Imre Lakatos proposed that science advances not by isolated theories or by simple falsification, but through competing “research programmes.” Each programme has two parts:

• Hard core: the programme’s central, fundamental assumptions that scientists protect from revision.
• Protective belt: auxiliary hypotheses and assumptions that can be modified to defend the hard core from apparent refuting evidence.

He also distinguished heuristics:

• Negative heuristic: tells researchers not to question the hard core.
• Positive heuristic: guides the development of new auxiliary hypotheses, predictions, and extensions to increase the programme’s empirical reach.

Progress occurs when a programme is “progressive”: it predicts novel facts and leads to theoretical and empirical growth. A programme is “degenerating” if it only makes ad hoc adjustments to the protective belt to accommodate anomalies without producing new, corroborated predictions. Competing programmes are rationally appraised by their progressive or degenerative character over time rather than by single critical tests.

Key implications:

• Science is rational and cumulative but complex: refutations are handled by modifying the protective belt rather than immediate abandonment.
• Historical and methodological evaluation matters: success is judged over a sequence of theory changes, not one-off experiments.

Further reading: Lakatos, “Falsification and the Methodology of Scientific Research Programmes” (in Criticism and the Growth of Knowledge, 1970).

Karl Popper’s The Logic of Scientific Discovery argues that the central demarcation between science and non-science is falsifiability: a theory is scientific only if it makes bold, testable predictions that could, in principle, be refuted by observation or experiment. Popper rejects inductivism (the idea that science proceeds by accumulating confirmed observations to establish universal laws). Instead he proposes conjectures and refutations: scientists propose tentative hypotheses and subject them to severe tests; surviving hypotheses are corroborated but never finally proven. Progress consists in replacing less adequate theories with better ones that explain anomalies. Popper also emphasizes the logical asymmetry between verification and falsification (no number of positive instances can conclusively verify a universal law, but a single counterinstance can falsify it) and criticizes ad hoc adjustments that immunize theories from refutation. Key influences include his focus on critical rationalism, the role of theory in observation, and his skepticism about probability-based confirmations.

Suggested further reading: Popper, K. R. (1959/2002) The Logic of Scientific Discovery; also Lakatos’ “Research Programmes” and Kuhn’s “Structure of Scientific Revolutions” for contrasting views.

1. Logical Positivism / Logical Empiricism

• Core idea: Scientific knowledge is grounded in empirical observation and logical analysis; meaningful statements are either empirically verifiable or analytic (truths of logic/language).
• Key claims: Emphasis on observational language, confirmation via data, reduction of theoretical terms to observational ones.
• Criticisms: Verification principle is self-refuting; theory-ladenness of observation; failure to account for theoretical virtues.
• Representative figures: Carnap, Neurath, early Vienna Circle.
• Sources: A. J. Ayer, Rudolf Carnap.

2. Popperian Falsificationism

• Core idea: Science progresses by bold conjectures and refutations; hypotheses are scientific only if falsifiable.
• Key claims: Verification is impossible; empirical testing aims to falsify rather than confirm; corroboration not equal to truth.
• Criticisms: Serious theories are rarely strictly falsified (auxiliary hypotheses); histories of science show theory choice isn’t purely logical.
• Representative figure: Karl Popper.
• Sources: Karl Popper, The Logic of Scientific Discovery.

3. Kuhnian Paradigms and Scientific Revolutions

• Core idea: Science advances through paradigm-driven normal science punctuated by revolutions that replace one paradigm with another.
• Key claims: Paradigms shape what counts as problems, methods, and standards; paradigm shifts involve incommensurability and are not purely rational.
• Criticisms: Exaggerates discontinuity; understates rational criteria for theory choice.
• Representative figure: Thomas Kuhn.
• Sources: Thomas Kuhn, The Structure of Scientific Revolutions.

4. Lakatosian Research Programmes

• Core idea: Middle path between Popper and Kuhn: science advances via competing research programmes with a hard core of assumptions protected by auxiliary hypotheses.
• Key claims: Progressive programmes predict novel facts; degenerating ones only accommodate anomalies ad hoc.
• Criticisms: Difficulty in demarcating progress; sociological influences remain underplayed.
• Representative figure: Imre Lakatos.
• Sources: Imre Lakatos, “Falsification and the Methodology of Scientific Research Programmes.”

5. Feyerabend’s Epistemological Anarchism

• Core idea: There is no single scientific method; “anything goes”—historical creativity and pluralism matter more than strict rules.
• Key claims: Methodological pluralism fosters progress; methodological constraints can stifle innovation.
• Criticisms: Risks relativism and undermining science’s authority.
• Representative figure: Paul Feyerabend.
• Sources: Paul Feyerabend, Against Method.

6. Scientific Realism

• Core idea: Best scientific theories aim to describe a mind-independent reality; unobservable entities posited by successful theories are (approximately) real.
• Key claims: Success of science is best explained by truth or approximate truth of theories.
• Criticisms: Pessimistic meta-induction (many past successful theories were false); underdetermination by data.
• Representative figures: Putnam, Boyd.
• Sources: Richard Boyd; Hilary Putnam.

7. Scientific Anti-Realism / Instrumentalism / Constructive Empiricism

• Core idea: Theories are tools for organizing observations and making predictions; commitment to unobservables is unnecessary or unwarranted.
• Key claims: Acceptance of theories equals empirical adequacy, not belief in literal truth.
• Criticisms: May underplay explanatory ambitions and success of theory-laden predictions.
• Representative figure: Bas van Fraassen (constructive empiricism).
• Sources: Bas van Fraassen, The Scientific Image.

8. Social Constructivism and Sociology of Scientific Knowledge (SSK)

• Core idea: Scientific knowledge is influenced or constructed by social, cultural, and institutional factors; facts are, in part, socially negotiated.
• Key claims: Emphasizes role of scientific communities, power, and interests in shaping what counts as knowledge.
• Criticisms: Strong forms can slide into relativism or deny empirical constraints.
• Representative figures: David Bloor, Bruno Latour.
• Sources: David Bloor, Knowledge and Social Imagery; Bruno Latour, Science in Action.

Brief note on method and unity/diversity: Debates also concern whether there is a single scientific method or multiple methods, and whether sciences form a unified enterprise (unity of science) or a heterogeneous set of practices.

For further reading: Kuhn, Popper, Lakatos, Feyerabend, van Fraassen, and Stanford’s SEP entries on “Scientific Realism” and “Philosophy of Science.”

Values In the philosophy of science, “values” are the social, moral, political, and personal considerations that influence scientific practice. They operate at different stages: choice of research topics, funding priorities, methodological trade-offs (e.g., simplicity vs. completeness), interpretation of data, and application of findings. Philosophers distinguish between cognitive (epistemic) values—such as accuracy, simplicity, explanatory power, coherence—that guide theory choice, and non‑epistemic values—such as social benefit, safety, economic interests, or moral commitments—that can legitimately (or contentiously) influence decisions about what to study and how to apply results. Debates center on how and when non‑epistemic values should shape science without undermining its reliability.

Objectivity Objectivity is the ideal that scientific knowledge should be unbiased, not dependent on particular observers’ viewpoints, and grounded in evidence and reasons accessible to others. Philosophical accounts vary:

• Correspondence and truth‑tracking views tie objectivity to methods that reliably lead to truth.
• Value‑neutrality views hold that science should be insulated from non‑epistemic values to preserve impartiality.
• Social and procedural views (e.g., Helen Longino) argue objectivity is achieved through critical interaction: diverse perspectives, open criticism, and institutional norms that constrain individual bias. Contemporary consensus: complete value‑neutrality is unrealistic; objectivity is better construed as robust procedures and social practices that manage biases and make claims publicly accountable.

Ethics Ethics in science concerns responsibilities of scientists and institutions: honesty in data collection and reporting, avoidance of fraud, respect for human/animal subjects, responsible communication of uncertainty, and consideration of societal consequences. Ethical issues intersect with values and objectivity when non‑epistemic concerns rightly shape research priorities (e.g., research on diseases) or when ethical constraints limit certain inquiries. Ethics also demands transparency about value judgments and conflicts of interest so that the epistemic integrity and social legitimacy of science are preserved.

