1. Introduction
Philosophers of science are increasingly attentive to scientific practice: its material culture, uses of models, and development of methods and technologies. One effect of this ‘practical turn’ is reexamination of oppositional dualisms that have traditionally delineated the domain of philosophy of science: justification versus discovery, fact versus value, rational versus social (e.g., Kitcher Reference Kitcher2001; Longino Reference Longino2002; essays in Schickore and Steinle Reference Schickore and Steinle2006). In this article, I reexamine an oppositional dualism that still holds powerful sway: social construction versus normative epistemology of science. I argue that if we aim to critically evaluate (and perhaps improve) scientific practice as a means of knowledge production, then we should endorse a version of social constructivism. Critical evaluation, rather than selective description or indiscriminate validation, has long been the aim of philosophy of science. Such evaluation requires epistemic norms, or standards. The challenge for philosophy of science, faced with the enormous diversity and complexity of scientific practices past and present, is to articulate epistemic norms that are useful—that engage “the kinds of contexts and inquiries we actually confront,” without restricting critique to narrowly circumscribed case studies or evaporating into trivial platitudes (Mayo and Miller Reference Mayo and Miller2008, 306). In this article, I argue that social constructivism, widely taken to oppose normative epistemology, can help us meet this challenge.
This claim needs some qualification. The term ‘social construction’ is notoriously ambiguous, encompassing diverse theses (see Hacking Reference Hacking1999). I begin by disambiguating varieties of social constructivism about scientific knowledge (Section 2). This distinguishes the social constructivist thesis I defend from its metaphysical cousins, which have heretofore received more attention. The thesis I defend, as part of an effective critical epistemology, is that standards of epistemic justification (SCj) are constructed in scientific practice. Using Ludwik Fleck's comparative epistemology as an exemplar, I clarify the notion of ‘standards of epistemic justification’ and explicate the claim that these standards are socially constructed (Section 3). The thesis of SCj is that epistemic standards for science are established and maintained by value-laden social interactions among practicing scientists; epistemic justification is embedded in the social organization of scientific communities. SCj is thus an alternative to abstract, idealized accounts of epistemic justification.
Having set out this alternative, I then argue that we should take it. The argument has a positive and a negative part. I begin with the negative: we cannot both reject SCj and effectively critique scientific practices (Section 4). Effective critique requires epistemic standards with clear application to our practices. But if SCj is rejected, a gap opens between philosophers’ standards for science and those at work in scientific practice. This vitiates evaluation of the latter by the former. Conversely, insofar as epistemic norms do apply to cases of scientific practice, SCj cannot be decisively rejected. Using accounts by Laudan (Reference Laudan1987), Solomon (Reference Solomon2001), and Longino (Reference Longino2002) as exemplars, I show that even epistemic norms designed to apply to social history of science do not overcome this dilemma. Philosophers who aim at critical epistemology of science should thus endorse SCj.
This may seem too great a concession, given tensions between SCj and normative epistemology. Here my positive argument begins (Section 5). I propose an analytic framework for studying our scientific practices, based on general analyses of social action by Gilbert and Bratman. Within this framework, SCj can be fleshed out with sociohistorical studies to yield robust epistemic ideals implicit in our scientific practices. These epistemic ideals ground critical evaluation of those practices. What undermines such critique is not SCj but identification of epistemic standards for science with standards accepted as justificatory in specific sociohistorical contexts. One epistemic ideal derived from the SCj–social action framework (a version of scientific objectivity) rebuts this identification. So a thesis associated with social constructivism that does undermine critical epistemology is excluded by the SCj–social action approach. I conclude that the latter is our best starting point for a critical epistemology of scientific practice. The final section (Section 6) summarizes the overall argument.
2. Varieties of Social Construction
Though the ‘Science Wars’ have largely subsided, this owes more to exhaustion than to rapprochement. Opponents have retrenched behind respective disciplinary lines. The phrase ‘scientific knowledge is socially constructed’ tends to elicit either knee-jerk agreement or dismissal, depending on which camp one is in. Yet the slogan is more than a litmus test for disciplinary identity. Among its ambiguous implications are substantive and challenging theses. Arguments for and against social construction of scientific knowledge have often been in a metaphysical key, focusing on questions of realism, objectivity, and truth.Footnote 1 Another strand of debate examines the vexed relation between ‘the rational and the social’, focusing on methodological assumptions in the human sciences.Footnote 2 These debates expose differing metaphysical and methodological assumptions that motivate acceptance or rejection of constructivist claims. But in the welter of competing assumptions and scandalous pronouncements, certain epistemological aspects of social constructivism have been comparatively neglected. In this section, I disambiguate the social constructivist slogan so as to bring out these aspects and distinguish a variety of social constructivism that offers resources for critical epistemology of scientific practice.Footnote 3
On a minimal interpretation, if something is socially constructed, then it has undergone a process of development in some way mediated by social structures, interactions, or values. That is, it has a history in which human sociality is implicated. This weak constructivist thesis (though inheriting all the ambiguity of ‘social’) is uncontroversial. Evidently, scientific knowledge is not given to us but results from our collective effort. These efforts necessarily involve social structures, interactions, and values. Controversy over social construction of scientific knowledge begins with a stronger thesis: that social structures, interactions, or values play a constitutive role in the development of scientific knowledge, such that this knowledge is of our making in an epistemologically significant sense. This stronger constructivist thesis is too vague to be fruitfully assessed. Here the familiar (though notoriously inadequate) tripartite analysis of knowledge is helpful (cf. Boghossian Reference Boghossian2006, 22–23). If any necessary component of scientific knowledge qua knowledge is socially constructed, then scientific knowledge is socially constructed in a strong and interesting sense. Three theses can thus be distinguished:
(1) Scientific facts/truths are socially constructed.
