Biology and Philosophy 15: 443–463, 2000. © 2000 Kluwer Academic Publishers. Printed in the Netherlands.
Evolutionary Epistemology, Social Epistemology, and the Demic Structure of Science TODD A. GRANTHAM Department of Philosophy College of Charleston Charleston, SC 29424 USA E-mail:
[email protected]
Abstract. One of the principal difficulties in assessing Science as a Process (Hull 1988) is determining the relationship between the various elements of Hull’s theory. In particular, it is hard to understand precisely how conceptual selection is related to Hull’s account of the social dynamics of science. This essay aims to clarify the relation between these aspects of his theory by examining his discussion of the “demic structure” of science. I conclude that the social account can do significant explanatory work independently of the selectionist account. Further, I maintain that Hull’s treatment of the demic structure of science points us toward an important set of issues in social epistemology. If my reading of Science as a Process is correct, then most of Hull’s critics (e.g., those who focus solely on his account of conceptual selection) have ignored promising aspects of his theory. Key words: David Hull, evolution, selection
1. Introduction Anyone who has even briefly perused Science as a Process knows that it is an ambitious book. Hull claims that the principal research programs within science studies (e.g., Lakatos’ methodology of scientific research programmes, Kuhn’s structuralist history, Merton’s functionalism, and social constructivism) are all inadequate. His aim is to replace these programs with a more accurate account of how reason, argument, and evidence – as opposed to power, prestige, and politics – influence conceptual change in science. As many reviewers noted, Science as a Process is also a sprawling book: long on narrative and brimming with intriguing suggestions, but short on analytic clarity (see, e.g., Ruse 1989; Donoghue 1990). One of the central difficulties is discerning the precise relationship between two of the central themes of the book: conceptual selection and the social dynamics of competition and cooperation in science (Kitcher 1988). In some places, Hull suggests
444 that the social mechanism of “curiosity, credit, and checking” is a selection process: The mechanism that I propose rests fundamentally on the relations which exist in science between credit, use, support, and mutual testing. Science functions the way that it does because of its social organization . . . This mechanism is an instance of a selection process, but it is social, not biological (1988, p. 281, emphasis added).1 In other places Hull states that the social mechanism “follows naturally from viewing science as a selection process” (p. 285). The second passage suggests a much looser relationship between the selection model and the social structure of science. This ambiguity raises a number of questions: What is the relationship between the social and selectionist aspects of Hull’s account? Does a selection process create the social structure? That is, are we to provide a selectionist (adaptationist?) account of how the social structure of science evolves? Is the social structure simply one dimension of the environment in which conceptual selection occurs? To what extent can the social and selection processes be conceived as independent elements of his theory? The principal goal of this essay is to clarify the relationship between the selectionist and social aspects of Hull’s theory. Attaining this clarity is a crucial step toward a more adequate assessment of Science as a Process. Most of Hull’s commentators – myself included (Grantham 1994) – have focused on Hull’s general analysis of selection and several disanalogies between the natural selection and conceptual change in science. However, without a clear account of the relationship between the social and selectionist aspects of the theory, this way of approaching Hull’s work can go wrong in two ways. First, if the social and selectionist aspects of the theory are truly independent, then criticisms of Hull’s selection account will simply fail to address a significant part of Hull’s theory. Second, if the aspects are intertwined, then a failure to recognize the connections may lead us to dismiss the theory without properly understanding it. The social and selectionist aspects of Hull’s theory might be related in one of three ways: 1. A single, unified theory. Either the social account is identical with the selectionist account, or the two are so closely intertwined that they stand (or fall) together. 2. Interdependent theories. Although the two theories could be teased apart, there are important evidential relations between the theories (e.g., one theory provides evidence for the other). 3. Independent theories. The truth of one theory has no bearing on the truth of the other theory.
445 The truth, I will suggest, lies somewhere between (2) and (3). The social and selectionist aspects of Hull’s theory fit together nicely in many ways. For instance, conceptual selection takes place within a social (as well as physical and conceptual) environment. Hull’s model of the social environment thus provides a crucial resource for his account of conceptual selection. Nonetheless, I will emphasize ways in which the two elements are relatively independent. In particular, I suggest that the social account has considerable merit independent of Hull’s selectionist account of science. It is difficult to address the relations between Hull’s social and selectionist theses in a fully general way: the issues are simply too large. My strategy is to explore the broader issues by analyzing one central theme: the “demic” structure of science. Scientists are generally organized into “demes” – small, relatively short-lived groups of scientists who primarily interact with one another (rather than members of other demes). Hull maintains that the demic structure of science is “extremely interesting” because “several of the most important features of science flow from it” (p. 15). Because demic structure is central to Hull’s theory, it provides a nice window through which we can view the relationship between the selectionist and social aspects of Hull’s theory. I pursue this project in three phases. Section 2 provides an overview of Hull’s theory. Section 3 examines Hull’s claim that conceptual selection has led to the demic structure of science. I argue that the general structure of Hull’s explanation can be preserved in a “folk psychological” account that does not presuppose his analysis of selection. Section 4 focuses on Hull’s claim that the demic structure of science is epistemically significant. I reconstruct his position as a version of social epistemology and argue that this position is (1) worthy of serious examination, (2) independent of the difficulties facing Hull’s account of conceptual selection, but (3) closely bound up with an evolutionary understanding of conceptual change.
