black swan suffices).
Demarcation: scientific theories are those that are falsifiable.
Unfalsifiable theories (Freudian psychology, Marxist history of his time) Popper labeled pseudo-science.
Methodologically, scientists should propose bold conjectures with high empirical content, then attempt to refute them.
Corroboration (how well a theory has stood up to tests) replaces confirmation but does not provide positive probability.
Critics noted that falsification is rarely clear-cut.
Theories rarely confront experience alone (Duhem-Quine).
Ad hoc rescues are always available.
Critical rationalism and post-Popperian methodology
Lakatos
Imre Lakatos’s “Falsification and the Methodology of Scientific Research Programmes” (1970, in Criticism and the Growth of Knowledge) developed a more nuanced framework.
Scientific research programmes consist of a hard core of central assumptions, surrounded by a protective belt of auxiliary hypotheses.
The negative heuristic deflects falsifying evidence into modifications of the belt.
Programmes are progressive if their modifications generate novel predictions confirmed by evidence.
They are degenerating otherwise.
Feyerabend
Paul Feyerabend’s Against Method: Outline of an Anarchistic Theory of Knowledge (1975) argued that methodological monism is empirically false and harmful.
His provocation: “anything goes.”
Historical case studies (Galileo’s defense of Copernicus involved ad hoc moves) show no consistent method underlies scientific success.
Feyerabend’s later work emphasized epistemic pluralism and the value of pre-scientific traditions.
Kuhn and paradigms
Thomas Kuhn’s The Structure of Scientific Revolutions (1962, 2nd ed. 1970 with postscript) reshaped the field.
Science alternates between phases.
Normal science is research within an accepted paradigm.
A paradigm is a worldview, exemplar problems and solutions, instruments, and training.
Crisis is accumulation of anomalies the paradigm cannot accommodate.
Revolution is paradigm shift to a new framework.
New normal science follows under the new paradigm.
Key concepts
Paradigm: Kuhn used the term in many ways.
He later preferred “disciplinary matrix” and distinguished exemplars (concrete problem solutions used as models).
Incommensurability: successive paradigms share no neutral observation language.
Terms shift meaning (mass in Newtonian vs relativistic physics).
Gestalt switch: paradigm shifts involve seeing the world differently rather than accumulating discoveries.
Kuhn was widely interpreted as anti-realist or relativist, a charge he resisted.
His influence extended far beyond philosophy into sociology, history, and the humanities.
Bayesian confirmation theory
Bayesian philosophy of science treats scientific reasoning as updating probabilistic credences.
Rudolf Carnap’s Logical Foundations of Probability (1950) sought a logical interpretation of probability with confirmation functions.
Colin Howson and Peter Urbach’s Scientific Reasoning: The Bayesian Approach (1989, 3rd ed. 2006) gave a systematic Bayesian account.
The book covers theory confirmation, hypothesis testing, and the Duhem-Quine problem.
Variants
Subjective Bayesianism: probabilities are agents’ degrees of belief, constrained only by probabilistic coherence.
Objective Bayesianism: additional constraints (maximum entropy, indifference).
Wesley Salmon, James Joyce, and others developed accuracy-first arguments for probabilism.
Strengths and difficulties
Strengths: handles confirmation by varied evidence, the role of priors, the problem of old evidence.
Difficulties: the problem of the priors (why these credences?), the problem of zero priors (cannot be raised by conditionalization), tractability for complex hypotheses.
Inference to the best explanation
Gilbert Harman’s “The Inference to the Best Explanation” (Philosophical Review 1965) argued that enumerative induction is a special case of IBE.
We infer to hypotheses that, if true, would best explain the data.
Peter Lipton’s Inference to the Best Explanation (1991, 2nd ed. 2004) gave a systematic defense.
He distinguished the inference to the likeliest from the inference to the loveliest explanation.
Recent debate: Igor Douven’s The Art of Abduction (2021) contests whether IBE is compatible with Bayesian conditionalization or a heuristic distinct from probability theory.
Underdetermination
The Duhem-Quine thesis (Pierre Duhem 1906, W. V. O. Quine “Two Dogmas of Empiricism” Philosophical Review 1951) holds that theoretical hypotheses cannot be tested in isolation.
They are tested only in bundles with auxiliary assumptions.
Any apparent refutation can be deflected by modifying an auxiliary.
Underdetermination of theory by data: for any body of evidence, multiple incompatible theories are compatible with it.
Quine’s holism radicalized the thesis.
In principle, any belief can be retained by sufficiently revising others.
Realism and anti-realism
A central debate concerns whether scientific theories should be interpreted as approximately true descriptions of mind-independent reality.
Scientific realism
The mainstream realist position holds the following.
Mature scientific theories are approximately true.
Their central theoretical terms genuinely refer.