(2) The beliefs of scientists (true or false) are socially constructed.
(3) Scientific standards of justification are socially constructed.
Philosophers have tended to focus on (1), which engages issues of scientific realism, reference, and representation (e.g., Giere Reference Giere1999; Goldman Reference Goldman1999; Lewens Reference Lewens2005; Boghossian Reference Boghossian2006). Though heavily criticized, the thesis is not nonsensical or self-defeating. The constructivist idea is not that we as social beings collectively extrude the material world like spiders spinning a web. It is rather that our interpretative webs, spun to capture features of the world, are necessarily shaped by our agency within it. This agency is socially as well as materially engaged. Scientific facts, as the outcome of these interactions, represent specific worldly features in a way that is necessarily influenced by social interests and values. The opposing view, objectivism about worldly facts or truths, conceives epistemic agency as passive: facts/truths are (largely) independent of our beliefs.Footnote 4 Objectivism of this sort is presupposed in most scientific (and everyday) practice. But it is not obvious that a philosophy of scientific practice must presuppose it. I shall not press the issue here, however. What follows does not presume (1).Footnote 5
Thesis (2) concerns rationality of scientific beliefs. The best-known interpretation of this thesis is that of the Strong Programme: satisfactory explanations of beliefs (true or false) must invoke social values or interests rather than absolute (context-independent) requirements of rationality (Barnes Reference Barnes1972, 373).Footnote 6 In this very strong form, (2) is difficult to defend: its rational/social dichotomy is too restrictive, while the nonexistence of absolute requirements of rationality cannot be decisively proven. It is tempting to deflate controversy over (2) by simply distinguishing between two projects: sociological description and explanation versus critical and evaluative philosophy of science.Footnote 7 Certainly the aims and methods of sociology of scientific knowledge differ from those of mainstream philosophy of science. But its challenge to the latter cannot be so easily deflected. Nonetheless, I will not defend (2). Controversy over sociology of scientific knowledge is not about beliefs, as such, but the epistemic standards by which those beliefs are evaluated. Barnes and Bloor cast their argument in terms of rationality, but their targets are context-independent epistemic norms grounding philosophical criticism of science. Their challenge is more clearly brought out by thesis (3), social construction of scientific standards of justification.
As stated, (3) needs clarification. The formulation in terms of justification is inherited from the general analysis of knowledge. However, technical notions of justification from analytic epistemology are of little use in clarifying constructivist ideas. In philosophy of science, justification is associated with logical validation and support of hypotheses by evidence (see Reichenbach Reference Reichenbach1938, 4–10). But (3) is not charitably understood as the claim that logical relations are socially constructed. Social constructivists take scientific practice as basic, its epistemic results (and philosophical rationalizations thereof) as derived. So (3) should be understood from this standpoint. Scientific practice consists of scientists in sociohistorical contexts using diverse materials, methods, and arguments to produce results that qualify as knowledge, or not. In a given context, certain features of scientific practice determine whether or not a given hypothesis qualifies as scientific knowledge.Footnote 8 These features constitute the standard of scientific justification for that context. Thesis (3) (hereafter SCj) is that standards of justification, in this sense, develop by means of social structures, interactions, or values. More specifically, justificatory standards for science are established and maintained via social interactions and negotiations of working scientists with one another, within a broader sociohistorical context. Standards for scientific justification are located within the historical unfolding of scientific inquiry and, more specifically, within the sphere of social agency.
3. Social Construction of Scientific Justification
An example will make this idea clearer. Ludwik Fleck's comparative epistemology, as articulated in his Genesis and Development of a Scientific Fact (Reference Fleck, Bradley, Trenn, J Trenn and Merton[1935] 1979), will serve as an exemplar of an epistemology based on SCj: empirical, historical, and focused on social dimensions of science.Footnote 9 Working through this example will clarify SCj's notion of epistemic justification as an alternative to abstract normative epistemology. I first summarize Fleck's view and then indicate the key features of any epistemology of scientific practice based on SCj.
Fleck's constructivist epistemology is grounded on a detailed historical case study of the concept of syphilis and the development of the Wassermann reaction as a serological test for that disease (see also Fleck Reference Fleck, Cohen and Schnelle[1927] 1986, Reference Fleck, Cohen and Schnelle[1929] 1986). Using this case, he argues that sociocultural factors impinge on the establishment of scientific facts such as “the Wasserman reaction is a test for syphilis.” Specifically, systems of concepts and facts accepted at a particular time and place form mutually supporting systems (“thought styles”) that both constrain and make possible individual cognition within “thought collectives”: social groups whose members share a distinctive thought style grounded in their traditions and ongoing interactions. On this view, the history of science is the history of changes in thought style, and establishment of new scientific facts is the result of human social action in a particular context. Since Fleck presupposes that all branches of science aim to establish definite, (relatively) permanent, objective facts, his constructivist account of their development amounts to an epistemology of scientific practice: a “critique of the methods used to establish [a fact]” (Reference Fleck, Bradley, Trenn, J Trenn and Merton[1935] 1979, xxvii).
Fleck's critique, however, is descriptive rather than normative; he explicitly rejects “normative epistemology” as “obsolete” (Reference Fleck, Bradley, Trenn, J Trenn and Merton[1935] 1979, 51). The rejection is an essential feature of his developmental theory of scientific inquiry. On this view, new scientific facts develop in three stages: “at first there is a signal of resistance in the chaotic initial thinking, then a definite thought constraint, and finally a form to be directly perceived” (95). Facts have active and passive components, corresponding to constraints imposed by a thought style and their necessary (“objective”) consequences, respectively. Within a “branch of knowledge,” the number of facts inevitably increases over time, and differences of opinion decrease as new ideas are thereby constrained (79, 83). Fleck uses the image of an increasingly intricate network of intersecting strands of thought, in which facts serve as nodes for convergence and initiation of diverse strands of research. As controversy is ‘damped’ over time; production of new facts becomes more efficient. In this sense, the growth of knowledge accelerates within a line of inquiry.