2. A précis of Science as a Process Hull offers, in the words of his subtitle, an evolutionary account of social and conceptual change in science. His theory is built on two central pillars: an analysis of selection processes and an account of competition and cooperation among scientists. 2.1. The first pillar: Hull’s analysis of selection processes According to Hull, conceptual selection and natural selection share the same abstract structure. Just as some genes are selected and passed on because
446 they code for traits that are better adapted to the physical and social environment, so some ideas are selected and passed on because they “fit” the facts revealed by experiments and/or are better adapted to the social and conceptual environment of the relevant scientific community. Theories that treat conceptual change as an analog of biological selection have been seriously criticized for their failure to recognize certain differences between these processes (see Bradie 1986). In order to answer these objections, Hull presents a general analysis of selection processes which identify the essential features of selection and abstracts away from idiosyncratic features of biological selection. Conceptual change isn’t merely analogous to natural selection; it is a selection process. According to Hull, selection processes involve two distinct kinds of entities – replicators and interactors – which are defined by their functional roles. Replicators pass on their structure through successive replications. By contrast, interactors are cohesive wholes that interact with their environment in a way that causes replication to be differential. Using these terms, Hull concludes that selection occurs when the differential success of interactors causes the differential perpetuation of replicators. In biological selection, genes are the most common sort of replicators, and a variety of entities (from genes to demes) can function as interactors (Hull 1980). In conceptual selection, a number of different kinds of “memes” function as replicators: substantive scientific beliefs about the world, beliefs about which methods are likely to work, beliefs about how and when to interact with other scientists, etc. These ideas can be carried in a variety of vehicles, including journal articles, conversations, lectures, textbooks, and scientists’ brains. While the ideas are carried in all these formats, scientists are crucial for the process of interaction. Scientists mediate the interaction of ideas with the world by setting up, performing, and interpreting the results of experiments. Furthermore, the way scientists interact in the social world influences the likelihood that an idea will be passed on. Thus, scientists (and perhaps scientific instruments) are the principal interactors in conceptual selection.2 By offering a general analysis of selection, Hull can reply to the critics who stress disanalogies between biological and conceptual selection. For example, many critics have argued that conceptual selection is fundamentally different from natural selection because whereas genetic mutations are “random”, new conceptual variants are introduced non-randomly. Even if conceptual and natural selection do differ in this way, Hull’s analysis helps us to see that this difference does not compromise the idea of conceptual selection; selection can occur whether or not the variants are randomly generated. (More on this below.) Although Hull’s general analysis of selection helps to set aside a number of disanalogies between natural and conceptual selec-
447 tion, it raises other issues. In particular, we must determine whether Hull’s proposed conceptual replicators and interactors play the same functional roles as biological replicators and interactors. According to Hull, selection processes require a very special relationship between replicators and interactors. Replicators bear information which can be preserved across successive replications; interactors are cohesive wholes which interact relatively directly with the environment. Differences in the interactors’ traits influence how well the interactor succeeds in its environment, thereby influencing its ability to pass on copies of the replicators. The questions I want to raise concern whether Hull’s general analysis adequately captures both the gene/organism relationship (paradigm replicators and interactors in the biological realm) and the idea/scientist relationship. These cases differ in two important ways. First, it is not clear whether ideas “code for” the traits of scientists in the way genes code for the traits of organisms. Replicators are thought to carry information – they are about something. Structural genes, an important class of replicators, are viewed as coding for the traits of an organism in the sense that differences in genes cause differences in organismic traits. But, given Hull’s view that scientists and scientific instruments are the principal interactors, it seems misleading to say that scientific ideas code for traits of interactors. Certainly, some scientific ideas are about how to build a more sensitive instruments or about how a scientist ought to behave in order to produce better theories. But many scientific beliefs are about the natural world, not about conceptual interactors. (See Sterelny 1994 for more on this issue.) The second issue concerns the nature of conceptual interaction. Biological interaction is complex. It involves a variety of entities at different hierarchical levels (e.g., genes, chromosomes, organisms, demes), each of which varies in a number of different properties that can influence the way that entity interacts with the environment. While biological interaction is far from simple, conceptual interaction appears significantly more complex. Conceptual interaction involves two distinct kinds of entities – scientists and scientific instruments – that don’t fall into a hierarchy where the larger entities encompass the smaller ones. Furthermore, conceptual selection seems to involve several distinct kinds of environmental interaction: ideas are tested for their fit with the natural world, with other ideas, and with elements of the social environment. Finally, scientists play a particularly complex role: they not only passively express the ideas in the way an organism phenotypically expresses its genes. Scientists also operationalize their concepts, actively create instruments to test their ideas, and interpret the results of the experiments. I don’t know whether these differences completely undermine the claim that biological interactors and “conceptual interactors” play the same functional role (I