The success of science is best explained by approximate truth.
Putnam’s “no miracle argument”: it would be miraculous if science were so predictively successful while being wholly wrong about underlying structure.
Stathis Psillos’s Scientific Realism: How Science Tracks Truth (1999) gives a systematic defense.
Richard Boyd’s “Lex Orandi est Lex Credendi” (1984) defends explanationist realism.
Structural realism
A middle position responds to the pessimistic meta-induction.
Theories of the past have been overturned, so current theories will be too.
John Worrall’s “Structural Realism: The Best of Both Worlds?” (Dialectica 1989) defends epistemic structural realism.
We know structure (mathematical relations), not the underlying nature of entities.
James Ladyman’s “What Is Structural Realism?” (Studies in History and Philosophy of Science 1998) and Ladyman and Don Ross’s Every Thing Must Go: Metaphysics Naturalized (2007) defend ontic structural realism.
On their view, structure is all there is.
Anti-realist positions
Constructive empiricism — Bas van Fraassen’s The Scientific Image (1980) — holds that science aims at empirically adequate theories, not at truth about unobservables.
Acceptance of a theory involves believing it empirically adequate (saves the phenomena) and committing to use it for inference and explanation, without believing it true.
Instrumentalism (Mach, Duhem) treats theories as computational instruments for predicting observations.
Entity realism — Ian Hacking’s Representing and Intervening (1983) — accepts entities we can manipulate (“if you can spray them, they’re real”) while remaining agnostic about theories.
Nancy Cartwright’s How the Laws of Physics Lie (1983) holds that phenomenological laws (with restricted scope) are true.
But fundamental laws are false because they hold only under ceteris paribus conditions never fully realized.
Laws of nature
Deductive-nomological model
Carl Hempel and Paul Oppenheim’s “Studies in the Logic of Explanation” (Philosophy of Science 1948) holds that to explain is to deduce the explanandum from laws plus initial conditions.
The covering-law model dominated mid-century philosophy of science.
Humean supervenience
David Lewis’s Counterfactuals (1973) and “Humean Supervenience Debugged” (Mind 1994) hold that laws supervene on the total Humean mosaic of local matters of particular fact.
The Best System Account (Mill-Ramsey-Lewis) holds that laws are theorems of the deductively closed system that best balances simplicity and informativeness.
Necessitarianism
David Armstrong’s What Is a Law of Nature? (1983); Fred Dretske’s “Laws of Nature” (1977); Michael Tooley’s “The Nature of Laws” (1977) hold that laws are second-order relations of nomic necessitation (N(F,G)) between universals.
Dispositional essentialism
Brian Ellis’s Scientific Essentialism (2001) and Alexander Bird’s Nature’s Metaphysics (2007) hold that laws derive from the essential dispositions of natural properties.
Primitivism
Tim Maudlin’s The Metaphysics Within Physics (2007) treats laws as primitive ontological entities.
Causation in science
Counterfactual: David Lewis’s “Causation” (Journal of Philosophy 1973) holds that c causes e iff in the closest worlds where c does not occur, e does not occur.
Interventionist / manipulationist: James Woodward’s Making Things Happen (2003) holds that X causes Y iff there is a possible intervention on X that changes Y.
This captures the practical sense of causation in experimental sciences.
Mechanistic: Stuart Glennan’s “Mechanisms and the Nature of Causation” (Erkenntnis 1996); Peter Machamer, Lindley Darden, and Carl Craver (MDC), “Thinking about Mechanisms” (Philosophy of Science 2000), hold that causation requires the operation of productive mechanisms.
Probabilistic: Patrick Suppes’s A Probabilistic Theory of Causality (1970); Nancy Cartwright; Wesley Salmon develop the probabilistic approach.
Scientific explanation
DN model (Hempel-Oppenheim 1948): explanation by deduction from laws.
Inductive-statistical (IS) model (Hempel): explanation of probable but non-certain events.
Statistical-relevance model (Wesley Salmon’s Statistical Explanation and Statistical Relevance 1971): explanation by citing factors that change the probability of the explanandum.
Causal-mechanical model (Salmon’s Scientific Explanation and the Causal Structure of the World 1984): explain by tracing causal processes and interactions.
Unification: Michael Friedman’s “Explanation and Scientific Understanding” (Journal of Philosophy 1974); Philip Kitcher’s “Explanatory Unification” (Philosophy of Science 1981): explanation unifies disparate phenomena under common patterns of derivation.
Pragmatic theories: Bas van Fraassen’s The Scientific Image (1980) ch. 5: explanations are answers to why-questions, relative to contrast classes and relevance relations.
Reductionism and unity of science
Ernest Nagel’s The Structure of Science (1961) treats reduction as derivation of one theory from another via bridge laws.