In this ongoing historical process, what distinguishes scientific fact from unjustified opinion is not a formal criterion or an abstract epistemic ideal or the brute resistance of nature. Justification is determined, rather, by the prevailing thought style of the scientific thought collective. Grounded in the contingent interactions among members of a thought collective, thought styles change over time and vary widely across fields and disciplines. These changes can be described, and explained in terms of “sociocogitative forces,” but cannot be epistemically evaluated (see esp. Fleck Reference Fleck, Bradley, Trenn, J Trenn and Merton[1935] 1979, Chapter 4, 178–179, note 5). By the same token, the justificatory status of facts cannot be compared across thought styles. Indeed, facts themselves become obsolete with the passing of their thought style. The aim of Fleck's comparative epistemology is to trace the history of changes in thought style and attendant discoveries of new scientific facts. But it does not (and cannot) evaluate these changes with reference to epistemic ideals detached from particular “sociocogitative” contexts. To do so, Fleck avers, would be an “empty play on words or an epistemology of the imagination” (21).
Abstracting from the particulars of Fleck's theory outlines an epistemology of scientific practice based on SCj. On such a view, scientific knowledge is conceived as resulting from interactions among epistemic agents. These agents have social values and interests, varying within and across cultural and historical contexts, which structure their epistemic interactions. Comprising their work, scientific practice is diverse, dynamic, and woven into a complex social fabric. Justificatory standards are of a piece with all this, responsive to social values and influences. Within a line of inquiry, a justificatory standard emerges gradually from the interplay of diverse sociocultural factors and “resistance” to collective human efforts. As social structures, values, and interests change over time, because of many contingent factors, standards demarcating justified from unjustified claims change in correlated ways. These changes in turn affect the epistemic results of science. In this way, justificatory standards for science are constructed by social interactions among epistemic agents, and, in this sense, scientific knowledge is socially constructed.
In Fleck's developmental account of scientific facts, thought styles grounded in interactions among members of a thought collective determine the justificatory standards for that thought collective. This is one way to explicate SCj. The ambiguity of ‘social’ admits many others. Though social constructivists often deliberately mingle metaphysical and epistemological claims, the latter may be distinguished as explications of SCj based on diverse interpretations of ‘social’.Footnote 10 A detailed exegesis of influential constructivist accounts in this vein is beyond the scope of this article. The following sections do not attempt to capture all the insights into, or reflect all the aims of, specific social constructivist accounts of science. Rather, I argue that one core thesis of many such accounts, SCj, is compatible with critical evaluation of scientific practice, that is, with normative epistemology in the sphere of philosophy of science.
These claims require some argument, given Fleck's explicit rejection of normative epistemology, echoed by many recent advocates of social constructivism. But SCj is not a rejection of normative epistemology tout court. What it excludes is any absolute or a priori standard that we can use to evaluate changes in standards of justification. This exclusion does present a challenge to normative epistemology. Given SCj, we can describe these changes in justificatory standards over time or compare them across contexts. But we have no overarching criterion by which to assess these changes as epistemically better or worse. We cannot appeal to further epistemic standards outside the sociocultural context in which a given standard of justification emerges and for a time persists (its epistemic environment, so to speak). Such a standard would transcend the variable sociohistorical contexts in which our scientific practices occur. But, the constructivist claims, we have no access to such transcendent standards. All we can do is engage the context in which we find ourselves.
SCj is compatible with certain features that distinguish justified and unjustified scientific claims being, as a matter of historical fact, invariant—shared across all scientific contexts.Footnote 11 But invariance as such does not ground epistemic evaluation across sociohistorical contexts, since by definition invariant features do not discriminate among contexts. We can describe patterns of similarity and difference in justificatory standards across scientific contexts and over time and explain these patterns in terms of (among other things) social processes and values. But all this is purely descriptive. Comparative understanding of diverse sociohistorical contexts yields no normative conclusions, no principled judgments as to the correct or incorrect way to proceed with inquiries, scientific or otherwise.
SCj does allow for epistemic norms restricted to specific sociohistorical contexts. Within socially constructed contexts of justification, scientific practices can be critically evaluated by the standards for those contexts. But such circumscribed critiques simply reprise the role of scientists themselves and, moreover, are ultimately dependent on acceptance of specific justificatory standards by a scientific community. If a justificatory standard is just what is accepted as such by a scientific community within a particular sociohistorical context, then normative epistemology is subsumed by social history of science. Furthermore, since acceptance of justificatory standards as authoritative is a feature of epistemic communities in general, conceiving them in this way elides epistemic distinctions between science and other human activities: law, religion, government, sport, philosophy, art, fashion. This approach, currently prominent in sociology of science, assimilates science to a broader social domain, effectively “disappearing” epistemic criticism of scientific practice as such (e.g., Jasanoff Reference Jasanoff2005; Frickel and Moore Reference Frickel and Moore2006).
Neither patterns of invariance revealed by comparative social history nor partisan acceptance of the epistemic norms of a particular scientific community within a broader sociohistorical framework can ground normative epistemology of scientific practice. Furthermore, SCj is compatible with epistemologies that exclude normative epistemic criticism. The simplest way to deal with such challenges is to reject SCj. Such an epistemology is committed to the thesis that relations of epistemic justification are not constructed in scientific practice by means of social structures, interactions, or values (¬SCj). The next section argues that, despite its initial appeal, this response will not do. We cannot both reject SCj and effectively critique scientific practices.