448 remain uncertain). But the task of developing an adequate account of conceptual interaction will be difficult. Given these difficulties, I would like to explore the possibility of defending Hull’s social account without relying on the selectionist framework. 2.2. The second pillar: The dynamics of cooperation and competition In keeping with a general trend toward viewing science as fundamentally social (see e.g., Bloor 1976; Kitcher 1993; Longino 1990; Shapin 1982; Solomon 1994), Hull argues that the social structure of science is not “external” to science. Rather, he argues that the reward structure of science and the dynamics of cooperation and cooperation are the engines that drive conceptual change in science. It is hard to do justice to the nuances of Hull’s account in a short compass; I will merely sketch three of the primary themes. First, Hull describes the reward structure of science. While Hull talks about formal recognition (e.g., Nobel prizes), he is much more concerned with a different kind of reward: use. According to Hull, the highest compliment is to have one’s ideas actively used – preferably with an explicit citation! Thus, Hull spends considerable time discussing citation practices. In addition, he describes the norms governing the punishment of scientific misbehavior (e.g., the fact that fraud is punished more severely than stealing the work of a graduate student). Second, Hull describes social dynamics among scientists as a struggle to maximize “conceptual inclusive fitness.” Scientists want to receive credit for their ideas. Because the reward system of science gives all the credit to the first person to make a discovery, there is pressure for scientists to conceal results from others, to claim all the credit (even when others helped), and to take short cuts (which may lead to sloppy work). These “selfish” motives are held in check by three features of the reward system of science. (1) Scientists can’t receive credit without making their results public. Thus, scientists are encouraged to publish results. (2) In order persuade others of the correctness of their analyses, scientists need to gather support for their ideas – and this usually involves citing other influential work. (3) Severe penalties for publishing sloppy data create pressure to present solid evidence. As Hull puts it: “one cannot pursue credit without risking blame; nor can one cite the work of another author in support of one’s own research without conferring credit” (p. 312). Just the biological struggle to survive and reproduce can lead to (apparently) altruistic behavior, the struggle for credit can lead to scientific cooperation (in the form of positive citations and collaborations). In sum, Hull presents the social dynamics of science as a struggle in which individual scientists – motivated by curiosity and a desire for credit – try to maximize their inclusive fitness within a social environment which is shaped
449 by the reward structure mentioned above. (Section 3 will develop this theme at greater length.) Finally (this is the third theme), Hull integrates the first two elements into a “functionalist” analysis of science. The social structure of science is a carefully tuned functional system that effectively channels the motivations of scientists. Given the reward structure of science, scientists who try to maximize their own conceptual inclusive fitness behave in ways that make science successful. Hull’s description of the reward structure of science is similar to the functionalist sociology of Merton (1971) and Hagstrom (1965). But while the early functionalists were content to simply describe the normative structure of science, Hull aspires to explain why scientists adhere to the norms of science: “It is not enough to specify the norms that characterize [science]. Why do scientists adhere to these norms?” (p. 281). Here is Hull’s answer: Once one identifies the operative norms of research science, the explanation for the high frequency with which individuals adhere to these norms becomes obvious. It is in their self-interest to do so . . . The institution is so structured that by and large, [scientists] are rewarded for doing what they are supposed to do and punished when they do not (p. 320). Hull’s functionalism blends descriptive and normative claims. For example, Hull describes the reward structure of science and explains the evolution of the social structure of science. But in addition to these descriptive and explanatory claims, he also claims that the social structure explains the success of science. Thus, Hull’s functionalist account of the social structure of science provides an overarching framework that combines two distinct projects: (1) a descriptive (explanatory) account of the evolution of the social structure of science, and (2) a normative account of how the social structure of science contributes to the success of science. (In Section 4, I offer a fuller account of why claims about the success of science should be understood as normative.) Let me sharpen this preliminary account by showing how both projects are present in Hull’s discussion of the demic structure of science. 2.3. The demic structure of science A biological “deme” is a population of conspecific organisms. Often these organisms share a discrete geographical location, but geographical isolation is not essential. What is essential is that the members of the deme are more likely to engage in ecological or reproductive interactions with members of that group than with members of other demes. To say that a species has a demic (or population) structure just means that the species is not one homogeneously interbreeding group, but is sub-divided into relatively distinct
450 breeding populations. Similarly, to say that science displays a demic structure means that scientific disciplines (analogs to species) are subdivided into smaller units. As Hull notes, The scientific community is not a seamless whole. It is broken down into nested and overlapping informal groups, what I have termed the demic structure of science. The smallest unit in science is the individual scientist. The next largest is the research team whose members are in daily contact. Scientific ‘demes’ are much larger, consisting of dozens to several hundred scientists, depending on the discipline. Invisible colleges are the largest groups of all (p. 366). Scientists in the same deme engage in more direct cooperation and competition (i.e., ecological interaction). For example, Hull treats the pheneticists and the cladists as two competing demes within the broader systematics community; scientists within each deme rely on one another’s ideas more than they depend on the ideas of scientists in opposing groups (p. 435). According to Hull, the fact that scientists are organized into groups influences the dynamics of science in several ways. The demic structure of science accelerates the pace of scientific change, provides resources to develop an idea through sympathetic criticism, encourages severe testing of ideas proposed by competing demes, and even influences the direction of scientific change. Hull often presents these claims in functionalist language. For example, he claims that “One of the chief functions of intellectual demes is to allow, even encourage, severe testing” (p. 378; see also p. 366). In sum, Hull’s theory is held together by a kind of functionalist sociology of science. His functionalism provides the context for integrating a variety of social and selectionist claims. Consider, for example, Hull’s explanation of the demic structure of science: demes have become common because scientists who form research groups are better able to develop and defend their views.3 As we’ll see in the next section, this explanation involves selection in a particular kind of social environment – suggesting close interconnections between the selectionist and social aspects of his theory. In addition to these essentially descriptive and explanatory aims, Hull’s functionalist framework also provides the setting for a variety of normative claims (e.g., demes contribute to scientific progress by providing a sympathetic audience for the early development of ideas and by encouraging severe testing of ideas in the broader scientific community). This, in a nutshell, is Hull’s account of science. The next two sections of the paper will develop this sketch more fully, focusing particularly on how the social and selectionist aspects of Hull’s account are related.
