Incomplete summaries of Curd & Cover, _Philosophy of Science: The Central Issues_ Notes are for my purposes only; use at your own risk October-November 2004 ***** Popper, "Science: Conjectures and Refutations" Popper's topic is the problem of demarcation: What separates science from pseudo-science (those empirical studies, such as astrology, that we don't want to call science)? Popper considers theories that have influenced him: Einstein, Marx, Freud, Adler. Of these, only Einstein's is scientific because the other three can all be made to fit with any observations. Only Einstein's theory is falsifiable. Science is about falsifiability, testability, disprovability. It's about testing theories, really trying to find counterexamples, and getting surprising results which confirm the theory. ***** Kuhn, "Logic of Discovery or Psychology of Research?" Kuhn on the demarcation problem: Popper gets it wrong; it's not falsifiability but puzzle-solving which demarcates science from pseudo-science. Scientists are puzzle-solvers. When hypotheses fails, scientists and pseudo-scientists can offer explanations, but only the scientist's explanation leads to new puzzles. The discrepancies between theory and observation open up new challenges and advance the science. Errors can be meaningful, leading the way to not an overthrow of theory, but to theory revision and refinement. ***** Lakatos, "Science and Pseudoscience" Lakatos: Popper's criteria (namely, falsifiability's central role in the scientific method) is wrong because in actuality scientists are often not prepared to abandon their theories if/when they are falsified. Lakatos' criteria is a methodology of scientific research programmes. Scientific theories are elaborate research programmes consisting of much more than a few central hypotheses. Scientific research programmes (in contrast with non-scientific programmes, such as Marxism) predict novel facts. They are progressive. ***** Thagard, "Why Astrology is a Pseudoscience" Thagard, focusing on astrology as an instance of pseudo-science, rejects the demarcation criteria of verificationists, Popper, Kuhn, and Lakatos. Science and pseudo-science are distinguished by theory, community, and historical context. Pseudo-science is less progressive than alternative theories, and psuedo-scientists do not seem concerned with progress nor with evaluating their theory against others. As for historical context, Thagard notes that a single theory may be science at one point in time and pseudo-science at another. ***** Ruse, "Creation-Science is Not Science" Ruse identifies defining characteristics of science for the purpose of showing that creation-science is not science (but rather religious doctrine which has no place in public schools). According to Ruse, science is empirical and is about finding natural regularities or laws. Laws are used in explanation and prediction. Theories are testable (both confirmable and falsifiable). Science is tentative and revisable. ***** Kuhn, "The Nature and Necessity of Scientific Revolutions" Drawing on examples from the history of science, Kuhn makes a number of points about theory commitment and progress. Theory changes in science are abrupt revolutions (much like political revolutions) that occur in times of crisis when an older theory is no longer considered adequate by some number of scientists. In a scientific revolution, an older theory is completely overthrown by a newer, incompatible theory. Revolutions involve replacing methodology, values, etc., as well as theories; that is, they represent not just a change in theory but a total paradigm shift. Theory/paradigm choice cannot be settled withing a single paradigm nor by logic and experiment alone; it involves a sociological (non-rational) element. The history of science is not that of linear progression; new paradigms are not always objectively better than older paradigms; often advances in one area are given up for (the promise of) progress in another area. ***** Kuhn, "Objectivity, Value Judgment, and Theory Choice" Kuhn attempts to clarify those things from Structure which have been misunderstood by his (Popperian) critics. Kuhn thinks his view has much in common with the standard view, and is not nearly as radical as it has been perceived to be. In particular, Kuhn clarifies his brand of subjectivity. He believes that his subjectivity is not (as is often charged) a threat to science's rationality; instead, it is compatible with rational theory choice. While subjectivity plays a role in Kuhn's account, so does objectivity. Scientists share some objective values (across paradigm shifts): roughly, accuracy, consistency, scope, simplicity, and fruitfulness. These are not objective rules for theory choice; they are values, and different scientists can disagree in their judgments as to how to apply these values. When scientists apply these values differently, they may arrive at the same outcome or they may not. This subjectivity is essential to theory choice. The succession of theories/paradigms requires that scientists are not always in agreement, that some scientists