Brief synthesis Values shape what science studies and how results are used; objectivity is an aspirational standard achieved through transparent, critical, and institutional practices that mitigate bias; ethics governs the responsibilities and societal effects of scientific activity. Recognizing the interplay among them helps maintain science that is reliable, socially responsible, and publicly trustworthy.

References (suggested further reading)

• Helen Longino, Science as Social Knowledge (1990).
• Heather Douglas, Science, Policy, and the Value‑Free Ideal (2009).
• Karl Popper, The Logic of Scientific Discovery (1959) for classical views on objectivity and method.

Scientific reasoning uses several related but distinct logical moves:

• Deduction

  • What it is: Reasoning from general laws or premises to specific consequences (if the premises are true, the conclusion must be true).
  • Role in science: Used to derive testable predictions from theories (theory → hypothesis → expected observation).
  • Strengths and limits: Logically certain given premises, but does not by itself justify the truth of the premises (it’s truth-preserving, not truth-producing). See Hempel’s covering-law model.

• Induction

  • What it is: Reasoning from particular observations to broader generalizations (observed cases → general law or probability).
  • Role in science: Used to form empirical generalizations and estimate parameters from data (e.g., from many measurements infer a law or statistical regularity).
  • Strengths and limits: Empirically indispensable but logically problematic (the “inductive problem” or Hume’s problem of induction): past regularities do not guarantee future ones. Bayesian approaches recast induction as probabilistic updating.

• Abduction (inference to the best explanation)

  • What it is: Reasoning from surprising or puzzling facts to the best explanatory hypothesis that would, if true, make the facts expectable.
  • Role in science: Central to theory generation and model selection—choosing hypotheses that best explain data given simplicity, coherence, explanatory power, and predictive success.
  • Strengths and limits: Pragmatic and inference-driven rather than deductively certain; criteria for “best” are partly normative and contested (see Peirce on abduction and more recent work on IBE).

How they interact in scientific practice

• Science typically cycles: abduction proposes hypotheses, deduction generates predictions from those hypotheses, and induction (or statistical inference/Bayesian updating) assesses how well observations support or revise hypotheses.
• Philosophical responses: Popper emphasized falsification (deductive testing), Bayesianism formalizes inductive support, and philosophers of science highlight the creative abductive step in theory choice (see Popper 1959; Peirce; Salmon; Kuhn; Hacking; Howson & Urbach).

Further reading (concise)

• C. S. Peirce on abduction; K. Popper, The Logic of Scientific Discovery (falsification); P. Achinstein, The Book of Evidence; I. Hacking, An Introduction to Probability and Inductive Logic; Howson & Urbach, Scientific Reasoning: The Bayesian Approach.

Contemporary issues in the philosophy of science examine how scientific knowledge is produced, validated, and used in complex social and epistemic contexts. Key topics include:

• Scientific realism vs. anti-realism: Debates about whether scientific theories genuinely describe unobservable reality (realism) or merely serve as instruments for prediction and manipulation (instrumentalism/constructive empiricism). Recent work focuses on selective realism (which parts of theories are true) and structural realism (we can know structure, not nature of unobservables). (See Psillos 1999; Worrall 1989.)

• Theory change and scientific revolutions: Beyond Kuhn’s paradigms, philosophers explore continuity vs. discontinuity in theory change, rationality of theory choice, and mechanisms (e.g., model-based inferences, Bayesian updating). (Kuhn 1962; Bird 2007.)

• Models, representation, and explanation: Contemporary accounts analyze how idealized and computational models represent phenomena, how explanations can be causal, mechanistic, or unification-based, and how multiple models offer complementary insights. (Cartwright 1983; Strevens 2008; Woodward 2003.)

• Scientific inference and uncertainty: Topics include the role of probabilities, Bayesian methods, inference to the best explanation (IBE), issues of confirmation, underdetermination, and the social dimension of evidence assessment. (Howson & Urbach 2006; Lipton 2004.)

• Evidence, values, and objectivity: Philosophers examine how social, ethical, and political values influence research agendas, theory choice, and risk assessment, raising questions about objectivity, impartiality, and trust in science. (Longino 1990; Douglas 2009.)

• Science and society: Issues include public understanding of science, science policy, the role of expertise, and problems like climate change and public health where scientific uncertainty intersects with democratic decision-making. (Kitcher 2011; Oreskes & Conway 2010.)

• Interdisciplinarity and big data: Philosophers investigate methodological challenges in interdisciplinary research, the epistemology of data-driven science, reproducibility crises, and the interpretation of large-scale statistical evidence. (Leonelli 2016; Ioannidis 2005.)

These contemporary issues show that philosophy of science now combines traditional metaphysical and epistemic questions with practical concerns about how science operates within society, technology, and policy.

The replication crisis refers to the widespread discovery that many published scientific results—especially in psychology, biomedical sciences, and some social sciences—fail to be reproduced when independent researchers repeat the same studies. Replication is a basic scientific test: if a finding reflects a real effect, other competent investigators using the same methods should obtain similar results. When replications fail, confidence in the original claim, its methods, or underlying theory is undermined.

Key causes

• Low statistical power: small sample sizes make false positives and effect-size exaggeration more likely.
• P-hacking and selective reporting: researchers (consciously or not) try many analyses and report only those that reach conventional significance (p < .05).
• Publication bias: journals favor novel, positive results over null or replication studies, skewing the literature.
• Poor methodological transparency: insufficient reporting of materials, data, and procedures prevents exact replication.
• Questionable research practices and incentives: career pressures prioritize quantity and novelty over rigor.

Consequences for reliability

• Many published effects are overestimated or false, reducing trust in disciplines and slowing cumulative knowledge.
• Meta-analyses and policy decisions built on biased literature may mislead practice.
• Replication failures prompt re-evaluation of methods, theories, and standards of evidence.

Responses and reforms

• Pre-registration of study plans and hypotheses to limit fishing for significance (Nosek et al., 2018).
• Open data, materials, and code to enable exact replication and reanalysis (Munafò et al., 2017).
• Larger, better-powered studies and multi-lab replications (Simons et al., 2014).
• Registered reports and journal reforms that commit to publishing based on methods rather than results.
• Improved statistical practices: emphasis on effect sizes, confidence intervals, Bayesian methods, and better correction for multiple testing.

Implication for philosophy of science The crisis highlights tensions between idealized models of scientific progress (cumulative, self-correcting) and the practice shaped by human, social, and institutional factors. It stresses the importance of methodological norms, transparency, and incentives for reliable knowledge production (Ioannidis, 2005).

Suggested readings

• Ioannidis, J. P. A. (2005). “Why Most Published Research Findings Are False.” PLoS Medicine.
• Nosek, B. A., et al. (2018). “The preregistration revolution.” Proceedings of the National Academy of Sciences.
• Open Science Collaboration (2015). “Estimating the reproducibility of psychological science.” Science.

Thomas Kuhn’s The Structure of Scientific Revolutions argues that science does not progress via a steady, cumulative accumulation of facts but through a cyclical process centered on paradigms — the shared theories, methods, and exemplars that guide normal scientific practice. In normal science, researchers solve puzzles within the existing paradigm. Persistent anomalies eventually accumulate, producing crisis. Crisis can lead to a scientific revolution: a paradigm shift in which the old framework is replaced by a new one that reinterprets data, poses different questions, and uses different standards. Paradigm shifts are incommensurable in important respects: proponents of different paradigms may talk past one another because they operate with different concepts and criteria for evaluation. Kuhn’s account emphasizes the social and historical dimensions of scientific change and challenges the view of scientific progress as purely rational or linear.

Key concepts: paradigm, normal science, anomaly, crisis, scientific revolution, incommensurability.

Primary source: Thomas S. Kuhn, The Structure of Scientific Revolutions (1962). For commentary: Ian Hacking, “Kuhn and History,” and Larry Laudan, Science and Values (for critiques).