451 3. Explaining demes: The origin of demic structure Hull explains the origin of demic structure in selectionist terms. Scientists have ideas that influence their social behavior. Some of these ideas, such as ideas about cooperating in research groups, generally lead to more productive research. Thus, ideas that promote cooperation are selected and become more common among scientists. Although Hull often discusses demic structure in selectionist terms, I maintain that the key features of Hull’s account of the origin of demes can be preserved in a social idiom that does not depend on the controversial details of Hull’s model of conceptual selection. If I am right, then criticisms of conceptual selection do not undermine Hull’s explanation of the origins of the demic structure of science. Let me begin by offering a more detailed statement of the selectionist explanation. In the course of their graduate education, budding scientists not only learn substantive chunks of science, they also absorb ideas about how to conduct research. Either through verbal instruction or through observation and imitation, graduate students develop a set of beliefs about how and when to engage in various forms of cooperation: sharing pre-prints and research proposals, refereeing papers and grant proposals, citing colleagues, co-authoring papers, etc. The ideas learned in graduate school are not the only relevant influences, of course. Insofar as personality traits like introversion/extroversion have a genetic basis, biological traits will influence scientific behavior. In addition, graduate students bring to their education a wide variety of other beliefs which influence their behavior. But this, Hull would say, is really no different than a gene influencing the phenotype against the backdrop of other genes in the genome. The process of graduate education is, then, one of the primary ways in which ideas are “inherited.” The end result is variation in inherited beliefs about how to successfully interact with other scientists. To complete his case for selection, Hull needs to demonstrate that these ideas affect conceptual fitness (i.e., the likelihood that an idea will survive and reproduce). Hull does not suppose that all forms of cooperation are selectively advantageous. Nonetheless, Hull (p. 286) and other sociologists (Blau 1978; Hagstrom 1965) argue that scientists who participate in research groups and/or co-author papers are the most productive. And, since productive scientists generally have greater access to graduate students, scientists who form demes will generally be more successful in passing on both their substantive ideas and their ideas about the value of forming research groups. Thus, it seems that the differential success of the interactors (scientists) will lead to the differential perpetuation of replicators (ideas that influence scientific behavior).
452 Is Hull’s explanation for the origin of the demic structure of science successful? There are at least two major worries about this account. The first challenges the completeness of his account, the second presents a more serious challenge. The completeness of Hull’s account. Hull suggests that scientists form a nested hierarchy of social groups: individual scientists form research groups, which in turn are grouped into larger demes, etc. While it is hard to deny that much of science is organized in this fashion, Hull presents data which suggest that in some fields, up to 50% of scientists operate as “isolates” who are not affiliated with any “clique” or research group (see also Blau 1978; Hagstrom, 1965).4 These findings present a serious problem for Hull. If conceptual selection is an effective force and if it generally favors at least some cooperation, then most scientists should respond to this selection pressure by forming research groups. Since conceptual selection seems to strongly favor research groups, it is doubtful that any selection explanation will suffice to explain why so many scientists do not form research groups. Hull’s selection account does not seem to answer all the relevant questions about demic structure. The intentionality objection. Critics often argue that evolutionary epistemologies fail because they ignore the intentionality of scientists (e.g., Oldroyd 1990; Sterelny 1994; Thagard 1980; Waters 1990). According to the selection account, scientists get their ideas about cooperation largely from memetic inheritance or through “mutations” in ideas. Memes for cooperating have increased in frequency because those individuals in past generations who cooperated were more successful in passing on memes to the current generation. But this account ignores (or at least discounts) the idea that scientists consciously and intentionally try to solve the problems they face. One way to frame the objection is to ask whether novel ideas are really “random mutations” or whether they arise through our conscious effort to solve problems. Consider an early modern scientist who is facing a complex scientific problem. He learns of another scientist who has developed a laboratory technique that may be usefully modified to address the problem he is working on. These scientists elect to cooperate and ultimately solve the problem using their combined expertise. The idea of cooperation seems to be most naturally explained as a function of human innovation and resourcefulness, not a history of selecting out those random variants which (coincidentally) happened to work best. This objection is complex, so I will not provide a definitive discussion. Briefly discussing this objection will, however, provide resources that clarify the role of the conceptual selection within Hull’s
453 account of science. One important version of this objection focuses on Campbell’s (1960) characterization of selection processes as involving “blind variation and selective retention.” Genetic mutations are “random” in the sense that the range of new variation is not correlated with selective pressures. While the thesis of randomness is well established in the biological realm, it seems implausible to regard new conceptual variants as “randomly generated.” Ideas are brought forward to solve a problem. Scientists don’t randomly scour the indefinitely large domain of conceptual space; they focus on particularly promising ideas. While this objection has considerable force, it is hard to assess the extent to which the process of producing new theories really hinges on non-random (directed) variation. The historical record concerns only a very small and select set of ideas: those ideas that were considered important enough to write down. Further, in looking at this historical record, we often emphasize the successful and new ideas, ignoring the mass of unsuccessful ideas (which might suggest more “random scattering” in conceptual space). More significantly, selection can occur even when directed mutation is occurring. For example, a small bias toward useful variations might be supplemented by selection of those useful mutations. In such a case, selection contributes to the process of adaptation without being the sole cause of adaptation. Thus, it seems possible to retain a more humble account of conceptual selection in which the self-conscious (non-random) search for novel ideas and subsequent selection of ideas both contribute to conceptual change in science. The more humble selectionist account – which recognizes the role of intentionality in scientific behavior – goes something like this. Scientists do not rely solely on the ideas they’ve “inherited” or on random mutations of those ideas. Instead, in their search to solve specific theoretical and experimental problems, scientists often look to see which social strategies have been successful in solving similar problems. Because scientists search for strategies that are tailored to their social and intellectual situations, this search will be decidedly non-random. When we look at this phenomenon from the perspective of the broader community of researchers, we can see a two stage process at work. First, the intentional search for effective strategies ensures that the pool of conceptual variants is not randomly generated. Second, the social process of critically evaluating new ideas selects out and differentially replicates the most successful ideas. Once we open the door to this more humble explanation, however, it seems possible to abandon the selectionist account altogether. The alternative (non-selectionist) account will be cast in terms of individual scientists pursuing their desires. Armed with Hull’s description of the reward structure, we explain the formation of demes as a result of scientists’ desire for credit and their beliefs about which strategies are likely to
454 be successful given the prevailing reward structure. In this way, both stages of the humble selectionist account can be incorporated into a non-selectionist idiom. Scientists cull out (“select”) those forms of behavior which appear most successful. [Because this account presupposes Hull’s rich description of the social and reward structure of science, it is not strictly an individualist explanation (see Kincaid 1997).] The point of these reflections should now be obvious: it is possible to retain the basic structure of Hull’s explanation without relying on the selection framework. The reward structure of science coupled with the desire for credit leads scientists to form demes. This explanation is independent of the selection account in the sense that we can accept it without resolving the problems that have plagued Hull’s analysis of conceptual selection. Hull would, of course, argue that there are benefits to viewing this process in selectionist terms. His theory provides a unifying framework that explains a wide variety of facts about the social organization of science. But the benefits of unification must be weighed against the well-known difficulties of fully articulating the selectionist account (Section 2.1). It is not my intention to weigh the costs and benefits of the selectionist framework. Rather, my aim is simply to clarify the role of conceptual selection within Hull’s more general theory. The lesson, I maintain, is clear: although Hull describes the evolution of demic structure in selectionist terms, his insights into the social dynamics of science can be preserved in a non-selectionist idiom.
4. Evaluating demes: The social epistemology of demes In this section, I examine and defend the claim that the demic structure of science contributes to the growth of scientific knowledge. Ignoring his invective against epistemology, I argue that Hull’s position is, at bottom, a form of “conservative social epistemology” (Kornblith 1994). Furthermore, while Hull’s social epistemology is intimately connected with an evolutionary view of science, it is independent of the details of Hull’s selectionism. Hence, the criticisms of Hull’s selection account do not undermine his social epistemology. 4.1. Conservative social epistemology Hull presents himself as developing an empirical “science of science”: I find a scientific theory of sociocultural evolution a vastly more significant goal that an evolutionary epistemology . . . My use of the appellation ‘evolutionary epistemology’ in this paper should not be taken as an
455 endorsement of any epistemologcal views whatsoever. If evolutionary epistemology were a genuine epistemological theory, I would not be in the least interested in it (1982, pp. 273–274). More pointedly, he asserts that the main problem with past work in evolutionary epistemology is not that it is evolutionary but that it is epistemology. As far as I can see, neither the content nor the methods of science can be “justified” in the sense that generations of epistemologists have attempted to justify them (pp. 12–13). Thus, Hull appears to disavow epistemology in favor of a descriptive (nonnormative?) science of science. But lurking beneath the anti-epistemology rhetoric, we find that Hull is committed to a number of normative epistemological theses (Gatens-Robinson 1993). For example, in a telling section entitled “What needs explaining” (pp. 285–303), Hull offers a long list of questions about science that, in his view, any adequate theory of science should answer. One of the central questions is: Why does science work so well? On Hull’s view, science aims to discover true and significant generalizations about the physical world. To presume (as Hull does) that science succeeds in its aims is to assume that science discovers significant truths about the world. Furthermore, Hull maintains that science is “globally progressive” (p. 464) and that the social structure of science contributes to the objectivity of science (pp. 3–4). By emphasizing normative epistemological notions of truth, progress, and objectivity, Hull turns his back on a strictly causal/explanatory account of demes. The remainder of this section argues that Hull’s normative project can be usefully seen as a form of social epistemology.5 Conservative social epistemology is a logical extension of the broader program of naturalizing epistemology (Kornblith 1994). According to the naturalists, scientific studies of psychological processes have epistemic import. For example, by helping us to understand biases in our reasoning processes, empirical studies of cognition contribute to the project of developing normative methodological guidelines. Just as the empirical study of psychological processes can reveal the epistemic strengths and weaknesses of certain cognitive processes, empirical studies of group processes can be epistemically revealing. Kornblith treats this program as a conservative program because, unlike advocates of the “strong programme” who suggest that we should set aside the notion of truth (e.g., Bloor 1976), this project is concerned to determine how social structures can either enhance or detract from our ability to achieve epistemic aims such as truth and objectivity. Hull’s discussion of the “functions” of demes is an exemplar of conservative social epistemology. As we have seen, Hull is concerned to show that the demic structure of science influences how well the scientific community
456 achieves its epistemic aims. In particular, he emphasizes three principal effects: 1. Biological species which are subdivided into semi-isolated populations evolve more quickly and are able to reach higher adaptive peaks. Similarly, a community divided into intellectual demes will explore the space of possible theories more quickly (see pp. 361, 388, 507). 2. Demes provide a safe place to try out new ideas and develop them before they face severe criticism: “One function of [research groups and demes] is to provide sympathetic criticism while a scientist develops his or her ideas” (p. 366). 3. Demic structure enhances critical inquiry because members of competing demes are likely to carefully test one another’s theories: “One of the chief functions of intellectual demes is to allow, even encourage, severe testing” (p. 378). These references to the “functions” of demes are epistemologically loaded. Hull is claiming that demes improve our ability to identify true (or at least promising) theories, eliminate false theories, and to do both of these things as quickly as possible. A community which is subdivided into demes provides small and relatively protected “niches” in which new theories can develop, but also provides for serious interdemic criticism to insure that ideas are retained only if they pass rigorous tests. Further, the division of labor will allow scientists to more effectively explore alternative theories than will a homogeneous (roughly Kuhnian) community in which all scientists accept all aspects of the disciplinary matrix.6 In many ways, Hull’s project is similar to Kitcher’s work on the division of cognitive labor. Kitcher (1990, 1993) claims that the way a community of inquirers is structured can influence its ability (as a community) to succeed in its epistemic aims. As one illustration, Kitcher explores a case in which a community is trying to discover the structure of a molecule. Suppose that there are two recognized methods for discovering the structure of the molecule, but that one is generally thought to be the better method. Each scientist (if he’s concerned to discover the truth) should pursue the better method. But a community that divides labor between the two methods will (under some conditions) be more likely to discover the structure of the molecule. Thus, dividing the cognitive labor – creating different groups that pursue different methods – can enhance our collective intellectual enterprise. Kitcher’s presentation of these ideas is markedly different from Hull’s: whereas Hull offers a rich (but non-quantitative) description of social dynamics, Kitcher uses formal models to address normative methodological issues. But beneath these differences, both identify ways in which social
457 structures influence the ability of the scientific community to succeed in its epistemic aims.7 4.2. Objections and replies Hull does not adequately develop his normative, epistemological analysis of science. Given Hull’s disavowal of epistemology, this may be a worry about my interpretation of Science as a Process rather than a criticism of Hull’s position. But insofar as Hull actually aims to address normative issues, this criticism has merit. Contrasting Hull’s claim that the demic structure of science contributes to the objectivity of science with Helen Longino’s social epistemology will help to clarify both the strengths and weaknesses of Hull’s approach. One important lesson to draw from this discussion is that even if Hull does not fully develop his account of the social epistemology of demes, empirical work on the social structure of demes is a crucial resource for future work in social epistemology. Hull asserts that “the objectivity that matters so much in science is not primarily a characteristic of individual scientists, but of scientific communities” (pp. 3–4). The idea is that even when most or all individual scientists are, say, biased in favor of their own hypotheses, the decisions reached by the scientific community can be objective: “The self-correction so important in science does not depend on scientists presenting totally unbiased results but on other scientists, with different biases, checking them” (p. 321). Despite the fact that the ideal of objectivity has been conceived in a variety of ways, Hull never presents an analysis of this controversial concept. By contrast, Longino (1990) presents a fuller analysis of objectivity that recognizes several distinct ways of conceptualizing objectivity. According to Longino’s contextual empiricism, all assessments of empirical data rest on (often implicit) background assumptions which may be value-laden. As a result, neither individual scientists nor scientific communities can guarantee that their conclusions are completely free of bias. While scientific communities cannot insure complete objectivity (i.e., complete freedom from bias), various social mechanisms can increase the objectivity of scientific inquiry. For Longino, the key is cultivating social mechanisms that allow the community to identify and critically examine its own background assumptions. Even though no individual or social group is completely free of bias, if the scientific community is organized in ways that increase the chances of identifying and critically examining our biases, then we can increase the objectivity of scientific inquiry. The connection to Hull’s analysis ought to be obvious: The demic structure of science is one mechanism for increasing intellectual diversity. Just as small, relatively isolated biological populations can “drift” away from the population, relatively isolated intellectual demes
458 will promote a greater diversity of criticism by allowing small groups to explore novel avenues of research. As Kornblith (1990) suggests, social epistemology must be grounded in careful empirical study. Even if the normative project is not sufficiently developed, Hull’s empirical work is a crucial resource for social epistemology. Hull presents the social structure of science as a complex functional system in which cooperation and competition are carefully balanced. This perspective has important implications for normative proposals to alter the balance between competition and cooperation: The functional perspective does lead one to be somewhat cautious in attempting to change a system . . . To the extent that a system is functionally organized, changes are sure to ramify, and these ramifications may be extensive, not to say unpredictable. Unless one is willing to risk the destruction of the system that one wants to change, caution is called for (pp. 355–356). Careful empirical study of the role(s) of competition is a necessary element in assessing proposals for a more cooperative social structure. Hull’s social and evolutionary models are inadequate because they abstract away from individual cognitive processes. As we have seen, Hull argues that the objectivity of science flows from the social organization of the scientific community. Hull’s social explanations do not depend on the cognitive states of individual scientists; instead, they assume only that scientists are motivated by curiosity and the desire for credit. Many critics of Hull’s social epistemology insist that a social account is not adequate. For example, Sterelny (1994) argues that the dynamics of science turn on the intentional characteristics of the participants in the domain. Hence we cannot idealize away from the intentional profile of scientific agents. That profile mediates both interaction and replication and plays a role in explaining the constraints on selection and the reasons for progress (p. 60). Similarly, Waters claims that The basic problem with the evolutionary model is that it leaves out cognition . . . [A]ny epistemology that is based on the analogy to biological evolution will tend to hide (by abstraction) the intentionality behind conceptual change and hence draws our attention away from the cognitive mechanisms (1990, p. 84). These worries are closely related to the intentionality objection. But whereas the intentionality objection is usually viewed as a criticism of Hull’s selection
459 model, I want to determine whether intentionality poses a problem for Hull’s social epistemology. Understanding individual-level cognition is part of any complete naturalistic account of science. As we’ve already seen, individual cognition and intentionality are important aspects of the process of scientific discovery. Sterelny (1994) develops this idea by suggesting that a scientist’s ability to recognize “promise” in a theory makes conceptual selection significantly different from biological selection. Whereas biological selection only takes into account realized fitness, scientists often pursue promising theories – even when they’re currently less “fit.” Any approach which completely abstracts away from individual cognition will miss this important fact. But similarly, any strictly individualistic account of science will also fail. As we have seen, Hull presents considerable evidence that intellectual demes are crucial components of any descriptively or normatively adequate account of science, a conclusion which is strongly reinforced by other work in social epistemology (e.g., Kitcher, Longino, Solomon). These brief reflections cry out for additional development, but on this occasion I will simply suggest that we have good reason to pursue an inter-level theory in which some of the explanatory and normative work is accomplished at the level of individuals, and some is accomplished by distinctively social theories (Kincaid 1997; Kornblith 1994). Hull seems to embrace this conclusion himself when he suggests that “we need a hierarchical theory of science” (p. 293). 