INTRODUCTION TO PHILOSOPHY OF SCIENCE
© Copyright 2018 Thomas J. Hickey
CHAPTER I – Overview
Both successful science and contemporary philosophy of science are pragmatic. In science, as in life, realistic pragmatism is what works successfully. This introductory book is a concise summary of the elementary principles of the contemporary pragmatist (or neopragmatist) philosophy of science, the philosophy that the twentieth century has bequeathed to the twenty-first century.
1.01 Aim of Philosophy of Science
The aim of contemporary pragmatist philosophy of science is to discover principles that explain successful practices of basic-science research, in order to advance contemporary science by application of the principles.
The principles are set forth as a metatheory. Basic science creates new language: new theories, new laws and new explanations. Applied science uses scientific explanations to change the real world, e.g., new technologies, new social policies and new therapies. Philosophy of science pertains to basic-science practices and language.
1.02 Computational Philosophy of Science
Computational philosophy of science is the design, development and application of computer systems that proceduralize and mechanize productive basic-research practices in science.
Philosophers of science can no longer be content with more rehearsing of the Popper-Kuhn debates of half a century ago, much less more debating ancient futile issues such as realism vs idealism.
Philosophy of science has taken the computational turn. Mechanized information processing has permeated almost every science, and is now belatedly intruding into philosophy of science. Today computerized discovery systems facilitate investigations in philosophy of science in a new specialty called “computational philosophy of science”.
The pragmatist philosophers Charles Sanders Peirce (1839-1914) and Norwood Russell Hanson (1924-1967) had described a nonprocedural analysis for developing theories. Some called this nonprocedural practice “abduction”, others “retroduction”. Today in computational philosophy of science procedural strategies for developing new theories are coded into computer systems.
1978 Nobel-laureate economist Herbert Alexander Simon (1916-2001), a founder of artificial intelligence, called such systems “discovery systems”. In the 1970’s Hickey (1940) called the mechanized approach “metascience”. In the 1980’s philosopher of science, Paul Thagard (1950), called it “computational philosophy of science”, a phrase that is more descriptive and therefore will probably prevail.
Mechanized simulation of successful episodes in the history of science is often used to test the plausibility of the discovery-system designs. But the proof of the pudding is in the eating: application of computer systems at the frontier of a science, where prediction is also production, in order to propose new empirically superior theories, further tests the systems. Now philosophers of science must practice what they preach by participating in basic-science research and producing contributions. Application of discovery systems gives the philosopher of science a participatory and consequential rôle in basic-science research.
1.03 Two Perspectives on Language
Philosophy of language supplies a coherent analytical framework that integrates contemporary philosophy of science. In philosophy of language philosophers have long distinguished two perspectives called “object language” and “metalanguage”.
Object language is discourse about nonlinguistic reality including domains that the particular sciences investigate as well as most of everyday discourse.
Metalanguage is language about language, either object language or metalanguage.
Much of the discourse in philosophy of science is in the metalinguistic perspective. Important metalinguistic terms include “theory”, “law”, “test design”, “observation report” and “explanation”, which are classifications of the uses of language. In the contemporary pragmatist philosophy a “Theory” is a universally quantified hypothesis proposed for empirical testing, and a “Test design” is a universally quantified discourse presumed for empirical testing a theory, in order to identify the subject of the theory independently of the theory and to describe the procedures for performing the test. The computer instructions coded in discovery systems are also metalinguistic expressions, because these systems input, process and output object language for the sciences.
1.04 Dimensions of Language
Using the metalinguistic perspective, philosophers analyze language into what Rudolf Carnap (1891-1970) called “dimensions” of language. They are syntax, semantics, ontology, and pragmatics.
Syntax refers to the structure of language. Syntax is arrangements of symbols such as linguistic ink marks on paper, which display structure. Examples of syntactical symbols include terms such as words and mathematical variables and the sentences and mathematical equations constructed with the terms.
Syntactical rules regulate construction of grammatical expressions such as sentences and equations usually arranged by concatenation into strings of terms.
Semantics refers to the meanings associated with syntactical symbols. Syntax without semantics is systematic but literally meaningless. Associating meanings to the symbols makes the syntax “semantically interpreted”.
Semantical rules describe the meanings associated with elementary syntactical symbols. In the metalinguistic perspective belief in the truth of semantically interpreted universally quantified sentences such as the heuristic stereotypic “Every raven is black” enables sentences to function as semantical rules displaying the complex meanings of the sentences’ component descriptive terms. Belief in the statement “Every raven is black” makes the phrase “black raven” redundant, thus displaying the meaning of “black” as a component part of the meaning of “raven”. The lexical entries in a unilingual dictionary exemplify an inventory of semantical rules for a language.
Ontology refers to the aspects of reality described by semantically interpreted sentences believed to be true, especially belief due to experience or systematic empirical testing. This is the thesis of ontological relativity. Ontology is typically of greater interest to philosophers than to linguists.
Semantics is knowledge of reality, while ontology is reality as known, i.e. semantics is the perspectivist signification of reality, and ontology is the aspects of reality signified by semantics. Ontology is the aspect of mind-independent reality that is cognitively captured with the perspective revealed by semantics.
Not all discourses are equally realistic; the semantics and ontologies of discourses are as realistic as they are empirically adequate. Since all semantics is relativized and ultimately comes from sense stimuli, there is no semantically interpreted syntax of language that is utterly devoid of any associated ontology.
Pragmatics in philosophy of science refers to how scientists use language, namely to create and to test theories, and thereby develop scientific laws used in test designs for testing and in scientific explanations.
1.05 Classification of Functional Topics
Basic-science research practices can be classified into four essential functions performed in basic research. They are also topics typically discussed in the academic literature of philosophy of science. They are:
1. The institutionalized aim of basic science is the culturally shared aim that guides development of explanations, which in turn are the final products of basic-scientific research. The institutionalized views and values of science have evolved considerably over the last several centuries, and will continue to evolve episodically in unforeseeable ways with future advancements of science.
2. Discovery refers to the processes of constructing new theories. Pragmatists define theory language pragmatically, i.e., functionally, as universally quantified statements expressible in conditional form including equations (a.k.a. “models”) that are proposed for empirical testing.
Theories are universally quantified hypotheses proposed for empirical testing.
3. Criticism refers to the decision criteria used for the evaluation of theories. Pragmatists accept only the empirical criterion. Empirical testing is the pragmatics of theory language, and it uses modus tollens conditional deductive argument form. A scientific law in the conditional deduction is an empirically tested and nonfalsified theory.
Test-designs are universally quantified discourse presumed for empirical testing a theory, in order to identify the subject of the theory independently of the theory and to describe the procedures for performing the test.
4. Explanation is language that describes the occurrence of individual events and conditions as caused by the occurrence of other described events and conditions according to law statements. It uses modus ponens conditional deductive argument form, which includes universally quantified statements expressible in conditional form that are scientific laws. Whenever possible the explanation is predictive of future events or of evidence for past events. A scientific law is a former theory that has been tested with a nonfalsifying outcome.
1.06 Classification of Modern Philosophies
Twentieth-century philosophies of science may be classified into three generic types. They are romanticism, positivism and contemporary pragmatism. Romanticism is a philosophy for social and cultural sciences. Positivism is a philosophy for all sciences inspired by reflection on Newtonian physics. Contemporary pragmatism is a philosophy for all sciences inspired by reflection on quantum physics. Each generic type has many representative authors advocating philosophies expressing similar concepts for such metalinguistic terms as “theory”, “law” and “explanation”. Philosophies within each generic classification have their differences, but they are much more similar to each other than to those in either of the two other types. These various concepts are related as follows:
Aim of Science
Chapter 2. Modern Philosophies
This chapter sketches the three generic types of twentieth-century philosophy of science in terms of the four functions.
Philosophy of language will be taken up in Chapter 3. Then all these elements will be integrated in a discussion of the four functional topics in Chapter 4.
Romanticism has effectively no representation in the natural sciences today, but it is still widely represented in the social sciences including economics and sociology. It has its roots in the eighteenth-century German idealist philosophers including notably Immanuel Kant (1770-1831), progenitor of romanticism, and especially Georg Hegel (1724-1804) with the latter’s exclusive emphasis on ideas in social culture. The idealist philosophies are of purely antiquarian interest to most philosophers of science today. Romantics have historically defaulted to the positivist philosophy for the natural sciences, but they reject imitating the positivist philosophy for the social sciences. Romantics maintain that there is a fundamental difference between sciences of nature and sciences of culture.
Aim of science:
For romantics the aim of the social sciences is an investigation of culture that yields an “interpretative understanding” of “human action”, by which is meant explanation of social interactions in terms of subjective mental states, i.e., ideas and motives, views and values that are culturally shared by members of social groups.
This concept of the aim of science and of explanation is called a “foundational agenda”, because it requires reduction of the social sciences to a social-psychology foundation, i.e., description of observed social behavior by reference to subjective social-psychological mental states.
For romantics the creation of “theory” in social science may originate either (1) in the social scientist’s introspective reflection on his own ideas and motivations originating in his actual or imaginary personal experiences, which ideas and motives are then imputed to social members, or (2) in empirical survey research findings reporting social members’ expressed ideas and motivations.
Romantics define “social theory” as language describing subjective mental states, notably culturally shared ideas and motivations.
Some romantics call the imputed motives based in introspective reflection “substantive reasoning” or “interpretative understanding”. But all romantic social scientists deny that social theory can be developed by data analysis exclusively or by observation of overt behavior alone.
Romantics therefore oppose their view of the aim of science to that of the positivists’ such as the sociologist George Lundberg (1933) and the behavioristic psychologist B.F. Skinner (1904-1990). Romantics therefore say that they explain consciously purposeful and motivated “human action”, while the behaviorists say they explain publicly observable “human behavior”. Some romantics call the empathetic interpretation of written texts describing human action “hermeneutics”.
For romantics the criterion for criticism is “convincing interpretative understanding” that “makes substantive sense” of conscious motivations, which are deemed to be the underlying “causal mechanisms” of observed human action.
Causality is an ontological concept, and all romantics impose their mentalistic ontology as the criterion for criticism, while making empirical or statistical analyses at most optional and supplementary.
Furthermore many romantic social scientists require as a criterion that a social theory must be recognizable in the particular investigator’s own introspectively known subjective personal experience. In Max Weber’s (1864-1920) terms this is called verstehen. It is the thesis that empathetic insight is a necessary valuable tool in the study of human behavior, which is without counterpart in the physical sciences. It effectively makes all sociology what has been called “folk sociology” or “pop sociology”.
Romantics maintain that only “theory”, i.e., language describing subjective motives, can “explain” conscious human action.
Motives are the “mechanisms” referenced in “causal” explanations, which are also called “theoretical” explanations. Observed regularities are deemed incapable of “explaining”, even if they enable correct predictions.
Some formerly romantic social scientists such as the Institutionalist economist Wesley Mitchell (1874-1948) and the functionalist sociologist Robert Merton (1910-2003) have chosen to focus on objective outcomes rather than subjective motives. This emphasis enhances the testability and thus the scientific status of sociology.
Historically positivism was a reaction against romanticism, but more recently positivism has been relegated to history of philosophy. Positivism’s origins are in the eighteenth-century British empiricist philosophers including notably David Hume (1711-1776). But not until the late nineteenth century did positivism get its name from the French philosopher Auguste Comte (1798-1857), who also founded sociology.
The “neopositivists” were the last incarnation of positivism. They attempted to apply the symbolic logic fabricated by Bertrand Russell (1872-1990) and Alfred Whitehead (1861-1947) early in the twentieth century, because they had fantasized that the Russellian truth-functional symbolic logic could serve philosophy, as mathematics has served physics. They are therefore also called “logical positivists”.
Contrary to romantics, positivists believe that all sciences including the social sciences share the same philosophy of science. They therefore reject the romantics’ dichotomy of sciences of nature and sciences of culture. And the positivists’ ideas about science originated in their reflection upon Newtonian physics.
Aim of science:
For positivists the aim of science is to produce explanations having objectivity grounded in “observation language”, which by its nature describes observed phenomena.
Their concept of the aim of science is thus also called a “foundational agenda”, although the required foundation is quite different from that of the romantics.
Positivists believed that empirical laws are inferentially discovered by inductive generalization based on repeated observations.
They define empirical laws as universally quantified statements containing only “observation terms” describing observable entities and/or phenomena.
Early positivists such as Ernst Mach (1826-1916) recognized only empirical laws for valid scientific explanations. But after Einstein’s achievements neopositivists such as Rudolf Carnap (1836-1970) recognized hypothetical theories for valid scientific explanations, if the theories could be linguistically related to language used to report the relevant observations.
Unlike empirical laws, theories are not produced by induction from singular observations. Positivists believed that theories are discovered by creative imagination, but they left unexplained the creative process of developing theories.
The neopositivists define theories as universally quantified statements containing any “theoretical terms”, i.e., terms not describing observable entities or phenomena.
The positivists’ criterion for criticism is publicly accessible observation expressed in language containing only “observation terms”, which are terms that by their nature describe only observable entities or phenomena.
Positivists maintain that theories are indirectly and tentatively warranted by empirical laws, when valid laws can be logically derived from the theories.
Like Hume they deny that either laws or theories can be permanently validated empirically, but they require that the general laws be founded in observation language as a condition for the objectivity needed for valid science. And they maintain that particularly quantified observation statements describing singular events are incorrigible and beyond revision.
Positivists reject the romantics’ verstehen thesis of criticism. They argue that empathy is not a reliable tool, and that the methods of obtaining knowledge in the social sciences are the same as those used in the physical sciences. They complain that subjective verstehen may easily involve erroneous imputation of the idiosyncrasies of the observer’s experiences to the subject of inquiry.
Positivists and specifically Carl Hempel (1905-1997) and Paul Oppenheim (1885-1977) in their paper “Logic of Explanation” in the journal Philosophy of Science (1948) advocate the “covering-law” schema for explanation.
According to the “covering-law” schema for explanation, statements predicting observable individual events are derived deductively from observation-language statements describing observable individual events together with “covering”, i.e., universally quantified empirical laws.
This concept of explanation has been called the “deductive-nomological model”.
Positivists also maintained that theories explain laws, when the theories are premises from which the empirical laws are deductively derived as theorems with the mediation of “correspondence rules” also called “bridge principles”. Correspondence rules are sentences that relate the theoretical terms in a theory to the observation terms in the empirical laws. The paradigmatic case is Kepler’s laws derived from Newton’s theory.
2.03 Contemporary Pragmatism
We are now said to be in a “postpositivist’ era in the history of Western philosophy, but this term merely says that positivism has been relegated to history; it says nothing of what has replaced it. What has emerged in reaction to positivism is a new coherent master narrative appropriately called “contemporary pragmatism” or “neopragmatism”, which was occasioned by reflection on quantum theory, and is currently the ascendant philosophy in American academia.
Pragmatism has earlier versions in the classical pragmatists, notably those of Charles Peirce, William James (1842-1910) and John Dewey (1859-1952). Some theses in classical pragmatism such as the importance of belief have been carried forward into the new.
In contemporary pragmatism belief is strategic, because it controls relativized semantics, which signifies a correspondingly relativized ontology that is realistic to the degree that a belief is empirically adequate.
Especially important is Dewey’s emphasis on participation and his pragmatic thesis that the logical distinctions and methods of scientific inquiry develop out of scientists’ successful problem-solving processes.
The provenance of the contemporary pragmatist philosophy of science is 1932 Nobel-laureate physicist Werner Heisenberg’s (1901-1976) reflections on the language in his revolutionary quantum theory in microphysics. There have been various alternative semantics and thus ontologies proposed for the quantum theory. Most physicists today have accepted one that has ambiguously been called the “Copenhagen interpretation”.
There are two versions of the Copenhagen interpretation. Contrary to the alternative “hidden variables” view of David Bohm (1917-1992), both Copenhagen versions assert a thesis called “duality”. The duality thesis is that the wave and particle manifestations of the electron are two aspects of the same entity, as Heisenberg says, rather than two separate entities, as Bohm says.
1922 Nobel-laureate Niels Bohr (1885-1962), founder of the Copenhagen Institute for Physics, proposed a version called “complementarity”. His version says that the mathematical equations of quantum theory must be viewed instrumentally instead of descriptively, because only ordinary discourse and its refinement in the language of classical physics can describe physical reality. Instrumentalism is the doctrine that scientific theories are not descriptions of reality, but are merely useful linguistic instruments that enable correct predictions.
The quantum theory says that the electron has both wave and particle properties, but in classical physics the semantics of the terms “wave” and “particle” are mutually exclusive – a wave is spread out in space while a particle is a concentrated point. Therefore Bohr maintained that description of the electron’s duality as both “wave” and “particle” is an indispensable semantic antilogy that he called “complementarity”.
Heisenberg, a colleague of Bohr at the Copenhagen Institute, proposed his alternative version of the Copenhagen interpretation. His version also contains the idea of the wave-particle duality, but he said that the mathematical expression of the quantum theory is realistic and descriptive rather than merely instrumental. And since the equations describing both the wave and particle properties of the electron are mathematically consistent, he disliked Bohr’s complementarity antilogy.
Years later Yale University’s Russell Hanson said that Bohr maintained a “naïve epistemology”. Duality is a thesis in physics while complementarity is a thesis in philosophy of language. However, the term “complementarity” has since acquired some conventionality to signify duality, and is now ambiguous as to the issue between Bohr and Heisenberg, since physicists typically disregard the linguistic issue.
These two versions of the Copenhagen interpretation differ in their philosophy of language. Bohr’s philosophy is called a “naturalistic” view of semantics, which requires what in his Atomic Physics and the Description of Nature (1934) he called “forms of perception”. Heisenberg’s philosophy is called an “artifactual” or a “conventionalist” view of semantics, in which the equations of the quantum theory supply the linguistic context, which defines the concepts that the physicist uses for observation.
1921 Nobel-laureate physicist Albert Einstein (1879-1955) had influenced Heisenberg’s philosophy of language, which has been incorporated into the contemporary pragmatist philosophy of language. And consistent with his relativized semantics Heisenberg effectively practiced ontological relativity and maintained that the quantum reality exists as “potentia” prior to determination by execution of a measurement operation. For Heisenberg indeterminacy is real.
Contemporary pragmatism is a general philosophy for all empirical sciences, both social and natural sciences. The distinctive linguistic philosophy of Einstein and especially Heisenberg as incorporated into the contemporary pragmatist philosophy of science can be summarized in three theses, which may be taken as basic principles in contemporary pragmatism:
Thesis I: Relativized semantics.
The seminal work is “Quantum Mechanics and a Talk with Einstein (1925-1926)” in Heisenberg’s Physics and Beyond (1971). There Heisenberg relates that in April 1925, when he presented his matrix-mechanics quantum physics to the prestigious Physics Colloquium at the University of Berlin, Einstein, who was in the assembly, afterward invited Heisenberg to chat in his home that evening. In their conversation Einstein said that he no longer accepts the positivist view of observation including such positivist ideas as operational definitions. Instead he issued the aphorism: “the theory decides what the physicist can observe”.
Einstein’s aphorism about observation contradicts the fundamental positivist thesis that there is a dichotomous separation between the semantics of observation language and that of theory language. Positivists believed that the objectivity of science requires that the vocabulary and semantics used for incorrigible observation must be uncontaminated by the vocabulary and semantics of speculative and provisional theory.
In the next chapter titled “Fresh Fields (1926-1927)” in the same book Heisenberg reports that Einstein’s discussion with him in Berlin had later occasioned his own reconsideration of observation. Heisenberg recognized that classical Newtonian physical theory had led him to conceptualize the observed track of the electron as continuous in the cloud chamber – an instrument for microphysical observation developed by 1927 Nobel-laureate C.T.R. Wilson (1869-1961) – and to misconceive the electron as simultaneously having a definite position and momentum like all Newtonian bodies in motion.
Recalling Einstein’s aphorism that the theory decides what the physicist can observe, Heisenberg reconsidered what is observed in the cloud chamber. He rephrased his question about the electron tracks in the cloud chamber using the concepts of the new quantum theory instead of those of the classical Newtonian theory. He therefore reports that he asked himself: Can the quantum mechanics represent the fact that an electron finds itself approximately in a given place and that it moves approximately at a given momentum? In answer to this newly formulated question he found that these approximations can be represented mathematically. He reports that he then developed this mathematical representation, which he called the “uncertainty relations”, the historic contribution for which he received the Nobel Prize in 1932.
Later Russell Hanson expressed Einstein’s aphorism that the physical theory decides what the physicist can observe by saying observation is “theory-laden” and likewise Popper (1902-1994) by saying it is “theory-impregnated”.
Thus for pragmatists the semantics of all descriptive terms is determined by the linguistic context consisting of universally quantified statements believed to be true.
In his Against Method (1975, Ch. 2-3) Paul Feyerabend (1924-1994) also recognized employment of relativized semantics to create new observation language for discovery, and he called the practice “counterinduction”. To understand counterinduction, it is necessary to understand the pragmatic concept of “theory”: theories are universally quantified statements that are proposed for testing. Feyerabend found that Galileo had practiced counterinduction in the Dialogue Concerning the Two Chief World Systems (1632), where Galileo (1564-1642) reinterpreted apparently falsifying observations in common experience by using the concepts from the apparently falsified heliocentric theory instead of the concepts from the geocentric theory. Likewise Heisenberg had also practiced counterinduction to reconceptualize the perceived sense stimuli observed as the electron track in the cloud chamber by using quantum concepts instead of classical Newtonian concepts.
Counterinduction is using the semantics of an apparently falsified theory to revise the test design language that had supplied the semantics of the language describing the apparently falsifying observations, and thereby to produce new observation language.
Like Einstein, contemporary pragmatists say that the theory decides what the scientist can observe. Thus semantics is relativized in the sense that the meanings of descriptive terms used for reporting observations are not just names or labels for phenomena, but rather are determined by the context in which they occur.
More specifically in “Five Milestones of Empiricism” in his Theories and Things the pragmatist philosopher of language Willard van Quine (1908-2000) says that the meanings of words are abstractions from the truth conditions of the sentences that contain them, and that it was this recognition of the semantic primacy of sentences that give us contextual definition.
Essentially the defining context consists of universally quantified statements that proponents believe to be true.
The significance is that the acceptance of a new theory superseding an earlier one and sharing some of the same descriptive terms produces a semantical change in the descriptive terms shared by the theories and by their observation reports.
Thus Einstein for example changed the meanings of such terms as “space” and “time”, which occur in both the Newtonian and relativity theories. And Heisenberg changed the meanings of the terms “wave” and “particle”. Feyerabend calls the semantical change due to the relative nature of semantics “meaning variance”.
Thesis II: Empirical underdetermination.
Empirical underdetermination refers to the limited ability of the semantics of language at any given time to signify reality.
Measurement error and conceptual vagueness, which can be reduced indefinitely but never completely eliminated, exemplify the omnipresent and ever-present empirical underdetermination of descriptive language that occasions observational ambiguity and theoretical pluralism. Experience is replete with instances for which it is unclear as to whether or not a descriptive term may apply. Einstein recognized that a plurality of alternative but empirically adequate theories could be consistent with the same observational description, a situation that he called “an embarrassment of riches”.
Additional context including law statements in improved test-design language contributes additional semantics to the observational description in the test designs, thus reducing while never completely eliminating empirical underdetermination. In his Word and Object (1960) Quine introduced the phrase “empirical underdetermination”, and wrote that the positivists’ “theoretical” terms are merely more empirically underdetermined than terms they called “observation” terms. Thus contrary to the positivists the types of terms are not qualitatively different.