Scientific theories explain phenomena by positing models, laws, and mechanisms that show how and why certain events occur. Explanation typically involves fitting particular observations into general patterns: a theory identifies relevant variables, states regularities (laws) or causal mechanisms, and situates a phenomenon as an instance of those general principles. Two common explanatory schemas are the deductive-nomological (DN) model—where explanations subsume events under general laws via logical deduction—and mechanistic explanations—where one describes component parts and their interactions that produce the phenomenon (Salmon; Machamer, Darden, Craver).

Prediction follows from the theory’s structure and empirical content. Given initial conditions and the theory’s laws or equations, one can derive expectations about future or unobserved states. Reliable prediction depends on the theory’s scope, accuracy of auxiliary assumptions (background conditions and measurements), and idealizations. When a theory furnishes both successful explanations and novel, testable predictions, it gains credibility; failures prompt revision or replacement (Popper on falsification; Kuhn on paradigms).

Key points:

• Explanation = showing how a phenomenon fits general laws or mechanisms.
• Prediction = deriving specific outcomes from those laws plus initial/boundary conditions.
• Both depend on auxiliary assumptions, idealizations, and empirical testing. References: Carl Hempel (DN model), Wesley Salmon; Machamer, Darden & Craver on mechanisms; Karl Popper; Thomas Kuhn.

• Science policy Science policy encompasses government, institutional, and organizational decisions that shape how scientific research is funded, regulated, and applied. It covers priorities for research investment, ethical and safety rules (e.g., clinical trial regulations, biosecurity), intellectual property, and how evidence is used in policy-making. Science policy mediates between scientific knowledge and social goals: it determines what kinds of questions get investigated, how results are translated into technology or regulation, and who benefits. Key references: the OECD’s work on science policy, and Sheila Jasanoff’s writings on co-production of science and policy.

• Public trust Public trust in science is the confidence citizens place in scientific institutions, experts, and findings. Trust depends on perceived competence (experts know what they’re doing), integrity (honest and unbiased practices), and benevolence (work serves public good). Trust is crucial for uptake of scientific recommendations (e.g., vaccination campaigns) and for legitimacy of science-informed policy. Erosion of trust can follow perceived conflicts of interest, lack of transparency, politicization of science, or failures in communication. See classic discussions by Helen Longino and more recent empirical work on trust and expertise (e.g., Wynne, 1992; Oreskes & Conway, 2010).

• Science communication Science communication is the practice of translating scientific knowledge for various audiences: policymakers, journalists, educators, and the public. It includes reporting results, explaining uncertainties and methods, and engaging in dialogue. Effective science communication balances accuracy with clarity, acknowledges uncertainty without undermining credibility, and considers audience values and cultural context. Approaches range from deficit models (one-way information transfer) to dialogical and participatory models that build mutual understanding and trust. Key sources: Brian Wynne on public engagement, and recent guides on risk communication (e.g., National Academies of Sciences, Engineering, and Medicine).

Interrelations (brief) Science policy sets the institutional context for research and communication; effective science communication helps build and sustain public trust; and public trust, in turn, influences the success of science policy. Failures in any one area can weaken the others—for example, poor communication about uncertainty can fuel mistrust and politicized policy debates.

For further reading:

• Sheila Jasanoff, States of Knowledge: The Co-Production of Science and the Social Order
• Naomi Oreskes & Erik Conway, Merchants of Doubt
• National Academies, Communicating Science Effectively: A Research Agenda

Pragmatic and pluralist approaches hold that there is no single correct kind of scientific explanation that fits every situation. Instead, the appropriate explanatory form depends on the questions being asked, the aims of the investigators, the audience, and the available methods and evidence. Key points:

• Context-sensitivity: Explanations are chosen for their usefulness relative to particular goals (prediction, control, understanding, policy, education). What counts as a good explanation in basic research (e.g., mechanistic detail) may differ from what matters in applied settings (e.g., reliable models for intervention).

• Multiple legitimate forms: Different sciences and problems employ distinct explanatory types—mechanistic explanations (how parts produce phenomena), causal-statistical explanations (probabilities and causes), functional explanations (why a trait persists), mathematical or unifying explanations (deriving phenomena from general laws), and model-based explanations (idealizations and simulations). Pluralism treats these as complementary tools rather than competitors.

• Practical constraints and trade-offs: Choice of explanation reflects limitations (data, tractability) and pragmatic trade-offs (precision vs. generality, realism vs. simplicity). For instance, idealized models can yield understanding despite false assumptions because they illuminate causal structure or dependencies.

• Epistemic and normative roles: Pragmatism emphasizes that explanatory standards also serve epistemic aims (warrant, reliability) and normative ones (policy relevance, communication). Hence debates about “the best” explanation often reduce to disputes about aims and context.

Relevant references: Salmon and Hempel for classic accounts; recent discussions in pluralist and pragmatic literature by Bas van Fraassen, Peter Achinstein, and Heather Douglas; see also Make It Better: pluralism in philosophy of science (e.g., Ioannidis on models).

Peter Godfrey-Smith’s Theory and Reality is a concise, accessible introduction to central issues in the philosophy of science. It presents competing accounts of scientific knowledge and method, emphasizing clarity and impartial exposition over doctrinaire advocacy.

Key features:

• Comparative overview: Surveys major positions—logical positivism, falsificationism (Popper), confirmation theory (Bayesians), scientific realism and anti-realism, Kuhnian paradigms, Lakatosian research programmes, and the sociology and history-informed critiques of science.
• Emphasis on realism debate: Carefully distinguishes varieties of scientific realism (e.g., entity realism, structural realism) and assesses arguments for and against the claim that successful theories are true or approximately true.
• Role of evidence and inference: Explores the logic of confirmation, Bayesianism, and problems like underdetermination and theory-ladenness.
• Historical and sociological context: Incorporates lessons from the history of science (Kuhn, Feyerabend) and contemporary work in practice-based philosophy, showing how scientific practice can inform philosophical accounts.
• Philosophical balance and pedagogy: Written for students and non-specialists; provides clear examples, critical discussion, and suggested readings.

Why it matters: Theory and Reality functions as both a survey and a critical guide—helpful for understanding debates about what science aims to do, how scientific reasoning works, and whether scientific theories really describe the world. Its clarity and breadth make it a standard introductory text in philosophy of science courses.

Recommended further reading: van Fraassen, The Scientific Image (1980); Kuhn, The Structure of Scientific Revolutions (1962); Popper, Conjectures and Refutations (1963).

The covering-law model, chiefly developed by Carl Hempel, explains how scientific explanations work by showing that the event or phenomenon to be explained (the explanandum) is logically or probabilistically deduced from general laws and specific circumstances (the explanans). In the deductive-nomological (D-N) version, an explanation takes the form of a valid deduction: given at least one general law and relevant initial conditions, the occurrence follows necessarily. The inductive-statistical (I-S) version handles probabilistic cases by showing that the explanans renders the explanandum highly probable rather than logically certain.

Key features

• Subsumption: Explaining means subsuming a phenomenon under general laws—“this happened because it falls under this law plus these conditions.”
• Logical structure: Explanations are judged by logical relations (deduction or probabilistic support), not by causal storytelling.
• Explanatory asymmetry: The model had trouble with cases (e.g., flagpole shadow) where deduction works both ways but only one is explanatory; Hempel and later commentators addressed such issues with added conditions (e.g., lawlikeness, relevance).
• Limitations: Critics point out problems with relevance, singular causal explanations, and the role of mechanisms; alternative accounts (causal, mechanistic, unificationist) arose in response.

Sources: Hempel, C. G. (1965) “Aspects of Scientific Explanation”; Salmon, W. C. (1984) Scientific Explanation and the Causal Structure of the World.

Scientific realism is the view that our best scientific theories do more than organize observations: they (approximately) truly describe both observable and unobservable aspects of the world. Proponents like Hilary Putnam and Richard Boyd argue that the success of science—its explanatory power, predictability, and technological fruitfulness—is best explained by supposing that theories correctly capture real entities and structures (electrons, genes, fields) even when those entities are not directly observable.

Key points

• Approximate truth: Scientific claims need not be perfect; realism allows that theories are approximately true and improve over time (mature theories converge on truth).
• Unobservables considered real: Entities posited by successful theories are taken to exist because positing them provides the best explanation of empirical success.
• No-miracle argument: Putnam famously argued that the success of science would be a “miracle” if theories were not at least approximately true.
• Inference to the best explanation (IBE): Boyd and others frame acceptance of unobservables as an inference—because theories that posit them best explain and predict phenomena, we are justified in believing those entities exist.