Oppenheim and Putnam’s “Unity of Science as a Working Hypothesis” (1958) describes a hierarchy of levels (elementary particles → atoms → molecules → cells → organisms → social groups) with each reducible to the next.
Multiple realizability argument: Jerry Fodor’s “Special Sciences (or: The Disunity of Science as a Working Hypothesis)” (Synthese 1974) holds that psychological types can be realized by indefinitely many physical types.
So psychology is not reducible to physics.
Special sciences enjoy autonomy.
Supervenience: Donald Davidson’s “Mental Events” (1970) holds that mental properties supervene on physical without being reducible.
Pluralism
John Dupré’s The Disorder of Things (1993) defends promiscuous realism — no single classification scheme is privileged.
Sandra Mitchell’s Unsimple Truths (2009) defends pluralism in biology — multiple compatible models for complex systems.
Nancy Cartwright’s The Dappled World (1999) defends a patchwork of local laws rather than a universal nomic order.
Models, idealization, simulation
A flourishing area since the 1990s recognizes models as central to scientific practice.
Roman Frigg and Stephan Hartmann’s “Models in Science” (SEP updated; Models as Mediators Morgan and Morrison eds. 1999) treats models as mediators between theory and the world.
Models have partial autonomy from both.
Idealization: Michael Weisberg’s Simulation and Similarity (2013); Angela Potochnik’s Idealization and the Aims of Science (2017).
Computer simulation as a distinct epistemic tool: Eric Winsberg’s Science in the Age of Computer Simulation (2010).
Values in science
Helen Longino’s Science as Social Knowledge (1990) and The Fate of Knowledge (2002) hold that objectivity is achieved through critical discourse among diverse perspectives.
Values are ineliminable from science on her view.
Heather Douglas’s Science, Policy, and the Value-Free Ideal (2009) addresses inductive risk.
Scientists must judge how much evidence is enough, and these judgments depend on the costs of error, which are evaluative.
Standpoint epistemology: Sandra Harding’s Whose Science? Whose Knowledge? (1991), Sciences from Below (2008).
Philosophy of physics
Interpretations of quantum mechanics
Copenhagen: Bohr, Heisenberg; measurement collapses the wave function; complementarity.
Many-worlds: Hugh Everett, “Relative State Formulation of Quantum Mechanics” (Reviews of Modern Physics 1957); Bryce DeWitt 1970 popularization — all branches of the wave function are real.
Pilot wave / de Broglie-Bohm: David Bohm, “A Suggested Interpretation of the Quantum Theory” (Physical Review 1952); deterministic non-local hidden variables.
John Bell’s “On the Einstein Podolsky Rosen Paradox” (Physics 1964) showed any local hidden-variable theory contradicts QM predictions.
Consistent histories: Robert Griffiths, Roland Omnès.
QBism (quantum Bayesianism): Christopher Fuchs and collaborators; the wave function represents an agent’s degrees of belief.
Time and relativity
Tim Maudlin’s Quantum Non-Locality and Relativity (1994) and Philosophy of Physics: Space and Time (2012) address the relation between QM and relativity.
The block universe debate continues.
Time direction is treated via the second law and entropy.
Cosmology and multiverse
Fine-tuning arguments, anthropic principle, and Boltzmann brains are central topics.
Philosophy of biology
David Hull’s “The Effect of Essentialism on Taxonomy” (British Journal for the Philosophy of Science 1965) holds that species are individuals (spatiotemporally extended lineages), not natural kinds.
Elliott Sober’s The Nature of Selection (1984) and Philosophy of Biology (1993, 2nd ed. 2000) address units of selection and the propensity interpretation of fitness.
Ernst Mayr distinguished proximate vs ultimate causes and population thinking vs typological thinking.
Richard Lewontin’s The Triple Helix (2000) holds that gene, organism, environment co-determine development.
Teleology and function
Larry Wright’s “Functions” (Philosophical Review 1973); Robert Cummins’s “Functional Analysis” (Journal of Philosophy 1975); Ruth Millikan’s Language, Thought, and Other Biological Categories (1984) develop competing accounts.
Evo-devo philosophy addresses developmental constraints and modularity.
Peter Godfrey-Smith’s Theory and Reality: An Introduction to the Philosophy of Science (2003, 2nd ed. 2021) and Other Minds (2016) on cephalopod cognition are widely read.
Philosophy of mathematics
Major positions on the nature of mathematical objects and truth:
Logicism
Gottlob Frege’s Grundgesetze der Arithmetik (1893, 1903); Bertrand Russell and A. N. Whitehead’s Principia Mathematica (1910-1913) held that mathematics is reducible to logic.
Frege’s program was blocked by Russell’s paradox.
Neologicism (Crispin Wright, Bob Hale) has revived a modified version.
