Herbert L. Roitblat (1994) A Representational View of Science. Psycoloquy: 5(44) Scientific Cognition (7)

Volume: 5 (next, prev) Issue: 44 (next, prev) Article: 7 (next prev first) Alternate versions: ASCII Summary
Topic:
Article:
PSYCOLOQUY (ISSN 1055-0143) is sponsored by the American Psychological Association (APA).
Psycoloquy 5(44): A Representational View of Science

A REPRESENTATIONAL VIEW OF SCIENCE
Book review of Giere on Scientific Cognition

Herbert L. Roitblat
Department of Psychology
University of Hawaii
2430 Campus Road
Honolulu, HI 96822

roitblat@hawaii.edu

Abstract

Cognitive science can provide the tools for examining and understanding conceptual development in scientists. Science develops nonalgorithmically and nonlinearly. Scientists in different paradigms use different ontological representations.

Keywords

Cognitive science, philosophy of science, cognitive models, artificial intelligence, computer science, cognitve neuroscience.
1. The positivists attempted to conceive of science as a purely formal process. They were reacting to the revolution in physics that replaced Newtonian mechanics with relativistic quantum mechanics. They attributed the failure of Newtonian mechanics to the infiltration of sloppy language, which allowed physicists to use terms that were only poorly defined. In their view, these sloppy terms allowed physicists to deceive themselves into thinking that they understood terms when actually the terms did not mean anything. Logical positivism was based on two assumptions: (1) there can be infallible observations, and (2) deductions from these observations are sufficient to capture all reliable and necessary theorizing in science. Only those terms in the observation language (e.g., see Grandy) and those derived via deductions from those observations were to be admissible as reliable scientific terms, because only these could have a fixed and reliable meaning. Logical positivism cast scientific theory building as a purely formal (i.e., syntactic) algorithmic process.

2. Over the last few decades the philosophy of science has taken a decidedly naturalistic turn. Instead of emphasizing a rational prescriptive approach to science, many, or even most philosophers of science have focused on how scientists actually conduct their work. When scientific process was examined, it became clear that the actual scientific practice of discovery had little to do with the formal processes prescribed by the positivists. It became necessary to develop conceptual tools that would allow philosophers to think about the process of doing science, and allow them to understand how scientists combine their human cognitive abilities with the conceptual resources available to them as members of a scientific community to create and communicate new scientific representations (Nersessian).

3. Kuhn (1970) proposed the approach of viewing science as a perceptual problem within the Gestalt tradition. Gestalt perception consists of organizing perceptual information into a coherent sensible whole or Gestalt. The particular Gestalt that was reached was affected by the context, expectations, and prior information of the perceiver. Kuhn argued forcibly that similar psychological processes were essential to science and that scientific practice depended not only on specific facts, but also on psychological features of the scientist (Giere).

4. A key feature of Kuhn's approach to science was the radical shifts in perspective that he called paradigm shifts. Paradigm shifts entail a change in world view, which includes the set of problems the field finds important, the units of analysis by which those problems are addressed, and examples of successful solutions. The change from Newtonian mechanics to relativistic physics is one example of such a paradigm shift. Kuhn also argued that these paradigm shifts are typically so radical that adherents of one paradigm have a difficult time understanding the issues and arguments of the other. The two paradigms are said to be incommensurable in that the problems addressed by the two camps and the tools used to address them are substantially different. Because of the incommensurability, it is impossible, he argued, to systematically compare the two positions and decide rationally (i.e., prove) whether one approach is better than the other. In the place of formal methods, scientists, in Kuhn's view, rely on heuristics and sociological factors to choose paradigms. Many of Kuhn's critics argued that his view of comparing paradigms depended on factors that were irrational and therefore outside the realm of philosophical or scientific investigation. There was nothing, therefore, to demarcate science from nonscientific approaches to knowledge. Others tried to defend the notion of paradigms by finding a way to couch paradigm comparisons in a systematic if not purely algorithmic light.

5. An important theme in Giere's (1992, 1993) book is that cognitive science provides a way to characterize these paradigm shifts. The apparent mystery of paradigm shifts is due to the inadequate means used to understand a scientist's thought process rather than to any inherent irrationality in the process. They are irrational only in comparison to the deductive methodology prescribed by the positivists. In the rest of this review I will focus on the issue of conceptual representation in science and on the changes that occur when scientists switch between paradigms. My primary concern will be the chapters by Carey, Chi, Gooding, and Nersessian.