4. Rejecting SCj: Alternatives
4.1. Normative Naturalism
Suppose we endorse ¬SCj. Then there are two positive options for normative epistemology: either standards of epistemic justification are part of the historical unfolding of science or they obtain independently of this history. In the former case (‘antisocial constructivism’, if you will), justificatory standards are situated in the historical unfolding of our scientific practices yet are not socially constructed. Laudan's (Reference Laudan1987, Reference Laudan1990) normative naturalism exemplifies this approach.
Laudan is committed to the normative aim of “discovering the most effective strategies for investigating the natural world” (Reference Laudan1987, 27) and to its achievement via an empirical methodology continuous with that of the natural sciences. He conceives epistemic norms as methodological rules in the form of hypothetical imperatives linking means and ends (e.g., “if you want theories with high predictive reliability, reject ad hoc hypotheses”). The consequent of such an imperative specifies a justificatory standard for cases in which the antecedent is satisfied. Epistemic norms are assessed on the basis of empirical correlations (relative frequencies) linking epistemic means and outcomes (Laudan Reference Laudan1987, 24; Reference Laudan1990, 46). Critical, normative epistemology is thus grounded on theories about scientific inquiry and, like theories in science proper, changes in response to new information. Neither theories nor rules have epistemic priority; rather, the two play complementary roles in the historical unfolding of scientific inquiry. So both epistemic justification and its normative assessment (metamethodology) are of a piece with scientific practice.
This account has been criticized as descriptive rather than normative, since assessment of epistemic norms is relative to epistemic ends that vary across scientific contexts and over time. “If two communities … pose different aims of inquiry, is there any sense in which methodology can rationally adjudicate between them rather than merely describe the effectiveness of their respective ‘best strategies'?” (Psillos Reference Psillos1997, 707). This is the same challenge raised by SCj (see Section 3). If normative naturalism is committed to a thoroughly historicized, context-dependent notion of epistemic justification, then the normative naturalist has no reason to reject SCj—at least not with respect to securing the normative dimension of the view. On the empirical side, a large body of historical case studies implicates social structures, interests, and values in the development and maintenance of justificatory standards in scientific inquiry.Footnote 12 To characterize scientific justification as historical while endorsing ¬SCj requires that one ignore or dismiss this body of evidence from sociohistorical studies of science. This would be an uncomfortable position for the historically inclined normative naturalist, to say the least.
Laudan points to three features of his view that secure its normative aspect. First, given an epistemic end, correlations between means and outcomes have normative consequences: one ought to choose means that best promote one's ends. But this is compatible with SCj. If correlations of means and ends are established via social interactions or values, then epistemic norms are socially constructed.Footnote 13 Second, axiology constrains permissible epistemic ends. Laudan's proposed constraints are minimal: realizability and preservation of past scientific exemplars as scientific. Both are compatible with SCj. Indeed, Laudan's defense of their empirical grounding and warrant casts “the overlapping canon of great science … as certifier or de-certifier for new proposals about the aims of science” (Reference Laudan1990, 53; italics in original). Axiological vetting, on his view, occurs at the level of scientific communities. The processes of selection and transmission of scientific exemplars implicate diverse social factors. This interpretation of axiology is congenial to SCj.
The third feature with which Laudan defends the normative aspect of his view is rule R1, an invariant epistemic norm common to all theories of methodology:
(R1): If actions of a particular sort, m, have consistently promoted certain cognitive ends, e, in the past, and rival actions, n, have failed to do so, then assume that future actions following the rule “if your aim is e, you ought to do m” are more likely to promote these ends than actions based on the rule “if your aim is e, you ought to do n.” (Reference Laudan1987, 25)
R1 “can be invoked as a neutral and impartial vehicle” to criticize epistemic norms of diverse communities (25). Use of this normative standard hinges on “evidence that the means proposed in the rule promotes its associated cognitive end better than its extant rivals” (26). To build up a methodological system based on principles of evidential support, we need only look for such evidence, in the form of statistical correlations among epistemic means and outcomes.
As noted above (Section 3), invariance as such does not ground normative criticism. And if R1 were socially constructed, so would Laudan's proposed methodological system be, raising all the concerns discussed above. But this is not the way Laudan conceives R1. On his view, it is a principle of induction that transcends specific justificatory contexts. Its epistemic status is either a priori or vindicated in the historical long run—in either case, outside the specific justificatory contexts of scientific practice. This construal of R1 is incompatible with SCj and underwrites endorsement of ¬SCj by normative naturalists. But, by the same token, it makes Laudan's view an appeal to justificatory standards independent of the historical unfolding of scientific inquiry—the second positive option. The first appears unmotivated; I shall not discuss it further.
Laudan's R1 exemplifies an ‘exogenous’ philosophical standard by which the justificatory standards of practicing scientists can be critically evaluated. Let us suppose, contrary to fact, that there is no philosophical dispute concerning R1, that it is impeccable in the abstract. Applying R1 to scientific episodes still presents a problem, in this philosophical utopia. Using R1 as an epistemic standard requires an account of “which methods have promoted, or failed to promote, which sorts of cognitive ends in the past” (Laudan Reference Laudan1987, 27). And, as Mayo and Miller have recently noted, these productive connections cannot simply be read off of statistical correlations between justificatory standards satisfied or violated in a particular case and the (long-term) epistemic outcome of that case (Reference Mayo and Miller2008, 308–309). To critically deploy R1, we need an account of how science works, which associates specific standards with specific outcomes, based on a representative sample of historical cases. Not only do we currently lack such a sample, but obtaining the evidence needed to critically deploy R1 presupposes the very account of the epistemic workings of science that critical deployment of R1 is intended to produce. That is, if committed to ¬SCj, normative naturalism suffers from a debilitating circularity (cf. Worrall Reference Worrall1999, 339).