Intuitionism
L. E. J. Brouwer’s “Intuitionism and Formalism” (1912) and Arend Heyting hold that mathematics is mental construction.
The view rejects the law of excluded middle for infinite domains.
Formalism
David Hilbert’s “On the Foundations of Logic and Arithmetic” (1904) launched the formalist program.
The program aimed to prove the consistency of mathematics by purely finitary means.
Incompleteness
Kurt Gödel’s “Über formal unentscheidbare Sätze der Principia Mathematica und verwandter Systeme I” (1931) proved the incompleteness theorems.
Any consistent formal system rich enough to express arithmetic contains true statements unprovable within the system.
The system cannot prove its own consistency.
This blocked Hilbert’s program.
Platonism
Gödel’s “What Is Cantor’s Continuum Problem?” (1947) defends Platonism.
Abstract mathematical objects exist; we perceive them through mathematical intuition.
Structuralism
Michael Resnik’s Mathematics as a Science of Patterns (1997) and Stewart Shapiro’s Philosophy of Mathematics: Structure and Ontology (1997) hold that mathematics studies structures, not particular objects.
Nominalism
Hartry Field’s Science Without Numbers (1980) holds that mathematics is a useful fiction.
The program nominalizes physics by dispensing with mathematical objects.
Yablo and Bueno extend the program in different directions.
Philosophy of social science
Wilhelm Dilthey distinguished Naturwissenschaften (natural sciences, explanation) from Geisteswissenschaften (human sciences, understanding/Verstehen).
Max Weber’s Economy and Society (1922) defended methodological individualism, ideal types, and interpretive sociology.
Frankfurt School critical theory: Theodor Adorno and Max Horkheimer’s Dialectic of Enlightenment (1947); Jürgen Habermas’s Knowledge and Human Interests (1968) and The Theory of Communicative Action (1981); Axel Honneth’s The Struggle for Recognition (1992).
The positivism dispute (Positivismusstreit) of the 1960s pitted Adorno/Habermas against Popper/Albert.
Hermeneutics: Hans-Georg Gadamer’s Truth and Method (1960) and Paul Ricoeur’s work.
Recent developments
The Stanford Encyclopedia of Philosophy (Edward Zalta, ed., 1995-present) is a peer-reviewed open-access reference of unprecedented quality and scope.
Contemporary leading figures include David Chalmers, Steven Weinstein, Jenann Ismael (How Physics Makes Us Free 2016), Tim Maudlin, Hilary Putnam, Bas van Fraassen, Nancy Cartwright, Helen Longino, Sandra Harding, Heather Douglas, Kyle Stanford (Exceeding Our Grasp 2006 on unconceived alternatives), and Anjan Chakravartty.
Increasing engagement with practice-based philosophy of science, philosophy of medicine, philosophy of climate science, and philosophy of AI/ML is reshaping the field.
Philosophy of medicine
Philosophy of medicine is an emerging subfield.
Key topics include the nature of disease, evidence-based medicine, randomized controlled trials, causation in epidemiology, and the role of values in clinical decision-making.
Christopher Boorse’s “Health as a Theoretical Concept” (Philosophy of Science 1977) defends a biostatistical theory of disease.
Disease is statistically subnormal species-typical functioning.
Naturalist and normativist accounts of health remain in tension.
The hierarchy of evidence in evidence-based medicine has been examined by Jeremy Howick (The Philosophy of Evidence-Based Medicine 2011) and Nancy Cartwright.
Philosophy of cognitive science
This subfield investigates representation, computation, modularity, and the relation between cognitive science and neuroscience.
Daniel Dennett, Andy Clark, and Patricia Churchland have shaped its development.
Recent debates concern deep learning and whether neural-network models constitute genuine cognitive theories.
Philosophy of climate science
A young area.
Wendy Parker’s work on climate model ensembles and Eric Winsberg’s Philosophy and Climate Science (2018) address how to interpret model disagreement and uncertainty.
The relation between models, simulations, and observational data is central.
Philosophy of data and machine learning
A rapidly developing field.
Topics include: what counts as data; the role of inductive bias in learning; the interpretability and explanation problems for deep networks; the epistemology of large-scale ML; algorithmic fairness.
Authors include Sabina Leonelli (Data-Centric Biology 2016), Brian Cantwell Smith, Cathy O’Neil, and recent work in machine learning theory and philosophy.
Recent topical trends
Philosophy of science in the 21st century has shown several distinctive trends.
Greater attention to scientific practice rather than reconstructed logic.
Closer engagement with specific sciences (climate science, neuroscience, ML, synthetic biology).
Renewed interest in the social epistemology of science.
Critical examination of replication crises in psychology and biomedicine.
Philosophical work on registered reports, preprints, and open science.