4.3. Conceptual selection and conceptual evolution In my view, the social epistemology of Science as a Process does not depend on Hull’s analysis of conceptual selection in terms of replicators and interactors. Consider the claim that a subdivided intellectual community will lead to more robust criticism. This idea can certainly be formulated in ways that sound selectionist. But, as with Hull’s explanation of demic structure, the appeal to selection is not essential. Interdemic criticism will not be effective in leading to the truth unless scientists can conduct empirical tests that “select” those ideas which best fit with nature. This confrontation with data is clearly necessary for the kind of selection Hull imagines. But one need not invoke all the details of Hull’s interactor/replicator analysis of selection in order to understand this testing (selection) process. Thus, I maintain that Hull’s social epistemology can and should be evaluated independently of Hull’s claim that science is a selection process. While Hull’s social epistemology does not require a selection account, it is intrinsically evolutionary. In biology, “evolution” has two central meanings: change in gene frequency over time and descent with modification. Hull’s project in social epistemology is evolutionary in both senses. First, once we
460 recognize the intellectual diversity within scientific communities, then we are led to view conceptual change as a change in frequency of certain beliefs within scientific communities. Science as a Process provides a rich body of evidence on the diversity of opinion within research programs. Grand research programs like Darwinism quite openly house significant differences of opinion. What is more surprising, though, is the degree of diversity even within much smaller demes and research groups (e.g., the numerical taxonomists and the cladists). While almost any two members of a research program agree on a large number of points, they will also disagree on a few points. And the points of agreement are not uniform throughout the community. Thus, there is no set of “essential” claims which defines the research program. Instead, the community is composed of a set of researchers who hold a variety of different versions of the theory – versions that overlap and diverge to varying degrees. Thus, research programs evolve as various elements of the theory either increase or decrease in frequency within the population of researchers. In Mayr’s terminology, Hull’s social epistemology requires us to abandon essentialist models of conceptual change in favor of “population thinking.” Second, theories and research programs should be seen as historical entities that evolve over time. Based on his case studies, Hull concludes that scientists never present their views fully formed. Science is a social process in which scientists evaluate and criticize each other’s work, leading to successive improvement. The appearance of some great scientist presenting a particular revolutionary view fully formed in a single seminal work is primarily a function of retrospective reconstruction (pp. 361–362). Theories and research programs are historical entities; they evolve in response to social interactions such as referee’s reports (for grants and journals), published criticisms, criticisms at conferences, and informal feedback from colleagues. As the theory is passed on, it is transformed (perhaps in a number of different, even incompatible, ways) through criticism. Hull’s case studies provide considerable evidence that any descriptively adequate theory of science will be evolutionary in these senses.
5. Conclusion The principal motivation of this study has been to clarify the relationship between Hull’s account of the social dynamics of science and his selectionism. Let me begin by emphasizing ways in which the two aspects of his position are intertwined. First, the selection explanation of the origin of demic
461 structure presupposes Hull’s description of the reward structure of science. In a sense, Hull’s account of the reward structure describes the ecology of conceptual selection: it enriches our understanding of the environment within which selection occurs. But not all features of the social environment are simply presupposed. By explaining some dimensions of the social environment (e.g., the demic structure of science) in selectionist terms, Hull provides a second (stronger) link between the social and selectionist projects. In spite of these connections, worries about the viability of Hull’s selectionist account led me to argue that some elements of Hull’s project can be retained without the selectionist framework. The burden of this paper has been to draw attention to aspects of Hull’s account which have not yet attracted widespread attention but which hold out real promise. In particular, Hull provides a detailed description of the demic structure of science and, based on this account, argues that demic structure is epistemically significant. Together, these elements point toward a fruitful project in social epistemology which does not depend on the more controversial elements of Hull’s account of conceptual selection. Although Hull’s project is rooted in a detailed empirical account of the social structure of science, his naturalism does not prevent us from developing a normative stance. Rather, we can view Hull’s work as a form of conservative social epistemology – an epistemology in which we try to determine the epistemic strengths and limitations of different forms of social organization in science. In short, I am pressing for a program that is less ambitious than Hull’s but which, to my mind at least, has greater chance of succeeding. While Hull’s social epistemology does not depend on the details of his model of conceptual selection, it is committed to an evolutionary perspective on conceptual change. By highlighting the conceptual variation within even the smallest research groups, Hull forces us to abandon essentialist accounts in favor of population thinking. Further, he suggests that future research in science studies should treat both social groups and theories as historical entities. If my assessment of Hull’s project is even approximately correct, then criticisms of Hull’s account of conceptual selection miss much of the insight in Science as a Process. Acknowledgments Thanks to Steve Downes, Paul Griffiths, and David Hull for helpful comments on an earlier draft of this paper. The relationship between biological parents and offspring is not always easy; the same holds true for the relationship between conceptual parents/offspring. But David has graciously endured my
462 intellectual adolescence. Even after suffering through countless drafts full of half-baked ideas, he has remained supportive of me and my work. Thanks, David.
Notes 1 All page references are to Hull (1988) unless otherwise noted. 2 In his recent work, Hull has begun to shift the emphasis from entities (repli-
cators/interactors) to the processes of replication and interaction (see, e.g., Hull et al. 2000). 3 I remain uncertain whether Hull views this as individual or group selection. On the face
of it, Hull provides an individual selection explanation: individual scientists who participate in groups have greater conceptual inclusive fitness. Alternatively, one might offer a group selection account in which more tightly integrated or more cooperative groups tend to outperform less well-organized research groups. (Hull hints at this kind of picture on p. 395.) Given the complex and on-going dispute about the relationship between kin selection and group selection, Hull can be forgiven for not definitively answering this question. But in the long run, “memeticists” must provide techniques for answering this question. 4 The literature I cite is, admittedly, out of date. As far as I can tell, sociologists of science no longer collect this sort of data. 5 Hull’s principal aim is to develop an adequate descriptive account of science. A solid empirical account of science would be a significant accomplishment, even if it did not address normative issues. Thus, I am not (like some critics of naturalism) insisting that good philosophy must address normative issues. I simply think it is important to identify and critically examine Hull’s implicit epistemological project. 6 Hull regards it as an open question whether current demic structures strike the best possible balance between criticism and cooperation. The intense inter-demic rivalry that Hull documents certainly has costs, including priority disputes and unfair parodying of one’s opponents. Nonetheless, Hull maintains that demic structure contributes to the success of science. It is precisely because a number of significant empirical and normative questions about demic structure remain unanswered that I think Hull is pointing us toward a rich and significant project in social epistemology. 7 The position I am defending is a version of normative naturalism, but is less ambitious than the position Laudan (1990) defends. In contrast to Laudan (who uses empirical experience to determine which goals we ought to pursue), neither Hull nor Kitcher tries to derive normative claims from empirical claims. 8 Although Waters and Sterelny both argue that it is a mistake to abstract away from the cognitive processes of scientists, they arrive at rather different assessments of Hull’s project.
References Blau, J.R.: 1978, ‘Sociometric Structures of a Scientific Discipline’, Research in Sociology of Knowledge, Sciences, and Art 1, 91–206. Bloor, D.: 1976, Knowledge and Social Imagery, Routledge, London. Bradie, M.: 1986, ‘Assessing Evolutionary Epistemology’, Biology and Philosophy 1, 401– 459.
463 Campbell, D.: 1960, ‘Blind Variation and Selective Retention in Creative Thought as in Other Knowledge Processes’, Psychological Review 67, 380–400. Donoghue, M.: 1990, ‘Sociology, Selection and Success: A Critique of David Hull’s Analysis of Science and Systematics’, Biology and Philosophy 5, 459–472. Gatens-Robinson, E.: 1993, ‘Why Falsification is the Wrong Paradigm for Evolutionary Epistemology: An Analysis of Hull’s Selection Theory’, Philosophy of Science 60(4): 535–557. Grantham, T.: 1994, ‘Does Science have a Global Goal?: A Critique of Hull’s View of Conceptual Progress’, Biology and Philosophy 9, 85–97. Hagstrom, W.O.: 1965, The Scientific Community, Southern Illinois University Press, Carbondale, IL. Hull, D.L.: 1980, ‘Individuality and Selection’, Annual Review of Ecology and Systematics 11, 311–332. Hull, D.L.: 1982, ‘The Naked Meme’, in H.C. Plotkin (ed.), Learning, Development, and Culture, Wiley & Sons, New York, NY, pp. 273–327. Hull, D.L.: 1988, Science as a Process, University of Chicago Press, Chicago, IL. Hull, D.L., Langman, R. and Glenn, S.S.: 2000, ‘A General Account of Selection: Biology, Immunology and Behavior’, Behavioral and Brain Sciences (to appear). Kincaid, H.: 1997, Individualism and the Unity of Science, Rowman & Littlefield, Lanham, MD. Kitcher, P.: 1988, ‘Selection among the Systematists’, Nature 336, 277–278. Kitcher, P.: 1990, ‘Division of Cognitive Labor’, Journal of Philosophy 87, 5–22. Kitcher, P.: 1993, The Advancement of Science, Oxford University Press, Oxford. Kornblith, H.: 1994, ‘Conservative Social Epistemology’, in F. Schmitt (ed.), Socializing Epistemology, Rowman & Littlefield, Lanham, MD, pp. 93–110. Laudan, L.: 1990, ‘Normative Naturalism’, Philosophy of Science 57, 44–59. Longino, H.: 1990, Science as Social Knowledge, Princeton University Press, Princeton, NJ. Merton, R.K.: 1973, The Sociology of Science, University of Chicago Press, Chicago, IL. Oldroyd, D.: 1990, ‘David Hull’s Evolutionary Model for the Progress and Process of Science’, Biology and Philosophy 5, 473–487. Ruse, M.: 1989, ‘Great Expectations’, Quarterly Review of Biology 64, 463–468. Shapin, S.: 1982, ‘History of Science and its Sociological Reconstruction’, History of Science 20, 157–211. Solomon, M.: 1994, ‘Toward a More Social Epistemology’, in F. Schmitt (ed.), Socializing Epistemology, Rowman & Littlefield, Lanham, MD, pp. 217–233. Sterelny, K.: 1994, ‘Science and Selection’, Biology and Philosophy 9, 45–62. Thagard, P.: 1980, ‘Against Evolutionary Epistemology’ in P.D. Asquith and R.N. Giere (eds.), PSA 1980: The Proceedings of the 1980 Biennial Meeting of the Philosophy of Science Association, Philosophy of Science Association, East Lansing, MI, pp. 187–196. Waters, C.K.: 1990, ‘Confessions of a Creationist’, in N. Rescher (ed.), Evolution, Cognition, and Realism, CPS Publications, University Press of America, Lanham, MD, pp. 79–90.