6. Scientists' thought is not easily captured as a formal syntactic algorithmic process. As mentioned earlier, one of the goals of the positivists was to develop an algorithmic approach to science that would ensure that scientists discovered only true descriptions of the world. Gooding calls this a "one-pass" view of science: theory building is incremental, analogous to constructing a house in which earlier work provides an always solid foundation for later work. The one-pass view derives in part from the reliance on the highly reconstructed accounts of science that are presented in journals and textbooks as the source material for philosophical analysis. The finished work presented in journal articles and other publications is intended to maximize communication and to conceal the thought processes by which the finished product was obtained. Gooding argues that this one-pass view allowed philosophers to focus on formal logical processes and ignore the substantial amount of qualitative and constructive work that actually generated it. Gooding goes on to argue that scientists manipulate material things as well as representations, objects as well as words, and that these manipulations also play an essential, though difficult to formalize role in scientific thought. Scientific thought is nonlinear, reflexive, and recursive, rather than neat and linear as the published work implies. Both Gooding and Nersessian point out that the relevant information for understanding how scientists think is not to obtainable exclusively from their finished work, but rather requires analysis of less formal records.

7. Nersessian echoes the theme that scientific thought is not scientific language. Much ill-formed and difficult to capture and describe work seems to occur before scientists can approximate their thoughts with expressions. The focus on scientific language has slowed progress in understanding science, and the belief that nonalgorithmic or at least nonlinguistic processes are too mysterious to be understood has further limited even attempts to understand the problem of how scientists think.

8. Another factor that has limited our understanding of scientific thought is the belief that the meanings of scientific terms can be captured by neat bundles of necessary and sufficient conditions. In fact, this is the traditional but inadequate view of word meanings. Many concepts, such as furniture, lack necessary and sufficient characteristics. All categories (Barsalou, 1987), including such technical concepts as numbers (Armstrong, Gleitman, & Gleitman, 1983), appear to have graded structure in that some items are better, more typical, and easier to identify members of a category than others. If all members of the category share necessary and sufficient features, it is difficult to see how some members of the category could be better examples than others (see Medin & Wattenmaker, 1987). Definitions of many working scientific terms are no more formal than those concepts in our everyday language. In psychology, for example, the concept of intelligence often receives one or another operational definition, but comparing operations requires that the investigators have some underlying notion of what intelligence is that is measured by intelligence tests; yet there is no satisfactory definition of the term. Similarly, Nersessian claims that there is no set of necessary and sufficient conditions for defining the concept of electromagnetic field, despite its long history of investigation in physics. Insisting that concepts be formal and rigid may make investigations of those concepts easier, but basing an analysis on inappropriate elements, no matter how accessible they are, is bound to be misleading.

9. The basic elements of scientific thought are not discrete formal units, nor are the thought processes that use these elements discrete and formal. Science depends strongly on the use of analogical reasoning, imagery, thought experiments, and limiting case analysis. These processes are nonalgorithmic (Nersessian). For example, despite a great deal of interest, there is no algorithmic account of how people understand metaphors and similes (cf. Gentner, 1983). The positivists recognized the need for deduction in scientific reasoning. Post- positivists such as Popper (1968) recognized that science necessarily entails induction. Attempts to limit psychological thought to these two kinds of processes is artificially limiting and misleading (Nersessian).

10. Changing paradigms entails changing ontologies. Processes of ontological change are difficult to use and difficult to describe. Another important contribution of this book to understanding scientific practice is the view it offers of paradigm shifts and comparisons. Recall Kuhn's (1970) argument that scientists pursue their subjects in the context of a paradigm, which, among other things, provides them with a world view, a set of tools, a set of problems requiring solution, and a set of examples of successful application of the paradigm's tools for solving those problems. Alternative paradigms for the same research area are said to be incommensurable because of the different vocabularies, world views, and problem sets recognized by each group of adherents.

11. Following Kitcher's (1988) analysis of Kuhn's (1982) discussion of incommensurability, Carey considers the possibility that incommensurability between two paradigms is due to adherents' use of the two paradigms of languages that are not mutually translatable. There are multiple methods for determining the referent of a term, including definitions, descriptions, theory-relative similarity, etc. Within a paradigm, each way of determining the referent determines the same item. Incommensurability between theories occurs when the items determined by one theory do not exist in the other, or when some of the uses of a term determine one item and some determine another item in the other theory. She suggests that the two paradigms use different ontologies. They cut the world into different elementary units and organize these units differently. In addition, because the terms are mutually dependent (e.g., property and theft) and learned together, it is difficult to understand any of them in isolation, so no straightforward lexical comparison is possible outside the ontology and world view of the paradigm. Chi agrees, arguing that paradigm shifts are characterized by an underlying ontological conceptual shift.

12. Carey attaches a great deal of importance to the psycholinguistic aspects of incommensurability, but it is not entirely clear that psycholinguistics is the appropriate model for scientific thought. It is not clear that translatability of languages has anything to do with incommensurability. It seems to me that Newtonian physics, for example, could easily be described in the language of relativity physics, although the two clearly use different ontologies. The problem seems to lie not so much with intertranslatability as with the very meaning of the terms and their organization.

13. Existing theories of meaning do not allow for the possibility of meaning change (Nersessian). This failure has produced many of the conundrums associated with the problem of incommensurability. If concepts are represented by neat bundles and individuated units (e.g., as sets of necessary and sufficient conditions) as has been traditionally held, then only replacement, not change, is possible (Nersessian).

