Empirical Underdetermination

         Another factor affecting decidability of empirical testing is the empirical underdetermination of language, with the result that empirical criteria cannot always result in unambiguous theory choice.  Mathematically expressed theories use measurement data containing some measurement error, which is a manifestation of empirical underdetermination.  Scientists like measurement and mathematically expressed theories, because they can measure the amount of error in the theory.  But separating the measurement error from the prediction error made by the equation constituting the theory can be problematic.  Repetition of the measurement procedure enables estimation of the degree or range of measurement error.  If the prediction made by the equation exhibits an error that is large relative to the estimated measurement error, then the theory is deemed conclusively falsified.  Otherwise the theory is either untestable or the test design is inadequate for the theory.  If there are several theories yielding prediction errors that are different but small relative to one another, and are also small enough to be within the estimated range of measurement error, then the inescapable empirical underdetermination inherent in language has imposed undecidability in the choice of alternative theories for the given test design.  The problem of empirical underdetermination also occurs in the testing of qualitative theories.  In such cases the empirical underdetermination is manifested as conceptual vagueness.  All concepts have vagueness, which can be reduced but can never be eliminated. Empirical tests are conclusive to the extent that measurement error and vagueness are small relative to the effect produced in the empirical test.
         Given the dilemma of having several alternative theories that are not falsified by empirical test due to empirical underdetermination, philosophers have proposed nonempirical criteria that may be operative in theory choice.  But no such nonempirical criteria enable scientists to predict which alternative nonfalsified theory will make more reliable predictions, when the degree of empirical underdetermination is reduced by improved observation practices or test designs.  And when scientists are confronted by such dilemmas, better observation practices with test designs having added clarifying information or more accurate measurements are in order.  The existence of several alternative theories that have survived an empirical test without having been falsified is thus endemic to science.  In the social sciences that use statistical techniques for testing this is a common outcome, but it also happens in natural sciences.  Einstein had described this situation in physics as an “embarrassment of riches.”  The resulting multiple explanations are equally scientific.  Different scientists may have distinctive reasons, such as aesthetic, circumstantial, or strategic reasons, for preferring one alternative explanation to another.  Thagard has noted three such criteria implemented in his ECHO system, his artificial-intelligence system specifically for theory selection.  He finds that the most important criterion is breadth of explanation, followed by simplicity of explanation, and finally analogy with previously accepted theories.  Where the empirical criteria are not decisive, theory selection becomes more of a professional career decision for the scientist rather than a purely scientific one.  For example knowing what a profession currently likes to see in new theories helps getting a paper published in the refereed literature.  Thagard considers these selection criteria to operate as inference to the “best” explanation.  But contemporary Pragmatists are inclined to exclude all such nonempirical criteria from the aim of science, because while relevant to persuasion, they are irrelevant to evidence.  They are like the psychological criteria that trial lawyers use to select and address juries in order to win lawsuits, but have nothing to do with courtroom evidence rules.

Scientific Pluralism

         Language is always empirically underdetermined by the real world, such that there is always the possibility of the development of a new theory that is empirically equal to or superior to currently accepted explanations of the same subject. This empirical underdetermination may be due to errors of measurement or to the residual vagueness always present in descriptive variables and terms, and it is often reduced by development of more adequate observation practices and technologies for test designs.  The undecidability of the resulting empirically adequate multiplicity of scientific explanations is recognized by the Pragmatist thesis of “scientific pluralism.”  Scientific pluralism is the undecidability among alternative laws and consequently explanations due to the empirical underdetermination of language.

Nonempirical Linguistic Constraints

         The constraint imposed by empirical test outcomes, the empirical constraint, is an institutionalized cultural value that is not viewed as an obstacle to be overcome, but rather is a condition to be respected for the advancement of science.  There are other constraints that are viewed as impediments that must be overcome for the advancement of science.  Some of these impediments are purely circumstantial.  They may be sociological, dogmatic, financial, political or academic.  These impediments are external to science.  There are two other nonempirical constraints that are internal to science in the sense that they are inherent in the nature of language. They are the cognition constraint and the communication constraint.
         The cognition constraint inhibits a theorist’s ability to construct new theories, and it is manifested as what is often mundanely referred to as lack of imagination.  Semantical rules are not just rules; they are also linguistic habits that enable fluency in both thought and speech.  The rules are such that the meaning of a descriptive subject term is determined by the set of universally quantified statements believed to be true.  Thus given the belief in certain universally quantified statements, the meanings of their constituent descriptive terms are determined.  Conversely given the established meaning of a descriptive term, certain conventions and beliefs are sustained, with the result that change of belief is made difficult by the need to change linguistic habits.  Accordingly the more revolutionary the revision of beliefs, the greater the impeding force of the cognition constraint imposed by psychological habit.  And if a new syntax is required such as an unfamiliar mathematics, then the semantical restructuring of the affected meaning complexes is even greater.  This follows from the relativistic semantics, which is opposed to the thesis that language is neutral in the sense of being merely a passive instrument for thought.  It is noteworthy that the use of discovery systems circumvents this problem, because the machines have no linguistic habits; they mechanically apply a generative grammar to inputted linguistic symbols.
         The communication constraint is similar to the cognition constrain; it is merely the impediment to understanding a new theory relative to those currently known due to prevailing linguistic habits.  The scientist must learn the new theory well enough to restructure the composite meaning complexes associated with the descriptive terms common to both the old theories he already knows and new theory to which he had recently been exposed.  And it may be noted that the scientist viewing the computerized discovery system output experiences the same communication impediment with the machine that he would have were the outputted theories developed by a fellow human scientist.
         If the differences between the old and new theories are very great, some members of the affected scientific profession are unwilling or unable to accomplish the learning adjustment required, and they become the rear guard defending the older conventional wisdom.  In the meanwhile the developers and advocates of the new theory, who have mastered the new theory’s semantics, assume the role of the avant garde until the new theory’s acceptance has become sufficiently widespread that it has become the new conventional wisdom appearing in the textbooks.  This is the conversion process described by Kuhn in revolutionary transitional episodes.  However, contrary to Kuhn the transition does not involve a complete semantical discontinuity.  Rather involves an extensive restructuring of the new theory’s semantical description of the domain common to both old and new theories as described by the semantics in their shared test design statements.

Scientific Explanation

         Whether viewed heuristically or historically the ultimate aim of basic science is the production of explanations.  One of the most obvious characteristics of an explanation is that it consists of language.  Thus it may be said that basic science produces a linguistic artifact.  This is what distinguishes basic or pure science from applied science and technology.  Applied science and technology produce nonlinguistic real products, such as engineered buildings, medical clinical therapies, and social policies affecting human activities.  So long as a tested theory has not been falsified, it is accepted as a scientific law, which may occur in an explanation.  Thus in the contemporary Pragmatist philosophy of science “explanation” is defined as follows: An explanation is a deduction containing one or several scientific law statements concluding to statements describing particular events or to universal statements.  Laws and theories are distinguished pragmatically.   A law statement is a former theory that has been tested by the most critical test design currently possible and is not yet falsified by the executed empirical tests.
         The statements or equations of an explanation, like those of a theory, are law statements that are universally quantified logically.  And the litmus test of the law’s universal claim is the prediction of future events or of currently unavailable evidence for past events in an explanation.  Prediction is the guarantee that the law is not ad hoc to its development sample.  Furthermore a motivating and social justification, which is external to basic science, is the control that is often yielded by basic science’s power of prediction.  Such control enables applied science and technology, which fundamentally distinguishes applied science from basic research science.

Summary

         This introduction started with discussion of four types of philosophy of science – Romanticism, Positivism, contemporary Pragmatism, and psychologistic computational philosophy of science.  It then took up philosophy of language – syntax, semantics, ontology, and pragmatics.  And it finally considered the four topical areas of philosophy of science – the aim of science, discovery, criticism, and explanation.  To facilitate an integrated understanding of these three discussions, the following recapitulation picks up the stick from the other end, as it were, and structures the whole discussion around the four topical areas.

Aim of Science:

         On the Romantic philosophy the natural and cultural sciences have different aims.  Romanticists do not object to the Positivist view of the aim of the natural sciences.  In fact it supplies Romantics with a stereotypic misunderstanding of natural science.  But Romantics maintain that the aim of the cultural sciences of human action consists of interpretative understanding in terms of the conscious views, values, norms and motives of human subjects.  Thus they require a mentalistic ontology for the cultural sciences. And they also deny that explanation in cultural science aims at prediction and control.
         On the Positivist philosophy the natural and social/behavioral sciences have the same aim.  That aim is to enable prediction and ideally control of phenomena by means of language either expressing or founded upon direct observation. Positivists reject a nonobservable mentalistic ontology for social sciences. 
         On the contemporary Pragmatist philosophy the aim of science is explanation that enables prediction and ideally control of the real world.  Like the Positivists they maintain the aim of science is the same for all sciences, but unlike the Positivists and the Romantics they reject commitment to any ontology prior to empirical testing, whether mentalistic or nonmentalistic.  Pragmatists permit but do not require mentalistic or nonmentalistic ontologies.
         On the cognitive psychology view the aim of science is whatever they find in history; they do not characterize it.  Philosophically they are eclectic.  They reject Behaviorism, which is Positivism in psychology, yet they distinguish observation terms, as those that are inputted to the system, from theoretical terms, as those that are generated and outputted by the system.  Like the Romantics they view the subject of their psychological investigations as mental representations, but contrary to the Romantics they equate human concepts to the data structures in their computer systems.  And unlike Positivists they are not nominalist.  Within these parameters they select criteria for scientific criticism according to what is needed for their systems to replicate the particular historical episodes that they simulate mechanically.