References

• Putnam, H. (1975). “What Is Realism?” In Mind, Language and Reality.
• Boyd, R. (1984). “The Current Status of Scientific Realism,” in PSA: Proceedings of the Biennial Meeting of the Philosophy of Science Association.

Science aims to produce objective knowledge: methods, evidence, and communal scrutiny are used to minimize individual bias and secure reliable claims about the world. Objectivity is pursued through repeatable experiments, transparent methods, peer review, and statistical standards so that results do not rest on any single scientist’s whims.

Yet choice points in scientific practice are shaped by both epistemic and non‑epistemic values:

• Epistemic values guide what counts as good science (e.g., accuracy, coherence, simplicity, explanatory power, predictive success). These values influence which hypotheses are favored, which methods are chosen, and how evidence is interpreted because they promote truth‑tracking and reliable inference. (See Laudan 1977; Kitcher 1993.)

• Non‑epistemic values (social, ethical, political, economic) enter decisions that are not determined by evidence alone: what questions to investigate, acceptable risk thresholds, how to balance costs and benefits, and how to apply results. For example, deciding whether to fund research into risky technologies, or setting clinical trial stopping rules, requires judgments about safety, justice, or public welfare. (See Douglas 2009; Longino 1990.)

Because these non‑epistemic values legitimately shape priorities and applications, complete detachment is neither possible nor desirable. The key is transparency and critical scrutiny: make value judgments explicit, justify them, and show how they influence methods and interpretations. This preserves the epistemic integrity of science while responsibly acknowledging its social embedding.

References (select):

• K. R. Popper, The Logic of Scientific Discovery (1959)
• Larry Laudan, Progress and Its Problems (1977)
• Heather Douglas, Science, Policy, and the Value‑Free Ideal (2009)
• Helen Longino, Science as Social Knowledge (1990)

Structuralism and model-based accounts hold that scientific progress is best understood not as accumulating literally true, complete descriptions of the world but as developing and refining models and structural representations that capture relevant relations and patterns. On this view (associated with figures like Frederick Suppe and Nancy Cartwright):

• Models are central: Scientists construct idealized, often simplified models that represent target systems. These models highlight structural features (relations, equations, causal patterns) while deliberately omitting or distorting other details to make systems tractable and to reveal explanatorily relevant relations.

• Structure over literal truth: What matters is the model’s structural fit to phenomena—its capacity to map relations and generate reliable predictions or interventions—rather than its being a literally true, complete statement about underlying reality. A model can be empirically successful even if many of its assumptions are false (idealizations, approximations).

• Plurality and context-sensitivity: Multiple models can coexist for the same phenomenon, each capturing different aspects or operating at different scales. Choice of model depends on explanatory aims, experimental context, and pragmatic criteria, not solely on convergence to a single true theory.

• Emphasis on practice and representation: Structuralism pays attention to the representational tools (mathematical structures, diagrams, simulations) and how they are used in practice to organize data, guide experiments, and support inference.

Key proponents and ideas:

• Frederick Suppe emphasized the role of structural models and theoretical structures in the semantics of scientific theories (see Suppe, The Structure of Scientific Theories).
• Nancy Cartwright argued that “the laws of physics lie” in the sense that true universal laws rarely hold without idealized ceteris paribus conditions; reliable knowledge arises from models and localized capacities (see Cartwright, How the Laws of Physics Lie).

In short, this view reframes scientific knowledge as model‑based and structurally organized: science advances by crafting and revising representations that reliably capture relations and produce useful, sometimes limited, truths about the world.

References: Frederick Suppe, The Structure of Scientific Theories (1977); Nancy Cartwright, How the Laws of Physics Lie (1983).

Models, simulations, and big data serve complementary epistemic roles in contemporary science by aiding representation, inference, and prediction.

• Models: Models are simplified, often idealized representations of phenomena that isolate relevant factors and relations. They make assumptions explicit, provide causal or structural hypotheses, and function as tools for explanation and understanding rather than literal replicas of reality. Philosophers emphasize their mediating role between theory and the world (e.g., Frigg & Hartmann 2020).

• Simulations: Simulations are computational implementations of models that explore system behavior over time or under varying conditions. They allow scientists to investigate complex, nonlinear, or multiscale dynamics that resist analytic solutions. Simulations produce data contingent on both model structure and numerical methods, so their epistemic authority depends on verification (correct implementation) and validation (empirical adequacy) (Winsberg 2010).

• Big data: Big data refers to large, heterogeneous datasets produced by high-throughput instruments, sensors, or digital traces. It shifts some emphasis from theory-driven hypothesis testing to pattern discovery, correlation detection, and algorithmic prediction. Big data can reveal previously inaccessible regularities, but without models it risks spurious correlations, selection biases, and opaque inference (Cioffi-Revilla 2013; Kitchin 2014).

Interactions and philosophical issues:

• Complementarity: Models and simulations structure interpretation of big data; big data can inform model building, calibration, and falsification.
• Epistemic opacity: Simulations and machine-learning tools can be hard to interpret, raising questions about understanding versus mere prediction (Hutchins & Evans; also Dennett’s concerns about cognitive opacity).
• Validation and trust: Reliance on large datasets and complex simulations requires robust methods for assessing reliability—sensitivity analyses, cross-validation, and identification of causal mechanisms.
• Values and normative concerns: Choices about data collection, model assumptions, and evaluation criteria embed value judgments that affect scientific outcomes.

Key references:

• Frigg, R., & Hartmann, S. (2020). Models in Science. Stanford Encyclopedia of Philosophy.
• Winsberg, E. (2010). Science in the Age of Computer Simulation.
• Kitchin, R. (2014). The Data Revolution.

The philosophy of science investigates foundational issues about scientific practice, aims, and knowledge. Central questions include:

• What is science? — How do we delineate science from non-science or pseudoscience (the demarcation problem)? Thinkers: Karl Popper (falsifiability), Thomas Kuhn (paradigms), Imre Lakatos (research programmes).

• How is scientific knowledge justified? — What counts as evidence, and how do observations and experiments support theories (confirmation, induction, Bayesianism)? Key concerns: induction problem (Hume), theory-ladenness of observation.

• What is the structure and status of scientific theories? — Are theories true, approximately true, instrumental useful fictions, or models? Debates: scientific realism vs. anti‑realism (van Fraassen, Putnam).

• How do scientific explanations work? — What makes an explanation satisfactory (causal, unification, mechanistic accounts)? Notable accounts: Hempel’s covering-law model, Woodward’s interventionist account, mechanistic explanations in biology.

• How do values and social factors influence science? — To what extent do social, ethical, and political values shape research questions, methods, and acceptance of results (social epistemology, feminist critiques)?

• How does science change over time? — What drives scientific revolutions, continuity, or progress? Key ideas: Kuhn’s paradigm shifts, Lakatos’s progressive vs. degenerative change.

• What is the role of models, idealizations, and instruments? — How do simplified or intentionally false models produce reliable knowledge, and what are the limits of measurement?

These questions guide analysis of how science produces reliable knowledge, its limits, and its role in society. For further reading: Karl Popper, The Logic of Scientific Discovery; Thomas Kuhn, The Structure of Scientific Revolutions; Bas van Fraassen, The Scientific Image; Peter Godfrey‑Smith, Theory and Reality.

Bayesianism is a framework in the philosophy of science and epistemology that treats degrees of belief as probabilities. Central to it is Bayes’ theorem, which prescribes how to update prior beliefs in light of new evidence to form posterior beliefs. Formally:

• Start with a prior probability for a hypothesis H, P(H), reflecting current credence before new data.
• When observing evidence E, use the likelihood P(E|H) — the probability of observing E if H is true.
• Update according to Bayes’ theorem: P(H|E) = P(E|H) P(H) / P(E). The denominator P(E) normalizes across all competing hypotheses.

Key points:

• Coherence: Beliefs should obey the probability calculus to avoid Dutch-book (sure-loss) vulnerabilities.
• Confirmation and learning: Evidence increases the probability of hypotheses that make the evidence more expected (higher likelihood).
• Prior dependence: Conclusions can depend on priors; rational debate often centers on how to choose or justify priors (objective vs. subjective approaches).
• Applications: Bayesian methods underpin statistical inference, model comparison, and scientific confirmation, offering a unified account of learning from data.

References: Thomas Bayes’ original idea; Bruno de Finetti on subjective probability; Harold Jeffreys and E.T. Jaynes for objective/epistemic Bayesianism; general treatments in Howson & Urbach, Scientific Reasoning (2006) and Jeffrey (2004).

Bas van Fraassen’s The Scientific Image develops and defends “constructive empiricism,” a position about the aims and interpretation of science. Van Fraassen agrees that science seeks theories that are empirically adequate — that is, they correctly account for observable phenomena — but denies that the aim of science is to give true descriptions of unobservable entities or mechanisms. On his view, acceptance of a theory means accepting that it is empirically adequate, not that its claims about unobservables are literally true.

Key points

• Empirical adequacy: A theory is acceptable when it correctly saves the observable phenomena; agreement with unobservables is not required.
• Anti‑realist stance: Van Fraassen opposes scientific realism (the view that science aims at—and often attains—truth about both observables and unobservables). He treats belief in unobservables as optional or pragmatic, not mandated by scientific practice.
• Theory acceptance vs. belief: One can accept a theory for use in explanation and prediction without committing metaphysically to the literal existence of theoretical entities.
• Constructive element: His account is “constructive” because it prescribes how scientists should relate to theories (accept empirically adequate ones) rather than merely describing what they in fact do.
• Critiques and influence: The book sparked extensive debate over realism, underdetermination, and the observability distinction. Critics challenge the observability/unobservability binary and argue that empirical success plausibly supports belief in unobservables (see Putnam, Boyd). Supporters develop more nuanced empiricist accounts.

Further reading

• van Fraassen, B. (1980). The Scientific Image. Oxford University Press.
• Stanford, P. K. (2006). Exceeding Our Grasp: Science, History, and the Problem of Unconceived Alternatives — critique of inference to the best explanation.
• Psillos, S. (1999). Scientific Realism: How Science Tracks Truth — a defense of realism responding to van Fraassen.

(Concise summary aimed at capturing the book’s central thesis and its significance in philosophy of science.)

The demarcation problem asks how to tell science apart from non‑science (including pseudoscience, metaphysics, and everyday theorizing). There is no single definitive test, but philosophers have proposed criteria that capture core features of scientific practice. Key approaches:

• Falsifiability (Karl Popper): A scientific theory must be testable and risk refutation by empirical observation. If no possible observation could contradict it, the claim is not scientific. Strength: emphasizes empirical testability. Limitation: many scientific theories are not straightforwardly falsified in isolation (Duhem–Quine problem) and some accepted theories were not immediately falsifiable.

• Empirical adequacy and evidence (Logical empiricists): Science builds on observation and aims for theories that systematize and predict observable phenomena. Emphasis on confirmation, measurement, and verifiability. Limitation: strict verification is unattainable (observations are theory‑laden).

• Methodological naturalism: Science seeks naturalistic explanations and uses methods (controlled observation, experimentation, mathematization, reproducibility) that allow intersubjective testing and correction. Strength: captures practice; limitation: does not sharply exclude some non‑scientific but empirical disciplines.

• Progressive vs. degenerative research programs (Imre Lakatos): Scientific programs are those that generate novel predictions and progressive problem‑solutions; pseudoscience tends to be degenerative, patching to accommodate anomalies without predictive gain.

• Scientific realism vs. instrumentalism: Debates about whether science aims at true descriptions of unobservable reality or merely empirically adequate instruments. This is less about demarcation than about interpretation of scientific success.

Practical criteria commonly used (pluralistic and fallible):

• Testability and empirical content (predictions open to observation).
• Reproducibility and methodological transparency.
• Use of controlled experiment or systematic observation.
• Openness to revision in light of evidence; willingness to abandon or modify hypotheses.
• Explanatory coherence and unifying power, relative to alternatives.

Conclusion: Demarcation is best treated not as a single silver‑bullet criterion but as a cluster of methodological and epistemic norms—testability, empirical grounding, reproducibility, and critical revision—that together distinguish scientific inquiry from non‑science. For further reading: Popper, The Logic of Scientific Discovery; Kuhn, The Structure of Scientific Revolutions; Lakatos, “Falsification and the Methodology of Scientific Research Programmes”; Hacking, Representing and Intervening.

Methodology in the philosophy of science refers to the systematic rules, procedures, and strategies scientists use to generate, test, and evaluate knowledge. It addresses questions such as which methods (experiments, observation, modeling, statistics) are appropriate for different problems, how hypotheses are chosen and structured, and what counts as good evidence. Competing methodological positions include inductivism (generalizing from observed cases), hypothetico-deductive methods (deriving testable predictions from hypotheses), Bayesianism (updating credences via probabilities), and methodological pluralism (using multiple complementary methods). Methodology also covers standards like reproducibility, control of bias, and the role of peer review.

Reasoning refers to the forms of inference scientists employ to move from data and theory to conclusions. Key types include:

• Induction: inferring general laws from specific observations (e.g., generalizing from repeated measurements). Its problem is the logical gap from finite observations to universal claims (Hume).
• Deduction: deriving specific predictions from general theories; if premises are true, conclusions follow necessarily (e.g., testing predictions from a physical theory).
• Abduction (inference to the best explanation): selecting the hypothesis that best explains the evidence; common in theory choice when multiple explanations fit data.
• Probabilistic reasoning: assessing how evidence changes confidence in hypotheses (Bayesian updating).

Philosophical issues tying methodology and reasoning together include the underdetermination of theory by data (multiple theories can fit the same evidence), theory-ladenness of observation (observations are influenced by prior theory), and the problem of induction. Debates focus on whether there is a single scientific method or a plurality tailored to context, and how normative methodological rules can be justified.

Further reading: Karl Popper, The Logic of Scientific Discovery; Thomas Kuhn, The Structure of Scientific Revolutions; Imre Lakatos, “Falsification and the Methodology of Scientific Research Programmes”; Bas van Fraassen, The Scientific Image.

Explanation

• Scientific explanation answers why or how a phenomenon occurs by citing causes, mechanisms, laws, or unifying principles that make the phenomenon intelligible. Prominent models:
  • Deductive-Nomological (D-N) model (Hempel & Oppenheim): an explanation shows that the event was to be expected by subsuming it under general laws plus particular antecedent conditions — the explanans logically entail the explanandum.
  • Causal/Mechanistic accounts: explanations appeal to causal relations or mechanisms—how parts and processes produce the effect—rather than mere logical deduction.
  • Unificationist approach (Friedman, Kitcher): explanations work by reducing the number of independent phenomena using general patterns or principles, thereby increasing understanding.
• Good explanations are accurate, informative, empirically testable, and ideally provide understanding rather than mere description. Explanations can be contrastive (why P rather than Q) and context-dependent.

Laws

• Laws of nature are generalizations or regularities that characterize patterns in nature. They play several roles in philosophy of science:
  • Descriptive role: compactly summarize observed regularities.
  • Explanatory role: serve as premises in D-N explanations that show why particular events occur.
  • Predictive role: allow us to forecast future occurrences under specified conditions.
• Debates about the metaphysical status of laws:
  • Humean regularism: laws are summaries of regularities (best-system account) — they are descriptive and derivative.
  • Nomic realism (non-Humean): laws are metaphysically fundamental, governing or constraining events (e.g., Maudlin).
  • Dispositionalist views: laws reflect dispositions or powers of entities rather than external governing rules.
• Connection between laws and explanation: In D-N-style explanations, laws are essential premises. Mechanistic and causal accounts may rely less on strict universal laws and more on causal relations and mechanisms, though laws often guide or constrain mechanisms.