remain committed to an established theories, while others are innovators. A few scientists take the risk and abandon accepted theories early on and as they show that new theories are promising, others follow. In short, Kuhn's subjectivity, far from being harmful, is both benign and beneficial to science's rationality. ***** Ernan McMullin, "Rationality and Paradigm Change in Science" After raising a few questions about objectivity under Kuhn's account, McMullin (a realist) turns his attention from Kuhn's alleged attack on the rationality of theory choice to his attack on realism in theory choice. Kuhn's (instrumentalist) values concern only predictive accuracy; he ignores the role of epistemic values. As an illustration, McMullin points to Kuhn's claim that the Copernican and Ptolemaic theories were equivalent on practical (by which Kuhn seems to mean predictive) grounds. Where Kuhn says the preference for the "more natural" Copernican account is an aesthetic preference, McMullin sees it as a practical epistemic distinction. The Copernican account had greater explanatory power (e.g., retrograde motion); it is more likely to be true, and truth plays a role in theory choice. As such, McMullin would expand the list of objective values to include coherence, fertility, and other such epistemic "superempirical" values. ***** Laudan, "Dissecting the Holist Picture of Scientific Change" Part 1: Laudan objects to Kuhn's claim that paradigms are a mix of theory, methods, and standard (or ontology, methodology, and values). According to Laudan, all three of these need not abruptly change at once (as in Kuhn's paradigm shift). Changes need not be holistic; there can be gradual piecemeal changes. Part 2: Laudan objects to Kuhn's statement of "localized" underdetermination (that is, methodological standards fail to pick out a single superior theory in cases of theory choice). Shared standards need not be ambiguous nor must they have internal tension nor must different paradigms have different standards of assessment nor must different scientists always have different senses in priorities. In short, Kuhn may have shown that methodological standards are not always sufficient to determine theory choice, but he has not shown that they can never do so. ***** Duhem, "Physical Theory and Experiment" Duhem begins with Bernard's claim that the results of an experiment must be accepted as they are without being biased by the theory that one wishes to (dis)confirm. Scientists should be healthy skeptics, as good science requires an open mind and no system is immune from refutation by experiment. Duhem notes that in physics, entirely theory-free experiments are impossible because instrumentation and measurement rely on theory. In physics, then, whole systems, not individual hypotheses, are tested; this is a holistic view. An experiment cannot disconfirm a single hypothesis; it tests and disconfirms a group of hypotheses without pointing to a single hypothesis. Decisive crucial experiments are impossible; disconfirming one system (or several systems) of hypotheses does not confirm as the set of alternate systems of hypotheses cannot be enumerated. Individual scientists must use good sense or judgment (not logic) to decide how to revise/replace a disconfirmed system; in this sense particular experiments can be crucial to particular scientists. ***** Quine, "Two Dogmas of Empiricism" The two dogmas that Quine argues against are (1) the analytic/synthetic distinction (that there is a distinction between truths which are grounded in meaning and those which are grounded in fact) and (2) reductionism (that every meaningful statement is equivalent to a statement in terms of immediate experience). Toward the first, Quine considers statements such as "No bachelor is married" and the ideas of synonymy, definitions, and interchangeability, before concluding that the notion of analyticity is ill-defined. Toward the second, Quine offers an alternate holistic account; statements about the external world are not to be taken individually, but together as a whole body. In the last section, Quine makes some important points including that no statement is immune to revision and that total science is underdetermined by experience. ***** Gillies, "The Duhem Thesis and the Quine Thesis" Gillies points out differences between Duhem and Quine's theses, assesses them, and then proposes a combined Duhem-Quine thesis. Duhem's holism is restricted to physics, whereas Quine's is about all beliefs (empirical sciences, logic, math). Duhem considers small groups of theories within physics to be empirically significant, whereas Quine's unit of meaning is the whole of science. Duhem, while maintaining that logic alone cannot cause a scientist to give up a theory, goes on to say that a combination of logic and scientific good sense can, whereas Quine does not address extra-logical considerations. ***** Carnap, "The Value of Laws: Explanation and Prediction" Driesch's entelechy (living force) theory (unlike the theory of magnetism) does not offer an explanation because it does not contain laws. Science theories can explain, using testable laws to do so. Laws of science provide explanations for observed facts and predictions for not-yet-observed facts. ***** Hempel, "Two Basic Types of Scientific Explanation" Hempel outlines two basic models of scientific explanation: the deductive-nomological form and the probabilistic-statistical form. Laws are a necessary part of both kinds of explanation. According to the D-N (aka, covering) model, explanation takes the form of a deductive argument whose premises (explanans) include circumstances and universal laws from which the explanandum can be logically deduced. The P-S model deals with explanations in which the relevant laws are statistical. In this case the premises do not logically imply the explanandum, but rather express likelihood; P-S explanations are inductive, not deductive. Many actual scientific explanations do not immediately appear fit these ideal forms because scientists (like mathematicians) omit certain propositions (which could be made explicit) or may provide only partial explanations. ***** Hempel, "The Thesis of Structural Identity" Hempel's thesis of structural identity of explanation and prediction: D-N explanation and prediction differ not in logical structure, but only in certain pragmatic respects. Every explanation (explanatory argument) of particular events is a potential prediction (predictive argument) and vice-versa. Hempel considers several objections, roughly: the question of insufficiency of causes (syphilis/paresis), narrative explanations (evolution), self-evidencing explanations (collapsing bridge), the confusion of correlation with causality (spots/measles). ***** Hempel, "Inductive-Statistical Explanation" Inductive-statistical explanations are fundamentally different from deductive-nomological explanations: I-S explanations are inductive (not deductive), express a high degree of probability (not certainty), and do not express objective truths. As an example, Hempel considers the recovery of a patient with a strep infection who is given penicillin. The I-S form can be expressed as p(R,S,P)=1-e, Sj & Pj; Likely-> Rj. I-S meets two problems of explanatory ambiguity; one is metaphysical and the other epistemological, but they are otherwise similar: additional premises (such as the particular form of strep infection) can make a different (even contradictory) conclusion follow from the explanation. Hempel's (based on Carnap's) answer to epistemic ambiguity is a requirement of maximal specificity; in forming an I-S explanation we should take into account all potentially explanatorily relevant information. Hempel's conclusion regarding non-epistemic ambiguity is epistemic relativity: I-S explanations are essentially relative to a given knowledge situation. ***** Ruben, "Arguments, Laws, and Explanation" Hempel's D-N requirements for explanation are neither sufficient nor necessary. Towards sufficiency, Ruben first addresses irrelevance counterexamples (eg., Jones eats poison which will kill him but is hit by a bus and dies first), considers adding ceteris paribus (all other things being equal) clauses, and concludes that this is not a solution (and also notes that pre-emptive causes provide prediction but not explanation). He then addresses symmetry counterexamples (eg., the length of flagpole is not explained by the length of its shadow), considers adding a causal requirement (putting the "cause" back into "because"), and decides that this doesn't address irrelevance (eg., blessed water dissolving sugar). Further, the causal requirement can render the nomological (lawlike) requirement (and the deductive requirement) unnecessary; some, but not all, explanations include laws. ***** Ayer, "What is a Law of Nature?" The central claim of the regularity approach to laws of nature is that laws are regularities which hold invariably and without exception. By itself, this is not sufficient to describe laws of nature. In particular, it does not explain how to distinguish accidental generalizations (generalizations of fact) from lawlike ones. Laws hold in all possible cases and support counterfactuals or subjunctive conditionals (if there were an X which P, then X would Q). Ayer further holds an epistemic regularity theory: the distinction between generalizations of law and generalizations of fact (All US Presidents are male) lies in the attitudes of those who believe the generalizations. Roughly (and only roughly), one's belief in a generalization of law is not destroyed by exceptions. This leads to the controversial conclusion that there are no unknown laws. ***** Dretske, "Laws of Nature" Dretske holds a necessitarian universals view on laws of nature. Although it is tempting to say that laws of nature are universal truths or universal truths plus something (confirmation, acceptance, explanatory potential, etc), Dretske maintains that this is an inaccurate account of laws of nature because it fails to meet several features of laws, including that the existence of laws does not depend on our knowledge of them, that laws imply counterfactuals, that laws are useful for prediction, that laws tell us what must happen, etc. Dretske suggests that a statement of a law is not a universal generalization "for all X, if X is F then X is G," but rather a singular statement about universals "F-ness necessarily yields G-ness". ***** Cartwright, "Do the Laws of Physics State the Facts?" Against both the regularity and necessitarian theories, Cartwright, an antirealist, challenges the claim that a generalization must be true for it to be a law (of physics). Instead, she says that there is a trade-off between factual content and explanatory power. Many laws that we use for explanation are (strictly speaking) false, and those laws which are qualified enough (limited in scope) to be true are not useful in explanations. Laws don't describe, for example, how bodies do or must behave; instead, they specify tendencies or dispositions. ***** Nagel, "Issues in the Logic of Reductive Explanations" For Nagel, the reduction of (the statements of) one scientific theory (T) to another (T') is a deductive explanation of T (and all of its observational consequences) in terms of T'. Nagel's reduction satisfies two conditions: consistency (the claims of T and T' must be logically consistent) and meaning invariance (the terms shared between T and T' must have a common meaning). Nagel distinguishes between two kinds of reduction: homogeneous reduction (such as the the reduction/explanation of Galileo's and Kepler's laws in terms of Newton's) and inhomogeneous reduction (such as the reduction of thermodynamics to the kinetic theory) in which there are some terms in T which are not in T', so bridge laws newed to be added to T' to establish a correspondence. Nagel considers Feyerabend's objection that reductions don't have the character of deductive explanation, but rather of radical replacements, and concludes that when laws are taken to be approximate, reductions are deductive. ***** Feyerabend, "How to Be a Good Empiricist" Feyerabend charges that modern empiricism, rather than eliminating dogmatic metaphysics, leads to dogmatic metaphysics in which theories themselves become part of the ontology. Instead, Feyerabend recommends a pluralism in which good empiricists work with many alternative theories. Feyerabend argues against Popper's/Nagel's account of explanation/reduction on the grounds that the conditions of both consistency and meaning invariance are inconsistent with an empiricist view of scientific change. In actual science, later theories do not subsume past theories but replace them. Further, later theories should not be judged/validated by their agreement with current theories, but by their agreement with the facts. The consistency condition is a barrier to good empiricism since it is a barrier to considering a plurality of alternate theories. ***** Nickles, "Two Concepts of Intertheoretic Reduction" Some theory succession is characterized by reduction, in which all the claims of one theory can be deduced from a more general theory. But in other cases succeeding theories contradict and replace their predecessors. Nickles' claim is that there is no single view of reduction that captures these both, but rather there are two different kinds of reduction. ***** Kitcher, "1953 and All That: A Tale of Two Sciences" Kitcher argues that classical genetics has not been (cannot be) reduced to molecular genetics, at least not if reduction is understood as the standard model of reduction. First, why the reduction account fails: Reduction requires (at least) three criteria conditions, none of which are satisfied by the relation between classical and molecular genetics. (1) Classical genetics contains general laws which can be derived from molecular genetics. Kitcher says the fundamental problem lies in the law requirement, which serves physical science well but not classical genetics. It is hard to identify laws in classical genetics and where it seems there might be laws (Mendel's), they do not seem to play an important role. Instead (following Kuhn), theories involve not just laws but also practices. (2) Bridge principles can link the vocabularies of classical and molecular genetics. Kitcher thinks this unlikely/impossible and considers the question of which segments of DNA count as genes. (3) A derivation of the consequences of classical genetics from molecular genetics would be an explanation. Kitcher (somewhat controversially) doesn't think that such a derivation would provide a relevant explanation (either because it doesn't provide an explanation at all or because it doesn't provide an enlightening or useful explanation). Second, Kitcher considers three ways in which molecular genetics connects with/contributes to classical genetics: (1) presupposition, as in providing an account of gene replication; (2) conceptual refinement, as in an account of mutation; (3) explanatory extension, as in an explanation of sickle-cell anemia. Explanation occurs at many different autonomous levels; lower-level theories can explain some consequences of higher-level theories, higher-level theories can explain lower-level theories, and explanations at one level can be independent of explanations at other levels. Kitcher concludes both that there is an important relation between classical and molecular genetics and that it is not one of reduction. ***** Maxwell, "The Ontological Status of Theoretical Entities" The primary questions of scientific realism are: Do scientific theories (attempt to) offer a true account of the world? and Do the entities postulated by theories (forces, genes, quarks) really exist? Maxwell argues against the antirealist position for (at least some) unobservable entities. He considers a fictional pre-microscope example: Jones postulates "crobes" which are useful for understanding how diseases are transmitted, but are not unobserved (so they are mere instrumental fictions for many antirealists). After the invention of the microscope, microbes are seen. Some antirealists become realists but others do not, instead holding either (1) claims about physical objects must be translatable into claims about sense data, (2) crobes were unobserved but not unobservable, or (3) observation through a microscope does not commit one to the existence of physical objects. Against these, he argues (1) sense data is problematic, (2) there is no clear separation between the observable and unobservable nor can a distinction between observed and unobserved solve the problem, and (3) if observation through a microscope is unsatisfactory, then likely so is observation through glasses or windows or anything. Maxwell concludes that there is no observation-theoretical dichotomy and any line drawn between the two has no ontological significance. ***** Van Fraassen, "Concerning Scientific Realism" Van Fraassen raises many interesting points, three of are (1) the formulation of constructive empiricism, (2) a discussion of the observable/unobservable dichotomy, and (3) a response to the miracle argument. Against scientific realism (science, via its theories, aims to give us a literally true story of what the world is like; acceptance of a theory involves the belief that it is true), van Fraassen offers the anti-realist (science need not give a literally true story; the acceptance of a theory need not involve a belief in truth) alternative of constructive empiricism: The aim of science is empirically adequate theories; acceptance involves a belief in empirical adequacy, not truth. Against Maxwell's claim that no line can be drawn between observable and unobservable objects, van Fraassen sees 'observable' as a vague predicate: although the distinction between observable and unobservable is not clear in general, some objects are clearly (un)observable. Against Putnam's miracle argument for scientific realism (realism is the only philosophy that doesn't make the success of science a miracle), van Fraassen offers an alternate Darwinist account for the success of science: Only the successful theories survive competition. ***** Brown, "Explaining the Success of Science" What explains the success of science (and its theories), and what are the ontological consequences of the explanation? For Brown, a successful theory is one that: (1) organizes/unifies a large variety of known phenomena, (2) systematizes empirical data more extensively than previous theories, and (3) has statistically significant predictive success. Van Fraassen's anti-realist/Darwinian account satisfies (1) and (2), but not (3), and also relies on an assumption that rational choice and success of theories go together. Realist accounts which rely on reference also fail, as theories (phlogiston) have been successful without reference, forcing the realist to hold the weaker position that reference is typical for success and the exceptions are miracles. Newton-Smith offers a more promising explanation of success, relying on a notion of verisimilitude (truthlikeness): a theory has greater verisimilitude if it has greater content and has greater true content than its rival. However, this satisfies (2) and (3), but not (1), does not allow for one of two false theories to be more true than the other, and is incompatible with history in that more successful theories may have more limited scope. Brown concludes that the difficulty in explaining the success of science is that Hempel's D-N model is the wrong model of explanation; instead, he advocates statistical-relevance and narrative explanations in which truth can play a role. ***** Hacking, "Experimentation and Scientific Realism" Hacking is an anti-realist about scientific theories, but a realist about some unobservable entities such as electrons. He claims that this position is held by many experimenters who are agnostic about many possible models, but are committed to the existence of unobservable entities. Experimental physics provides the strongest evidence for scientific realism about unobservables: experimenters manipulate unobservables to create new phenomena and rely on their understood causal properties to design successful instruments which are used to interfere (intervene) in other more hypothetical areas (as in Hacking's example of weak neutral forces and parity).