Discovery:

  The Romantics’ concept of scientific discovery for cultural science is based on their concept of scientific theory, which they define in terms of the mental states experienced by the social actors whose actions they investigate.  Acquiring this kind of knowledge is believed to require introspection by the researcher.  The Romantics therefore deny that social theory can be developed exclusively by analysis of empirically acquired social statistics, and they have a Luddite attitude toward computational theory development with mechanized discovery systems.
             The Positivist’ concept of scientific theory is also distinctive.  They dichotomize observation language and theory language.  The latter contains descriptive terms referencing entities that have never been observed, and thus given their nominalism they presume that theory language is not meaningful until reductively related to the observation language.  Discoveries expressible exclusively with observation terms are called empirical generalizations.  Generalizations are the product of induction resulting from recognition of similarities in repeated direct observations.  The Positivists offer no explanation for the discovery of theories except to note that theories are free creations of the imagination and are not generalizations based in observation.  
            The contemporary Pragmatists’ concept of theory differs from both the Romantics’ and the Positivists’ concept, because the Pragmatists reject associating theory with any prior ontology.  They define theory functionally as any universally quantified discourse proposed for testing.  This concept lends itself to computer processing, since any output from the discovery system is considered to be theory proposed for further empirical testing. 
            Finally the cognitive psychologists’ principal concern is with the development of computerized systems, with the objective of characterizing, proceduralizing and mechanizing the psychology of the discovery process.

Criticism:

            The Romantics require as a criterion for scientific criticism in cultural sciences, that the language describe a mentalistic ontology. Language that does not conform to this prior ontological criterion is rejected out of hand as “atheoretical” and as unsuitable for cultural science notwithstanding valid empirical findings.  Some Romantics furthermore require Weber’s verstehen criterion that the theory be empathetically or vicariously plausible in the personal experience of the researcher.  Such theories are said to “make sense.”
            The Positivists maintain that empirical generalizations are always provisional, and must be tested empirically.  The early Positivists such as Mach viewed theories as temporary expedients relegated to less than scientific status, to be replaced later by empirical generalizations based on direct observations as science progressed.  The later Positivists such as Carnap were more accepting of theories, but conditioned the acceptance of theories not only on the confirming outcome of scientists’ empirical test, but also conditioned the theory’s meaningfulness on the philosophers’ logical reduction to an observation language.
           The contemporary Pragmatists give absolute authority to the outcome of empirical testing as the criterion for theory acceptance and selection, so long as the observed effect is large enough to be distinguishable from error due to the empirical underdetermination of language. They view falsification as conclusive.  They deny that an empirical test outcome can establish a theory, but they accept nonfalsification as warranting belief in the theory’s ontological claims.  The empirical underdetermination manifested by measurement error or conceptual vagueness results in undecidability, such that more than one theory may be empirically nonfalsified.  This scientific pluralism permits the scientists to choose among the alternative tested and nonfalsified theories on the basis of other criteria, such as simplicity or familiarity.
           Cognitive psychology will consider any criteria for scientific criticism that their cognitive systems can successfully use to simulate historical episodes in the progress of science.  These have included empirical adequacy, breadth of explanation, simplicity of explanation, and analogy with established explanations.

 Explanation:

The above mentioned considerations flow through to the topic of scientific explanation.  For the Romantics explanation in cultural science is interpretative understanding.  Knowing the social actors’ misunderstanding is deemed more important than connecting their intentional action to its unintended consequences.  Romantic explanation is discourse having the required mentalistic ontology.
           The Positivists on the other hand are committed to an observational ontology traditionally called phenomena, sense data, or sensations.  They typically reject the mentalistic ontology of the Romantics.  On their philosophy scientific laws are either empirical generalizations containing only observation terms, or they are theories confirmed by empirical testing and found meaningful, because their theoretical terms have been logically related to a suitable observation-language reduction basis.
          The contemporary Pragmatists define scientific law as language that was formerly theory, because it has been empirically tested and has not yet been falsified.  And since nonfalsification warrants belief, the law and the explanations in which it is used describes its own ontology.
          The cognitive psychologists view an explanation as either the output of a cognitive discovery system or a primitive term in a theory-selection system, which is applied to a problem in basic research science.  Cognitive psychologists construe an explanation as a conceptual representation.  

Pages [1] [2] [3] [4] [5] [6]
NOTE: Pages do not corresponds with the actual pages from the book