Further reading (concise)

• Carl G. Hempel, “Aspects of Scientific Explanation” (1965)
• Philip Kitcher, “Explanatory Unification” (1981)
• David Lewis, “Counterfactuals” (1973) and the Humean Best System account
• Tim Maudlin, “The Metaphysics within Physics” (2012) for non-Humean views

Scientific realism and anti-realism offer opposing views about what scientific theories tell us about the world.

• Scientific realism: Claims that mature scientific theories aim to give true (or approximately true) descriptions of both observable and unobservable aspects of the world. Under realism, successful theories are taken to track real entities, structures, and causal mechanisms (e.g., electrons, genes, gravitational fields). The usual arguments for realism include the “no miracles” argument: the success and predictive power of science would be a miracle unless theories are at least approximately true (Putnam 1975; Boyd 1984). Realists also appeal to theory continuity: many successful past theories were approximately true in important respects, suggesting current successful theories are too.

• Scientific anti-realism: Denies that we should accept theoretical claims about unobservables as literally true. Varieties include instrumentalism (theories are tools for prediction, not descriptions), constructive empiricism (van Fraassen: science aims to produce empirically adequate theories—correct about observables—without commitment to unobservables), and positivist or pragmatist readings. Anti-realists stress past theory change (pessimistic meta-induction): many once-successful theories were later discarded, so success does not guarantee truth. They also emphasize underdetermination: the idea that multiple, empirically equivalent theories can fit the same data, so theory choice does not determine truth about unobservables.

Trade-offs and middle positions:

• Structural realism: Attempts a compromise by claiming we can know the structure or relations described by theories even if not the nature of unobservable entities (Worrall 1989).
• Selective realism: Accepts that some parts of theories (e.g., core mechanisms) are likely true while other parts are revision-prone.
• Pragmatic or pluralist views: Emphasize the role of models, practices, and instruments rather than grand metaphysical claims.

Key considerations:

• Empirical success vs. historical turnover
• Observability distinctions (how to draw the line)
• The role of explanation, prediction, and intervention in justifying belief

References (selection):

• Bas van Fraassen, The Scientific Image (1980).
• Hilary Putnam, “What Is Realism?” (1975).
• Stathis Psillos, Scientific Realism: How Science Tracks Truth (1999).
• James Ladyman & Don Ross, Every Thing Must Go: Metaphysics Naturalized (2007).
• Richard Boyd, “Scientific Realism” (1984).
• W. V. O. Quine and the problem of underdetermination; Worrall on structural realism (1989).

In brief: realism affirms that science uncovers truth about a mind-independent world (including unobservables); anti-realism treats theories primarily as instruments for organizing and predicting observations, withholding metaphysical commitment to unobservables.

Instrumentalism (a form of scientific anti‑realism, as defended by Bas van Fraassen) holds that scientific theories are primarily instruments for organizing observations and generating reliable predictions, not literal or final descriptions of an unobservable reality. According to this view:

• The epistemic aim of science is empirical adequacy — a theory is acceptable if it correctly accounts for observable phenomena — rather than truth about unobservable entities (electrons, quarks, fields).
• Theoretical terms and models are useful fictions or calculational devices: we should use them when they work to predict and control experience, but remain agnostic about whether they correspond to how the world “really is.”
• Acceptance of a theory, for the instrumentalist, does not require belief in the existence of its unobservable posits; it requires only confidence in the theory’s empirical success.

Key motivation: History shows recurring theory change (e.g., phlogiston → oxygen, Newtonian → relativistic mechanics). Instrumentalists argue that because successful past theories were later rejected, we have reason to avoid metaphysical commitment to current theories’ unobservables.

Reference: Bas van Fraassen, The Scientific Image (1980) — central defense of constructive empiricism, a prominent form of instrumentalism/anti‑realism.

Responsible research rests on three interlocking commitments:

• Transparency: Researchers should disclose methods, data, assumptions, and conflicts of interest so others can evaluate how results were produced. Transparency enables scrutiny, reduces biases, and builds public trust. (See: Stodden et al., “Toward Reproducible Computational Research,” 2013.)

• Reproducibility: Findings should be replicable by independent investigators using the same or similar methods and data. Reproducibility tests the reliability of results and helps distinguish robust knowledge from artefacts of chance, error, or selective reporting. (See: National Academies of Sciences, “Reproducibility and Replicability in Science,” 2019.)

• Ethical reflection: Researchers must consider broader moral implications—risks to participants, societal impacts, dual-use concerns, and justice in who benefits or is harmed. Ethical reflection guides choices about study design, dissemination, and application, preventing harm and aligning research with social values. (See: Resnik, “The Ethics of Science,” 1998; UNESCO Recommendation on Science and Scientific Researchers, 2017.)

Together, these elements ensure scientific practices produce reliable knowledge while respecting persons and society. Lack of any one undermines credibility: opacity hides errors or bias, irreproducible results waste resources and mislead, and absent ethical reflection can produce harmful or unjust applications.

Causal or mechanistic accounts of explanation hold that good scientific explanations identify the causes or the underlying mechanisms that produce the phenomenon to be explained. Rather than merely describing regularities or fitting data to laws, these accounts insist on specifying how parts, activities, and organizations—often across multiple levels—interact to bring about the effect. Two influential formulations:

• Wesley Salmon’s causal-relevance approach emphasizes causal processes and interactions: explanations pick out the relevant causal chains and processes that transmit conserved quantities (like energy or momentum) to produce the phenomenon. Explanatory power lies in showing how events are embedded in causal networks. (See Salmon, Scientific Explanation and the Causal Structure of the World.)

• Stuart Glennan and William Bechtel, and especially Peter K. Craver, articulate mechanistic explanation: a mechanism is a structure of entities and activities organized such that they produce or realize the phenomenon. Explaining is decomposing the system into components, detailing their operations and interactions, and situating those operations in a context (including higher- and lower-level mechanisms). Craver emphasizes how parts and their organization constitute the mechanism that answers “how-possibly” and “how-actually” questions. (See Craver, Explaining the Brain.)

Why this matters: Causal/mechanistic accounts connect explanation to intervention and prediction—if you know the mechanism, you can manipulate it and expect outcomes. They also handle explanations in special sciences (biology, neuroscience, engineering) where laws are sparse but mechanisms are central.

Further reading: Salmon 1984; Craver 2007, “Explaining the Brain.”

Constructivism and social epistemology hold that scientific knowledge is not produced in isolation by lone observers discovering neutral facts, but is shaped fundamentally by social processes, values, and institutions. Key claims:

• Knowledge as constructed: Scientific facts, theories, and classifications arise through human practices—experiments, instruments, language, and interpretive frameworks—rather than being simple mirrors of an independent nature. What counts as a convincing observation or a legitimate theory depends on shared methods and background assumptions (Kuhn: paradigms; Latour: laboratory practices and network-building).

• Role of communities and institutions: Scientific consensus is achieved through social mechanisms—peer review, training, funding, institutional norms, conferences—so scientific change often reflects shifts in community standards, interests, or power relations as much as empirical recalcitrance. Kuhn’s “scientific revolutions” emphasize paradigm shifts in which a community adopts new standards; Latour analyzes how networks of actors and instruments stabilize facts.

• Values and interests matter: Non-epistemic values (social, political, economic) can influence what questions are asked, which methods are privileged, and which results are pursued or accepted. Constructivists warn that scientific practice is value-laden and context-sensitive.

• Not radical relativism: Most contemporary constructivists and social epistemologists do not deny that empirical constraints exist; rather they stress that empirical data are theory-laden and that acceptance of claims depends on social validation. Science’s authority comes from the robustness of social processes that produce reliable knowledge, not from a purely objective, context-free gaze.

Key references: Thomas Kuhn, The Structure of Scientific Revolutions (1962); Bruno Latour and Steve Woolgar, Laboratory Life (1979); Helen Longino, Science as Social Knowledge (1990).

Karl Popper proposed that what separates scientific theories from non-scientific ones is falsifiability: a scientific theory must make risky, testable predictions that could in principle be shown false. Rather than seeking verification through confirming instances, science advances by proposing bold conjectures and then attempting rigorous attempts to refute them. A theory survives only provisionally—its status improves the more ways it has withstood serious attempts at falsification, but it is never finally proven true.

Key points

• Falsifiability criterion: A theory is scientific if it rules out possible observational outcomes (i.e., it can be empirically refuted).
• Emphasis on boldness: Good scientific theories are risky—they predict novel, improbable phenomena that, if observed, strongly support the theory; if not observed, they expose the theory to refutation.
• Conjectures and refutations: Progress occurs through a cycle of proposing hypotheses and critically testing them; failed tests lead to rejection or revision.
• Demarcation and critique: Popper used falsificationism to demarcate science from pseudoscience (e.g., he criticized astrology and psychoanalysis for being unfalsifiable).

Limitations (brief)

• Auxiliary hypotheses: Failures can be blamed on background assumptions rather than the core theory (Duhem–Quine problem).
• Historical practice: Scientists often retain theories despite anomalies and modify them rather than immediately discarding them (see Kuhn).
• Some valuable theories are probabilistic or model-based and resist simple binary falsification.

Further reading

• Popper, K. R. The Logic of Scientific Discovery (1959).
• Duhem, P., and Quine, W. V. O. on theory underdetermination; Kuhn, T. S., The Structure of Scientific Revolutions (for contrasting views).

Karl Popper proposed falsificationism as a criterion to demarcate science from non‑science: a theory is scientific only if it makes risky, testable predictions that could in principle be shown false. Rather than seeking to verify universal claims by accumulating confirming instances (the classic problem of induction), Popper insisted science advances by bold conjectures that survive attempts at refutation; theories that are repeatedly falsified are rejected or revised.

Popper used this standard to criticize disciplines he saw as unfalsifiable. He argued that astrology often makes vague, ad hoc predictions that accommodate any outcome, and that psychoanalytic theories (as practiced in his time) were insulated from disconfirmation because failures were reinterpreted as confirming evidence. For Popper, such theories lack genuine empirical content and so fall outside the domain of science—they are pseudoscientific.

Critiques of Popper’s demarcation include:

• Some genuine sciences use practices (like model adjustment, auxiliary hypotheses) that make strict falsification difficult (Duhem–Quine problem).
• Historical cases show that scientific theories sometimes survive apparent falsifications and are later vindicated (Kuhn, Lakatos).
• Popper’s criterion can exclude legitimate theoretical work (e.g., certain evolutionary or cosmological hypotheses) that are presently hard to test but scientifically valuable.

References: Popper, The Logic of Scientific Discovery (1959); Duhem and Quine on underdetermination; Kuhn, The Structure of Scientific Revolutions (1962).

When a scientific prediction fails, it rarely tests a lone theory in isolation. Scientific assertions are embedded in a web of background assumptions and auxiliary hypotheses—about experimental setup, instrument calibration, initial conditions, or other supporting laws. The Duhem–Quine problem points out that because any empirical test combines the core theory with these auxiliaries, a failed prediction underdetermines which component is at fault.

Two key implications:

• Underdetermination of theory refutation: You can often preserve a favored theory by revising an auxiliary hypothesis (e.g., “the detector was faulty” or “we neglected friction”), so a single failed test does not uniquely refute the core theory.
• Rational theory change is holistic: Empirical evidence confronts the whole network of belief; decisions about which part to adjust rely on scientific judgment, background commitments, and pragmatic factors (simplicity, scope, coherence), not a mechanical rule.

Examples:

• In 19th-century celestial mechanics, anomalous planetary motions led to either adjusting Newtonian theory or positing new bodies (the discovery of Neptune) — an instance of choosing which auxiliary to amend.
• In particle physics, unexpected detector signals may be attributed to background noise, calibration error, or new particles; deciding between these requires broader theoretical and experimental context.

References:

• Pierre Duhem, The Aim and Structure of Physical Theory (1914)
• Willard Van Orman Quine, “Two Dogmas of Empiricism” (1951)

Many scientific theories are not packaged as single, universally true/false claims but as probabilistic statements or collections of interlocking models. For example, a theory might say “smoking increases the probability of lung cancer,” or provide a family of models that approximate complex systems (climate models, epidemic models, particle-physics simulations). Such theories have three features that make simple binary falsification (true vs. false) inappropriate:

• Probabilistic claims admit degrees. A probabilistic theory does not predict a single outcome; it assigns probabilities to outcomes. A low-probability event (e.g., a heavy smoker who never develops cancer) does not falsify the theory, because the theory allows such outcomes. Disconfirmation is a matter of shifting credences, not an immediate refutation.

• Models are idealized and partial. Models simplify, idealize, or isolate mechanisms (neglecting friction, treating agents as rational, averaging over small-scale details). When a model’s predictions fail in some conditions, it may be because the idealizations break down, not because the whole theoretical framework is false. Scientists often revise assumptions, change parameter values, or restrict the domain of applicability rather than discard the theory outright.

• Complex systems and underdetermination. Many model-based theories involve adjustable parameters, auxiliary hypotheses, and measurement uncertainties. Empirical anomalies can often be accommodated by modifying auxiliary assumptions (the “protective belt” in Lakatos’s terms), making a single failed prediction insufficient to falsify the core theory.

Implications

• Evaluation is comparative and evidential: scientists compare model fits, predictive accuracy, and explanatory scope, often using statistical inference or Bayesian updating rather than single-shot refutation.
• Methodological pluralism: different models serve different purposes (prediction, explanation, policy guidance), so success is measured by utility, robustness, and domain-specific adequacy, not only by passing a binary falsification test.

References: Karl Popper (falsificationism) for the contrast; Imre Lakatos (research programmes) and Bas van Fraassen (confirmation and empiricism) for nuances about theory change; Nancy Cartwright and Margaret Morrison for work on models and idealization.

Duhem and Quine: theory underdetermination

• Pierre Duhem: In physics, Duhem argued that experiments never test a single hypothesis in isolation but rather a whole web of background assumptions and auxiliary hypotheses. When an experiment conflicts with prediction, you can adjust various parts of that web (instrument calibration, auxiliary assumptions, or the theory itself). Thus empirical data underdetermine which component to revise. (See Duhem, The Aim and Structure of Physical Theory.)
• W. V. O. Quine: Quine generalized Duhem’s point into the “Quine-Duhem thesis.” He argued that our scientific beliefs form a holistic network—from logical truths to particular observations—and any statement can be maintained by adjusting other parts of the network. Quine also challenged the analytic–synthetic distinction and emphasized that observation cannot fix theory uniquely; choices among rival theories involve pragmatic and holistic considerations. (See Quine, “Two Dogmas of Empiricism” and “On What There Is.”)

Core consequence (shared): Empirical evidence underdetermines theory choice; theories are revisable in multiple ways, so observation alone cannot definitively decide between competing theoretical frameworks.

Kuhn: the structure and dynamics of scientific change

• Thomas Kuhn offered a different, historically grounded picture in The Structure of Scientific Revolutions. He distinguished “normal science” (puzzle-solving within a dominant paradigm) from “revolutionary science” (paradigm shifts when anomalies accumulate). For Kuhn, paradigms are comprehensive frameworks—methods, exemplars, standards—that shape what counts as a legitimate question and evidence.
• Kuhn’s view complements underdetermination: paradigms are incommensurable to varying degrees, so proponents of different paradigms may talk past one another and evaluate evidence by different standards. Theory choice is influenced by pragmatic, aesthetic, and community-based factors (simplicity, scope, fertility), not by deductive logic alone.

Contrast and synthesis

• Scope and emphasis: Duhem and Quine focus on epistemic holism and the logical problem that data cannot uniquely determine theory. Kuhn focuses on historical, sociological, and normative dimensions—how scientific communities actually operate and shift between frameworks.
• Decision factors: Duhem–Quine highlight that background assumptions can be adjusted arbitrarily; Kuhn adds that choices are governed by paradigm-dependent standards and community consensus (so sociological and normative factors shape theory change).
• Compatibility: The views are compatible and complementary. Underdetermination describes a logical feature of theory testing; Kuhn supplies a descriptive account of how scientists respond to that feature in practice—often conservatively during normal science and collectively during revolutions.

Further reading (short)

• Duhem, P., The Aim and Structure of Physical Theory (1914)
• Quine, W. V. O., “Two Dogmas of Empiricism” (1951)
• Kuhn, T. S., The Structure of Scientific Revolutions (1962)

If you’d like, I can give a brief example (e.g., Ptolemy vs. Copernicus, or 19th‑century ether debates) that illustrates these points.

Karl R. Popper proposes falsifiability as the hallmark of scientific theories: a theory is scientific only if it makes risky predictions that could, in principle, be shown false by observation. He rejects induction as a logical justification of science (the idea that repeated observations can prove a universal law). Instead Popper views science as conjectures and refutations: scientists advance bold hypotheses and then attempt to refute them; surviving theories are corroborated but never finally verified. Popper also emphasizes methodological rigor—preferring risky tests to ad hoc adjustments—and criticizes historicist and totalizing philosophies that claim inevitable scientific progress. His view reshaped debates about demarcation, scientific method, and the rational appraisal of theories.

Key ideas to note

• Falsifiability over verifiability: empirical content comes from what could falsify a theory.
• Conjectures and refutations: knowledge grows through critical testing, not accumulation of confirmed instances.
• Corroboration ≠ truth: surviving tests increase a theory’s empirical standing but do not prove it true.
• Criticisms: neglects the role of auxiliary hypotheses, theory-ladenness of observation, and the complex practice of science (responses developed by Lakatos, Kuhn, and others).

Further reading

• Popper, K. R., The Logic of Scientific Discovery (1959)
• For critiques/alternatives: Kuhn’s Structure of Scientific Revolutions; Lakatos’ Methodology of Scientific Research Programmes.

Karl Popper’s idea of conjectures and refutations sees scientific progress as a cycle: scientists propose bold hypotheses (conjectures) and then subject them to rigorous tests aimed at falsifying them (refutations). A theory never achieves final verification; instead, it survives only as long as it withstands attempts to falsify it. Failed tests show where a conjecture is false or incomplete, prompting its rejection or revision and the formulation of improved hypotheses. This approach emphasizes critical testing, clear empirical risk (falsifiability), and the asymmetry between verification and falsification—no amount of positive instances can conclusively prove a universal law, but a single decisive counterinstance can refute it. Key source: Karl Popper, The Logic of Scientific Discovery (1959).

Historically, scientists typically do not abandon a theory at the first sign of anomaly. Instead they attempt to resolve discrepancies by adjusting auxiliary hypotheses, refining measurements, or extending the theory’s scope. Kuhn emphasized that during “normal science” researchers work within an accepted paradigm—solving puzzles, improving techniques, and protecting the core theory. Anomalies accumulate, but only when they become severe or when an alternative paradigm offers a better problem-solving framework does a scientific revolution—paradigm shift—occur. This practice reflects pragmatic values (stability, continuity, cumulative knowledge) and shows that theory choice is guided by more than immediate empirical falsification: theoretical virtues, available alternatives, and community judgment all matter.

Further reading: Thomas Kuhn, The Structure of Scientific Revolutions (1962).

Good scientific theories are “bold” because they make risky, novel predictions—claims that would be unlikely to be true by accident or under rival hypotheses. A theory that merely accommodates known facts is weak: almost any pattern can be fitted after the fact. By contrast, a bold theory stakes itself on predictions about phenomena not yet observed (or about specific parameter values, unexpected correlations, etc.). If such improbable predictions come true, the theory gains strong support because the coincidence of success is unlikely under alternatives. If the predictions fail, the theory is exposed to decisive refutation.

This idea underlies Popper’s falsificationist emphasis on conjectures and refutations: scientific progress requires daring hypotheses that can be tested and possibly falsified. It also explains why novel predictions (e.g., the existence of Neptune, the bending of light around the sun, or the neutrino) historically boosted confidence in the theories that anticipated them. Boldness thus serves both an epistemic function (providing severe tests that increase confirmatory power when passed) and a methodological one (driving theory development and empirical investigation).

References: Karl Popper, The Logic of Scientific Discovery; discussions of “novel predictions” in scientific confirmation (e.g., in Godfrey‑Smith, Theory and Reality).

Falsifiability (Popper): A theory counts as scientific only if it makes risky predictions that rule out some possible observational outcomes—i.e., there must be conceivable evidence that would show the theory to be false. The core idea is not that theories are conclusively proven true, but that scientific knowledge advances by bold conjectures subjected to attempts at refutation. A theory that accommodates every possible observation (is compatible with all outcomes) is untestable and therefore non-scientific by this criterion.

Why it matters

• Emphasizes empirical testability and accountability to observation.
• Encourages theories that expose themselves to potential disconfirmation rather than ad hoc adjustments.
• Clarifies the asymmetry between verification (hard to achieve) and refutation (observable in principle).

Limitations and responses

• Some genuine scientific claims (e.g., complex theories or those with auxiliary hypotheses) are difficult to test in isolation—leading to the Duhem-Quine problem: empirical failure can implicate background assumptions as well as the core theory.
• Popper’s view downplays confirmation and probabilistic reasoning; Bayesian and other approaches supplement or revise falsificationism.
• Historical science and model-based work show that researchers often judge theories by a mix of empirical tests, coherence, predictive novel results, and explanatory power.

For further reading: Karl Popper, The Logic of Scientific Discovery (1959); discussion in Peter Godfrey-Smith, Theory and Reality (2003).

Observation and theory interact bidirectionally: theories shape what we observe, and observations test and revise theories. The core claim of “theory-ladenness” is that what we take to be an observation is influenced by background beliefs, conceptual frameworks, experimental design, and instruments. Three main points:

• Perceptual interpretation: Raw sensory data are organized by concepts. Theories supply the categories and expectations that make sensations intelligible. For example, seeing “a planet” instead of “a wandering star” depends on astronomical theory (Kuhn 1962; Hanson 1958).

• Experimental mediation: Instruments and procedures are theory-laden. Choices about which instruments to build, how to calibrate them, and how to interpret outputs presuppose theoretical commitments (Hacking 1983). A thermometer reading becomes meaningful only within assumptions about what it measures.

• Underdetermination and revision: Because observations are interpreted through theory, empirical data underdetermine theory choice—different theoretical frameworks can accommodate the same observations. Conversely, persistent anomalous observations can prompt revision or replacement of theory (Duhem 1906; Laudan 1977).

Implications: Theory-ladenness challenges a naïve view of neutral, theory-free observation but does not imply total subjectivity. Scientists can converge on reliable observations through shared methods, intersubjective checks, and instrument calibration, even while acknowledging theory’s guiding role.

Suggested readings: Norwood Russell Hanson, Patterns of Discovery (1958); Thomas Kuhn, The Structure of Scientific Revolutions (1962); Ian Hacking, The Social Construction of What? (1999) and Representing and Intervening (1983); Pierre Duhem, The Aim and Structure of Physical Theory (1914/1954).

Induction is the reasoning method by which we infer general laws or future occurrences from a limited set of past observations (e.g., seeing many white swans and concluding “all swans are white” or expecting the sun to rise tomorrow because it has always risen). It underpins most empirical science and everyday prediction.

David Hume’s problem of induction: Hume argued that there is no non-circular rational justification for inductive inference. Two key points:

• Descriptive: We habitually form inductive beliefs because of psychological custom or habit—past conjunctions make us expect similar future conjunctions—but this is a statement about human practice, not a rational proof that induction yields truth.
• Normative: Any attempt to justify induction either (a) relies on a deduction (which can’t deliver the general-from-particular move induction requires), or (b) appeals to induction itself (e.g., “induction has worked in the past, so it will work in the future”), which is circular.

Consequences: Hume’s skeptical conclusion is that inductive reasoning lacks a rational, non-circular foundation; yet we cannot avoid using it. This problem has driven subsequent responses: pragmatic defenses, probabilistic accounts (e.g., Bayesianism), attempts to ground induction in naturalized epistemology (Quine), and claims about innate cognitive structures or justification through explanatory virtues (Peirce, Reichenbach, Popper’s falsificationism rejects induction as the basis of science).

For further reading: Hume, An Enquiry Concerning Human Understanding (Section IV); Karl Popper, The Logic of Scientific Discovery (on falsification); Wesley Salmon and C. G. Hempel on the logic of confirmation.