The problem of application is not unique to Laudan's view. It arises for any normative epistemology of science that imposes an exogenous epistemic norm onto our scientific practices in order to ground criticism of the latter. For any epistemic norm that is not continuous with the practices to be evaluated on its basis, the question arises: whence its prescriptive grip? How can we use it to critique our scientific practices? A normative epistemology that does not answer this question fails to engage the very practices it aims to clarify. Purely abstract standards for epistemic justification, lacking any connection to our practices, are at best pretty fictions, descriptions of imaginary science to dazzle the uninitiated and inspire novices. At worst, they mask relations of power and politics that determine what is accepted as scientific knowledge, protecting inequities from critique and lending social biases an appearance of inevitability.Footnote 14 To prevent such misuse of epistemic ideals, their connection with scientific practice must be articulated. The next sections argue that, as Laudan's example suggests, normative accounts committed to ¬SCj cannot do so. Such ideals are useless for criticizing our scientific practices (though, of course, they may serve other philosophical purposes).
4.2. Social Empiricism
The connection between idealized or abstract epistemic norms and scientific practices is typically made via case studies: intuitively plausible examples of epistemic success or failure in historical or contemporary scientific inquiry that conform to an idealized justificatory standard. Miriam Solomon's (Reference Solomon2001) account of social empiricism exemplifies this strategy. I argue, first, that social empiricism does not successfully link epistemic ideals with our scientific practices. I then show that (as with normative naturalism) the problem is not specific to Solomon's account but arises for any epistemic ideal imposed on our scientific practices from ‘without’.
Solomon explicates the epistemic standard for science in terms of distributions of decision vectors (DV), an inclusive term designating the various factors that influence scientific decisions regarding the theories on offer within a scientific community at a given time. Fundamental to this view is a distinction between two types of DV: empirical, which cause preference for theories with some genuine empirical success, and nonempirical, which cause preference for theories at best irrespective of their genuine empirical success. Scientific communities satisfy the epistemic norm of social empiricism to the degree that empirical DV are distributed among theories in proportion with the empirical success of those theories (“equitably distributed”), and nonempirical DV are evenly distributed among theories (“equally distributed”).Footnote 15 So social empiricism explicates the epistemic standard for science as a relation of two distributions: of genuine empirical success among theories and of DV among theories.
This epistemic standard is applied to various sociohistorical case studies: plate tectonics, the ‘central dogma’ of molecular biology, the variability hypothesis in sex research, surgical technique, the ovulation theory of menstruation, the Copenhagen interpretation of quantum mechanics, cold fusion, and treatment of peptic ulcers. In these examples, Solomon defers to historians of science to individuate DV, delineate community contexts, and describe scientists’ choices among alternative theories. So the distribution of DV among theories is a purely descriptive matter. But the distribution of genuine empirical success among theories is not. Here Solomon appeals to the further distinction between genuine empirical success and what is accepted as such in a given context. Genuine (robust and significant) empirical success of a theory consists in the success of observations, predictions, and explanations derived from that theory that is due neither to chance nor to selective treatment of evidence. Solomon positively characterizes genuine empirical success via “Whig realism”: a theory's genuine empirical success is the portion of its empirical success that is retrospectively explained by its “true substructures.” Only hindsight can distinguish these true explanatory substructures of a theory from false components supported by chance or biased treatment of evidence.
According to Whig realism, the distinction between empirical and nonempirical DV is a matter of causal fact: many factors influence theory choice, and some track genuine empirical success while others do not. So the epistemic standard for science is not socially constructed but determined by the distribution of scientists’ decisions about theories and the distribution of genuine empirical success among those theories. Suppose (as with Laudan's R1) that Whig realism is philosophically beyond reproach. How can its epistemic standard be applied to episodes of scientific practice? As noted above, Solomon appeals to recent history of science to individuate DV, but their classification as empirical or nonempirical is based on her own intuitions. The intuitions at work are widely shared in history and philosophy of science today,Footnote 16 and in practice, this seems the best one can do. However, the resultant critiques of scientific communities are perforce relative to currently prevalent intuitions about the epistemic value of various factors influencing theory choice. Such intuitions change over time and across scientific contexts, as evidenced by the same historical record to which Solomon appeals to identify distributions of DV. Why should the intuitions of historians and philosophers of science today determine the epistemic standards for communities of past and present scientists?
Whig realism, rather than providing the answer, undercuts social empiricism when the latter is put into practice. In practice, classification of DV as empirical or nonempirical depends, unavoidably, on our current understanding of theories’ genuine empirical success. Whig realism applies to this understanding of the empirical/nonempirical distinction, just as for any other theory; it is, ex hypothesi, philosophically impeccable. Given Whig realism, today's partition of theories into true substructures, false excrescences, and idle wheels itself mingles true and false components, distinguishable only with hindsight. Our critical evaluations of scientific episodes do not depend on theories’ genuine empirical success but on our idea of it. Genuine empirical success is an abstract ideal, realized (if at all) in the indefinitely far-off end of scientific practice. The epistemic standard for social empiricism, given Whig realism, is one we do not have in hand. This vitiates epistemic criticism of scientific practices based on that standard.Footnote 17 There is no reason to privilege the epistemic standards we do have in hand—our current understanding of the ideal—over epistemic standards held in different contexts.
One might defend the normative force of social empiricism's critiques within our current community. Insofar as they conform to currently prevalent epistemic standards for science, Solomon's assessments should have critical bite where those standards hold. But this lets social constructivism in through the backdoor, so to speak. The prescriptive bite of social empiricism stems from the authoritative acceptance of epistemic standards in our specific context, which determine the distribution of genuine empirical success among theories. But, given Whig realism, we cannot rule out the possibility that the epistemic standards we accept as authoritative are socially constructed: established and maintained via social interactions and values. So social empiricism cannot decisively reject SCj and effectively critique our scientific practices. Solomon's account either lacks grounds for epistemic criticism or admits SCj.
A version of this dilemma arises for any epistemic standard conceived as independent of the historical unfolding of scientific practice, whether the specific ideal invoked is truth tracking, reliability, rationality, or genuine empirical success. Whatever their virtues in the abstract, such idealized standards require further assumptions to connect them with our scientific practices. In social empiricism, these assumptions guide the classification of DV as empirical or nonempirical. More generally, assumptions are needed to specify which aspects of scientific episodes are epistemically relevant and whether this relevance is positive or negative. Different specifying assumptions make a difference to the results of epistemic criticism. If the choice among assumptions is arbitrary—if there is no reason to privilege one set of specifying assumptions over another—then the resulting epistemic assessments lack critical bite. A priori reasons cannot settle the issue because for any such reason the original question recurs: what is the connection with our scientific practices? A regress of ideals does not demonstrate their application but only makes for a higher castle in the air.
One reason to privilege a set of specifying assumptions is that epistemic ideals specified in this way are accepted as authoritative within one's epistemic community.Footnote 18 This gives evaluations of scientific practice based on them a critical bite, albeit circumscribed by community boundaries. But then SCj is not excluded; the possibility that epistemic standards accepted as authoritative are socially constructed cannot be ruled out. To rule out SCj in the case of standards one accepts as authoritative, one would have to show that their epistemic status stems from outside the social interactions and values of that context. This might be possible. But Whig realism does not accomplish it. A more promising approach is a normative account that directly engages the social aspects of our scientific practices, leaving less scope for SCj by normatively constraining the very features with which it is concerned. This is the approach taken in Helen Longino's Fate of Knowledge (Reference Longino2002), the final exemplar of normative epistemology I shall consider.
4.3. Critical Contextual Empiricism
Longino's critical constructive empiricism aims to be “responsive to the normative uses of the term ‘knowledge’ and to the social conditions in which scientific knowledge is produced” (Reference Longino2002, 1; italics in original). Her strategy is to reject the “rational-social dichotomy”—the thesis that cognitive rationality and the influence of social factors on cognition are mutually exclusive. If this dichotomy is assumed, the logical lacuna between theories and data allows subjective idiosyncrasies of personal or social origin to influence the content of empirical beliefs through background assumptions concerning evidential relations, thereby frustrating objective or rational empirical knowledge. Rejecting the rational-social dichotomy allows that cognitive rationality and sociality are interdependent rather than opposed. Distinguishing three senses of knowledge (knowledge as content, knowledge-producing processes, and the relation of knowing), Longino rejects dualisms associated with each sense, thereby articulating a “nondichotomous” account of reliable empirical knowledge.Footnote 19
Longino's disambiguating framework recasts philosophers’ normative concerns so as to reconcile their projects with (some) sociological data. But the framework is thoroughly abstract. For application to cases, Longino proposes epistemic norms for critical interaction within a community (CCE norms): public forums, ‘uptake’ or response to criticism, public standards for debate, and tempered equality of intellectual authority within the community (Longino Reference Longino2002, 128–135; see also Chapter 4 of Longino Reference Longino1990). Based on detailed case studies of human behavioral research, the CCE norms are designed to bridge both the rational-social gap and the divide between epistemic ideals and scientific practices. However, if SCj is rejected, Longino's CCE norms cannot span the latter.
CCE norms specify social conditions required for ‘interactive’ criticism within epistemic communities. These requirements complement rather than replace epistemic norms for the individuals within those communities. On Longino's view, the means to reliable empirical knowledge is pluralism. Idiosyncratic biases and preferences embedded in the background assumptions of individuals or subgroups within a community are eliminated by criticism of hypotheses and methods from diverse perspectives. So inclusion of diverse perspectives and transformative criticism of methods and results are required for the epistemic ideal of reliable empirical knowledge. CCE norms specify the features of scientific communities that satisfy these requirements for reliable empirical knowledge. They thus supply an epistemic standard for scientific communities.
One might object that this standard is too permissive for effective epistemic criticism. Consider a community of scientific creationists who have public forums for discussion and shared standards and take account of criticism from within the group in accordance with tempered equality of intellectual authority.Footnote 20 It would seem that Longino should say that this group could reliably know that the first metabolic pathways were intelligently designed. But this is too fast. On Longino's view, it is not enough to take account of criticism only within the group. Epistemic communities satisfying CCE norms must be responsive to criticism from outside the community as well as within. However, community boundaries must play some role in structuring epistemic criticism. If satisfaction of CCE norms demanded openness to criticism tout court, then no community could satisfy them.Footnote 21 Reliable empirical knowledge would be an unattainable ideal, divorced from our practices. So application of CCE norms to scientific episodes presupposes some specification of community boundaries, which ‘tempers’ the requisite openness to criticism. In this sense, normative assessments of scientific communities based on Longino's CCE norms depend on the boundaries of those communities.
Longino's solution to this problem hinges on tempered equality of intellectual authority. For a community to satisfy CCE norms, the full variety of available perspectives must be represented in proportion to their ‘track records’ vis-à-vis those norms. This is how equality of intellectual authority is tempered, to chart a course between extremes of permissive relativism and impossible absolutism. So equality of intellectual authority is not simply bounded by group membership but tempered by engagement with outside perspectives as well. This solution involves the history of critical interactions among epistemic communities. CCE norms apply to epistemic communities in the context of a history of interactions with ‘neighboring’ communities whose projects have brought them into critical contact with one another. Sociohistorical studies of science have traced the interactions establishing boundaries between epistemic communities in a number of scientific episodes.Footnote 22 Appealing to these studies to ground application of the CCE norms commits one to SCj.
If SCj is rejected, then these boundaries are not socially constructed. But then it is hard to see how CCE norms can be critically applied, since their application requires some specification of community boundaries. Longino's own account leaves this as a lacuna, to be filled in by empirical studies of science—for example, by sociological data on the intensity of interactions among inquirers (personal communication from Longino). But these patterns of interaction are enmeshed in sociocultural contexts, and numerous case studies implicate social interactions and values in the settlement of the boundaries of scientific communities. In this sense, evaluation based on the CCE norms depends on socially constructed community boundaries. Insofar as these norms apply to scientific practice, SCj cannot be excluded. The issue is not simply that critical constructive empiricism requires empirical input for its norms to apply to scientific episodes, as in parameter specification. The empirical input required cannot be specified independently of social history of science. The boundaries of scientific communities result from social negotiations and are continuously under revision. If the community boundaries that determine whether or not knowledge is reliable are the result of socioculturally enmeshed controversies and negotiations, then CCE norms are, in practice, committed to SCj.
4.4. SCj and Normative Epistemology
These exemplars indicate that we cannot both reject SCj and effectively critique scientific practices. If SCj is rejected, then a gap opens between our actual practices and epistemic justification for scientific claims, which cannot be bridged without vitiating the normative bite of exogenous standards. Insofar as social epistemic norms apply to episodes of scientific practice, SCj cannot be decisively rejected. This suggests that philosophers who share the goal of critical epistemology of science should thus endorse SCj. But then its challenge to normative epistemology reappears in full force (Section 3).
If SCj and normative epistemology are mutually exclusive, then endorsing SCj would be unacceptable. Assumptions of normative/descriptive dualism, a sharp divide between sociohistorical facts, and epistemic ideals fan epistemological controversy over social constructivism. But it is just this dichotomy that impedes an effective critical epistemology of scientific practice. Neither pure description of sociohistorical facts nor abstract ideals with no clear application to our practices can suffice. The problem is not that normative and descriptive perspectives on epistemic justification are distinct but that they are seen as conflicting. I next propose an account based on SCj, which avoids this dualism, instead casting descriptive social history and normative epistemology as collaborative partners in understanding science.
5. Scientific Practice as Social Action
My strategy is to integrate normative and descriptive approaches to scientific justification, effecting a collaboration between them. My starting point is analysis of the concept of collaboration itself, more specifically, collaborative effort among diverse participants. Scientific inquiry and the study of science are collaborative activities in this sense. The analysis of such activities by philosophers of social action provides a rigorous general perspective that is neutral with respect to debates over social constructivism. These analyses focus on the distinctive attitude of joint intention (also referred to as collective or shared intention) and vary considerably on points of social metaphysics. For example, on Bratman's (Reference Bratman1999) reductive account, shared intention is identified with interlocking individual intentions, whereas Gilbert (Reference Gilbert1989) analyzes the concept in terms of an irreducibly plural subject.Footnote 23 However, a thin consensus account of collaborative activity emerges from these diverse analyses. Collaborative activity involves (1) a shared goal, (2) participant means taken to it, (3) coordination of diverse means among participants, and (4) a public context for the preceding. Importantly, a shared goal is not merely had by multiple agents as tokens of a type. It is, rather, an end that participants in an activity are trying to achieve together. More specifically, what distinguishes collaborative activities is that participants view their actions as contributing to a goal shared with other participants, with whose actions theirs are coordinated.
The participatory relation, constructed in episodes of joint action, mediates between groups and their members. As an instrumental relation, it also connects epistemic standards and scientific practices: the former specify shared epistemic goals of the latter. Since groups may themselves be members of groups, this consensus view of collaborative action provides an integrative framework for studying multiple levels of social organization in science. This framework may be summarized in three premises:
(1) Scientific practices can be understood and evaluated in terms of ‘fit’ between shared goals and means taken to them by participants (minimal instrumentalism).
(2) Scientific practices have an epistemic goal, characteristics of which are specified by empirical study of science rather than philosophical presumptions.
(3) The epistemic goal of scientific practices is shared—seen by participants as something to be jointly achieved.
From these three premises, two requirements follow: the shared goal of a collaborative action is achievable by the group that has it, and means taken to the shared goal are coordinated among group members.Footnote 24 These requirements constrain the participatory relation in an instrumental framework presupposed in our familiar habits of explaining and evaluating human action. In any particular case, the participatory relation will be specified so as to conform to the requirements, which operate as preconditions for understanding social actions in terms of fit between goals and means.
These premises and the entailed requirements apply to collaborative action in general. Fleshing out this minimal framework requires empirical study of scientific practices. I have elsewhere discussed methods and the results of one such study (Fagan Reference Fagan2007, Reference Fagan2010). Here I briefly sketch the main points. The main methodological concern is to avoid imposing exogenous epistemic ideals onto scientific practices. Instead, a participant-driven narrative is constructed from published sources, personal interviews, and laboratory visits. The focus of the narrative is scientific success and the various social interactions that contribute to same. Diverse individual perspectives as well as multiple levels of social organization (small research teams, laboratory groups, interlab partnerships, and scientific fields) yield many different conceptions of scientific success and various important interactions. The role of the sociohistorical analyst is to accommodate diverse participant perspectives within a coherent, multilevel narrative, specifying participatory relations that contribute to scientific success for research teams, laboratory groups, fields, and subfields.
In such a narrative, group boundaries are flexible, shifting over time, and often porous. Yet, once a coherent, inclusive account is set out, robust patterns can be discerned. The patterns of interest here are robust to diverse levels of social organization and to incidents involving different participants within the narrative. These patterns are further corroborated by other sociohistorical case studies. One robust pattern that emerges is a characterization of scientific success, which may be conceived as a general mechanism for experimental success as recognized by participants (individuals or groups): construction of improved models of the subject matter under investigation coordinated with formation of new group boundaries. Over time, models constructed in one group context are given critical scrutiny from a new (though not wholly unrelated) perspective, as new group boundaries form. In many cases, this mechanism implicates familiar epistemic values: improved models are taken to increase the scope, consistency, precision, or accuracy of our knowledge; coordination of diverse lines of inquiry recalls epistemic virtues of consistency, coherence, and unification. These ties to epistemic ideals are abstract. But when this mechanism for scientific success is considered in its context of operation—the social history of science—a more concrete, robust feature is brought out. The pattern of group boundaries and improved models that unfolds historically, the path of realized scientific success, depends on specific, unanticipated experimental results.
From these robust results and the requirements of the social action framework (see above), epistemic standards for effective criticism can be derived. One is of particular significance here. Given the means taken in practice (the mechanism for scientific success) and the requirements of achievability and coordination, the shared epistemic goal of scientific practices is knowledge such as to satisfy epistemic standards not specifiable in advance. This rules out identification of standards for epistemic criticism of science with the justificatory standards accepted as authoritative in specific contexts. Rejection of SCj to secure normative criticism is motivated by this identification: if authoritative acceptance in a particular sociohistorical context is not sufficient for normative criticism and SCj allows for no other conception of justificatory standards than such authority, then critical epistemology of science must appeal to a standard that is not dependent on social aspects of our scientific practices (¬SCj). But the conclusion does not follow, since SCj in the social action framework does allow for another conception of justificatory standards, which is incompatible with identifying justificatory standards for science with authoritative acceptance of standards in a particular sociohistorical context.
6. Conclusion
I have argued that social constructivism, contrary to its reputation as an opponent of philosophy of science, can in fact contribute to an effective critical epistemology of our scientific practices. By this I do not mean that sociohistorical case studies by constructivist scholars provide the empirical substrate for philosophical criticism but that a social constructivist thesis (SCj) can be a starting point for critical epistemic evaluation of science. Moreover, there is reason to think that rejecting this thesis (¬SCj) is incompatible with an effective critical epistemology of our scientific practices. If we aim to critically evaluate, and perhaps improve, our scientific practices, rather than articulate abstract ideals with no clear application, then we should be social constructivists.
This is not to say, of course, that we should endorse social constructivist theses tout court. The first part of my argument used the tripartite analysis of knowledge to distinguish three varieties of social constructivism about scientific knowledge (Section 2). Though all are worth taking seriously, I defend only one: social constructivism about epistemic justification (SCj). Because this construal of epistemic justification is quite unlike the abstract formulations characteristic of the analytic tradition, SCj requires some clarification before its implications for epistemology can be assessed (Section 3). I used Fleck's comparative epistemology as an exemplar to clarify SCj. Its thesis is that epistemic standards for science are established and maintained by value-laden social interactions among practicing scientists, such that epistemic justification is embedded in the social organization of scientific communities.
So understood, SCj presents a significant challenge to abstract and idealized accounts of epistemic norms for science. In Section 4, I argued that this challenge is not met by rejecting SCj. If standards of epistemic justification are part of the historical unfolding of science, then ¬SCj is unmotivated and open to methodological criticism. If justificatory standards are independent of our scientific practices, then ¬SCj follows, and the problem of linking these exogenous standards to our scientific practices arises. Without such a link, epistemic ideals cannot be effectively deployed for critical evaluation of our scientific practices. Using recent accounts of normative naturalism (Laudan Reference Laudan1987, Reference Laudan1990), social epistemic norms (Longino Reference Longino2002), and social empiricism (Solomon Reference Solomon2001) as exemplars, I argued that we cannot both reject SCj and effectively critique scientific practices. If this negative argument succeeds, then philosophers who aim at critical epistemology of science should endorse SCj.
This conclusion may seem unacceptable, since SCj is also compatible with elision of any distinction between philosophy and social history or between science and society. If SCj is conjoined with a sharp divide between sociohistorical facts and epistemic ideals or a simplistic view of their relation, then the consequences are indeed unwelcome. But these restrictive assumptions are unnecessary. I propose an analytic framework that avoids them, based on general analyses of social action. Within this framework, SCj can be further explicated to yield epistemic ideals implicit in our scientific practices, providing grounds for their critical evaluation. The social action framework is neutral with respect to debates over social construction of scientific knowledge, is minimally assumptive, and requires fleshing out by sociohistorical studies of science. Such studies, focusing on the participation relation linking groups and their members, as well as means and ends, yield robust patterns associated with scientific success as recognized by participants. These robust results, combined with requirements of the social action framework, yield epistemic norms that are grounded in social aspects of scientific practice, without undermining their critical force.
A final note, by way of conclusion. Because the concepts and themes discussed in this article deal with core aims and long-standing debates in philosophy of science, many of its points have been raised before, albeit in different argumentative contexts. In particular, my approach is consonant with work in history and philosophy of science and in philosophy of biology by Richard Burian, Hasok Chang, Jordi Cat, Michael Friedman, Jim Griesemer, Don Howard, Jutta Schickore, and William Wimsatt.Footnote 25 Their and others’ accounts overlap and reinforce many of the ideas defended here, though to trace these connections is beyond the scope of this article. However, the argument that philosophers of science should endorse a form of social constructivism, SCj, is original, as is the positive approach that secures critical epistemology from a social constructivist stance. My aim, like that of experimental scientists, has not been to create a theory de novo but to combine existing materials in a new and useful way. Social constructivism should be part of the tool kit for philosophy of science.