14. Chi (p. 132) echoes this claim in the argument that no transformational mechanism, such as deletion or addition of features, analogy, generalization, or specialization, movement, surgery, etc. can transform a concept in one ontological category into a concept in another ontological category. Changes of ontological categories might be better described as the development of new concepts, with the old concepts remaining more or less intact, and perhaps eventually being abandoned. The difficulty of ontological change is that the new category must first be learned, understood, or induced, and then it must be realized that a concept does not belong to the old ontological category, but the new one (Chi). It cannot be changed, it must be replaced.

15. By this analysis, paradigm shifts are as difficult and relatively rare as they are because of the severe intellectual demands that ontological changes place on the scientists. The meanings of terms and what scientists say about those terms differ between paradigms. For example, phlogiston is not simply another name for oxygen; rather, it is a different kind of concept in an ontology different from that of elemental oxygen (Carey). Nevertheless, cognitive science appears to provide tools for investigating the representations scientists use when they use these terms. By understanding these representations, we can understand the scientific process and contribute to the education of our students.

16. A cognitive scientific approach to understanding scientific practice is not just an exercise in armchair intellectualism. As this book demonstrates, it has important implications for our understanding of scientific practice and, more important, for science education. For example, Chi points out that students come to science education with preconceived notions and ontologies. Learning to adopt the ontology of the field in which they are studying is often difficult and painful because it cannot be done feature by feature (e.g., by adding or subtracting features from familiar concepts); rather, it requires a conceptual reorganization that first has one construct new ontological structures and then abandon old familiar structures. The problem people have with ontological shifts is sufficient to explain the difficulty many students have with science education. The problem is confounded, however, by the epistemology that many students also bring to science education. For example, Perry (1981) noticed that university students tend to go through a series of stages in their educational epistemology. At the earliest stage they believe that knowledge consists of definite facts, about which the experts are certain, and their goal as students is to learn those facts as they are dispensed by the professor. Such an approach to knowledge makes ontological change difficult, because students try to memorize the definitions of specific terms rather than fitting them into a more elaborate ontology; they have difficulty conceiving explicitly of the possibility that there can be more than one ontology. At later stages of development they recognize that there are different points of view and that these can be compared. Finally, some students reach a stage at which they recognize that although there is no certainty they are right, they can evaluate alternative points of view and select (even if tentatively) the most accurate and compelling account of a phenomenon. It is interesting to note that the progression of epistemological stages described by Perry corresponds roughly to the transition that occurred from logical positivism (just facts and deductions from them) to the historicist position (and its descendent in this book) in which alternative paradigms can be compared and tentatively accepted with the awareness that subsequent information may necessitate their rejection.

REFERENCES

Armstrong, S.L., Gleitman, L.R. & Gleitman, H. (1983) On what some concepts might not be. Cognition, 13, 263-308.

Barsalou, L.W. (1987) The instability of graded structure: Implications for the nature of concepts. In U. Neisser (Ed.), Concepts and conceptual development: Ecological and intellectual factors in categorization. New York: Cambridge University Press.

Gentner D. (1983) Structure-mapping: A theoretical framework for analogy. Cognitive Science, 7, 155-170.

Giere, R.N. (1992) Cognitive Models of Science. Minnesota Studies in the Philosophy of Science, volume 15. Minneapolis: University of Minnesota Press.

Giere, R.N. (1993) Precis of Cognitive Models of Science. PSYCOLOQUY 4(56) scientific-cognition.1.giere.

Kitcher, P. (1988) The child as parent of the scientist. Mind and Language, 3, 217-228.

Kuhn, T. (1970) The structure of scientific revolutions. Chicago: University of Chicago Press.

Kuhn, T. (1982) Commensurability, comparability, communicability. In PSA 1982, vol. 3. pp. 669-688. East Lansing, MI: Philosophy of Science Association.

Medin, D.L. & Wattenmaker, W.D. (1987) Category cohesiveness, theories, and cognitive archeology. In U. Neisser (Ed.) Concepts and conceptual development: Ecological and intellectual factors in categorization. (pp. 25-62). Cambridge: Cambridge University Press.

Perry, W.G. Jr. (1981) Cognitive and ethical growth: The making of meaning. In A. W. Chickering and Associates (Eds.), The modern American college: Responding to the new realities of diverse students and a changing society. (pp. 76-116). San Francisco: Josey-Bass.

Popper, K. (1968) Conjectures and refutations. New York: Harper & Row.

Rips, L.J. (1989). Similarity, typicality, and categorization. In S. Vosniadou & A. Ortony (Eds.), Similarity and analogical reasoning. (pp. 21-59). Cambridge: Cambridge University Press.


Volume: 5 (next, prev) Issue: 44 (next, prev) Article: 7 (next prev first) Alternate versions: ASCII Summary
Topic:
Article: