Pragmatism,pluralism,and eclecticism: Sil and Katzenstein’s “Analytic eclecticism” in Beyond Paradigms

Sil and Katzenstein’s concept of the “paradigm”

In order to assess the argument that Sil and Katzenstein develop in Beyond Paradigms, it is worth trying to clarify precisely what they mean by the concept of a “paradigm.” Because of the enormous influence that Kuhn’s paradigm-based analysis of the history of science has had on the study of international politics, and the social sciences generally, many readers will presume that the authors use the term in Kuhn’s sense. Indeed, much of the text of Beyond Paradigms appears to confirm this. The frequent occurrences of the locution “paradigm-bound research” further the impression that they are referring to Kuhnian paradigms, which are indeed very “bound.” But Sil and Katzenstein are not using “paradigm” in the Kuhnian sense.

In the introduction to Beyond Paradigms the authors describe what they mean by the term “paradigm.” ⁶ They say that they do not follow Kuhn’s notion of paradigm, but mean the term to convey Larry Laudan’s notion of a “research tradition.” An author is always free to use terms any way desired, as long as the author provides a definition and adheres consistently to that definition. (There can, though, be a cost in terms of confusion, if an author uses a familiar term in an unfamiliar way.) In IR debates the term “paradigm” is understood in different ways by different people. Some read it in a strict Kuhnian way, while others read it in a very loose way, as simply a set of related theories. To define a paradigm as a Laudan-style research tradition is unusual.

For Laudan, “A [natural science] research tradition, at best, specifies a general ontology for nature, and a general method for solving natural problems within a given natural domain.” Laudan contrasts this definition with that of “theory,” which “articulates a very specific ontology and a number of specific and testable laws about nature.” ⁷ Sil and Katzenstein say that a research tradition has “long-enduring epistemological commitments that govern the scope and content of scientific research in any given field.” ⁸ These produce traditions, “each of which consists of: ‘(1) a set of beliefs about what sorts of entities and processes make up the domain of inquiry; and (2) a set of epistemic and methodological norms about how the domain is to be investigated, how theories are to be tested, how data are to be collected, and the like.’” ⁹

One way to clarify a general term is to examine its instantiations. Sil and Katzenstein rely on three examples of what they regard as IR paradigms—realism, liberalism, and constructivism, though they note that there are others, as well. Regarding constructivism, Sil and Katzenstein offer a helpful delimitation of the meaning of the very broad notion of the term at the end of chapter 2. ¹⁰ It is helpful because “constructivism” has been used to refer to a very wide range of types of theories. A still deeper assessment of possible differences would be of value. In any case, one might question the parallel character of the three examples. But for present purposes this paper takes at face value that realism, liberalism, and constructivism constitute three examples of the same sort of “thing,” specifically of a Laudan-style research tradition.

Who are the opponents of analytic eclecticism?

Sil and Katzenstein’s choice of the flexible, non-Kuhnian concept of paradigm raises a question about the logic of the argument: specifically, at what group of opponents is their argument aimed? Any successful argument begins by having a target audience. What audience are the authors trying to persuade? That is, what sorts of scholars who pick up Beyond Paradigms need persuading? Supporters of Kuhn in IR believe that paradigms have sharp borders and relatively clear identity conditions and that the theoretical and empirical content of one cannot be combined with the content of others. But this group is not likely to be persuaded by the argument for analytic eclecticism, in part because Sil and Katzenstein focus on Laudanian research traditions when they use the word “paradigm,” not the sort of incommensurable “thing” that they believe prevents cross-paradigm fraternization. Thus, they do not address the Kuhnian view of empirical science, according to which researchers work within incommensurable frameworks. Hence, the Kuhnian position can thereby be maintained by its supporters despite the argument Sil and Katzenstein develop. Another route to defeat Kuhnians would be to take on Kuhn’s philosophical view of the character of empirical science. Indeed, there are quite compelling arguments against IR Kuhnianism that are much more direct than that offered by analytic eclecticism, as I have tried to show elsewhere. ¹¹ But this route is not available, as Sil and Katzenstein disparage the role of philosophy of science as a framework or basis for substantive IR conclusions.

Sil and Katzenstein’s concept of a paradigm is, nevertheless, a useful one in terms of conveying what at least some IR scholars think about the sort of cluster of theories that realism and liberalism comprise. But by defining a paradigm in this way, their argument’s force is limited. Sil and Katzenstein emphasize that they do not oppose paradigm-bound research in general, only the notion that theories must, in all cases, be paradigm bound. Since they are discussing “research tradition paradigms” that are, by definition, not mutually exclusive, and since the authors are arguing only that variables from different research tradition paradigms will produce better theories than purely within-paradigm competitors, they can prove their point simply by identifying a few multi-paradigm theories that yield better explanations or prediction-supporting theories than their purebred paradigm competitors.

If these theoretical clusters lack sharp borders by definition, then one is not likely to be surprised by Sil and Katzenstein’s conclusion that the most effective solutions to some intellectual and practical problems will occasionally benefit from mixing variables from different (non-mutually exclusive) clusters, i.e., Laudanian research traditions. Put simply, the looser and less constricting that Sil and Katzenstein define the notion of a research tradition/paradigm, the less force there is to the conclusion that scholars can go beyond them.

Do IR authors regard themselves as unable to transcend paradigm boundaries?

Sil and Katzenstein observe that, according to the Teaching and Research in International Politics (TRIP) surveys of 2006 and 2009 (the most recent prior to Beyond Paradigms), a large portion of IR scholars self-identify as operating within a specific paradigm. ¹² This prevents scholars from producing more useful theories—though the authors add that non-paradigm-specific work is on the upswing. However, the TRIP surveys do not offer a definition of paradigm, and a fortiori do not discuss Laudanian research traditions. Thus, they do not specify that paradigm be understood in any way that prohibits cross-tradition combinations of variables or explanatory factors. As noted, some IR scholars use the term “realist paradigm” essentially as a synonym for “realist theory”—or perhaps “realist group of theories”—rather than in a way that fits any precise, technical definition of paradigm. That is, IR authors use the term without any implication that realists, liberals, etc., each has either its own un-translatable vocabulary for observation terms or their own distinct, well-defined set of criteria of theory choice.

TRIP respondents may well interpret paradigm more broadly, and even more flexibly, than as a Laudan-style research tradition (or Kuhnian paradigm). That is to say, many respondents very likely conceive of realism simply as a clearly delineated cluster of theories, but not in a Laudanian sense that requires them to have distinct epistemological principles and standards of evidence. If IR scholars adhere to a set of variables within a single theoretical tradition, it may well be owing to their graduate education—and scholarly experiences beyond—as they have been impressed by the empirical research of that theoretical tradition rather than a metatheoretical framework, like the epistemological principles of a theoretical tradition. So, for example, political realists may avoid institutionalist variables simply because their empirical research, drawing on variables like the distribution of power-capabilities, anarchy, balancing, etc., produces better explanations. It may have nothing to do with any commitment they have, as supporters of that realist tradition, to recognize a set of paradigmatic—or research tradition—boundaries (involving methodological prescriptions and ontological commitments) of the sort Sil and Katzenstein associate with Laudanian research traditions.

Further evidence that many IR scholars understand the term “paradigm” in a sense much broader than Sil and Katzenstein’s Laudanian sense (or Kuhn’s) can be seen by noting the following. If many IR researchers adhered to a set of variables from a single research tradition for metatheoretical reasons, then we should expect to see many authors rejecting criticisms of their theories from those working in other research traditions by drawing on metatheoretical reasons. That is, they would point out that the critics from the other paradigms fail to adhere to metatheoretical principles that Sil and Katzenstein note in the introduction to Beyond Paradigms. This is at least true of debates between authors in two of the three paradigms, realism and liberalism (which raises the question of whether the three examples of paradigms Sil and Katzenstein offer are really species of the same kinds of intellectual entity). Or if they were indeed adhered to their paradigm for metatheoretical reasons, scholars would point out that opponents’ criticisms are misguided because they apply criteria and the like from outside of their own paradigm. But these defences are hardly ever seen in IR debates. Rather, in reading IR debates one can see that many authors advance their own theories in part by taking seriously rival theories and arguing against them. ¹³ That is to say, many authors defend their own views with empirical evidence that supports those theories and by engaging with other research traditions or paradigms by pointing out flaws in the critics’ logic or by developing new empirical tests, etc., all of which are reasonable responses only if the authors and the critics regard the researchers from the opposing traditions as engaging in fully meaningfully dialogue, not hindered by underlying epistemological disagreements.

IR research requires premises about what inquiry is and is not legitimate—we cannot avoid metatheory in the study of IR

Sil and Katzenstein criticize IR scholarship for focusing too much on metatheory generally, and epistemology, in particular, at the expense of practically oriented, eclectic work. They maintain that there are ways to evade these philosophical questions. But like others who have taken this tack, such as Jackson and Monteiro and Ruby, the arguments are flawed. ¹⁴ All academic work that defends a particular answer to a research question––whether in physics, mathematics, economics, or IR—must draw on metatheory in various ways. For example, all such work builds on many unstated premises about the nature and limits of legitimate knowledge claims (epistemology), the possibility of establishing causal or constitutive relations (metaphysics), the status of the things our theories describe (ontology), and, as just noted, the proper criteria of theory choice. ¹⁵ Philosophers of science endorse different sets of criteria and, among the criteria they share, they prioritize them very differently. However, any debate in IR over the most appropriate criteria for a researcher’s purposes is in reality a debate in metatheory. Indeed, Sil and Katzenstein are engaging in metatheory by criticizing the way IR research is conducted today and, in particular in the way are doing so, by proposing their own criterion, namely, the criterion of producing useful knowledge.

While downgrading the role of metatheory generally, Sil and Katzenstein say that, to the extent it matters, ontology is more central than epistemology. ¹⁶ This is misplaced; epistemology is no less central than ontology, and, in fact, ontology seems to be less central, at least in the natural sciences. While philosophers of science disagree on many points, virtually all agree that what makes some disciplines scientific and others not scientific is found in principles of knowledge production––an epistemological consideration. In contrast, there is no consensus on the importance of ontology. Moreover, if considerations of ontology are important, then we should be able to identify which ontological criteria are key to aiding us in choosing the best answer to questions posed in IR.

The claim of ontology’s importance unravels as soon as we seek those criteria, as the following highlights. IR debates typically include as evidence a set of observations. Researchers eliminate competing theories that fail to satisfy the major criteria. We return, then, to the question of the criteria that should be used. Which among them that help us choose the best theory have anything to do with ontology? The answers are hard to find. One criterion would be that theories may not posit forces that are somehow mutually exclusive. But beyond that, it is hard to imagine what it would take for a scholar to say, “Theory T₁ performs better on all other criteria, but we choose theory T₂ because T₁ fails to satisfy the ontology criterion.” Other than pointing out internal inconsistency, can we imagine scholars accepting the refutation of a theory on the grounds that there is something “wrong” with a theory’s ontology?

The ontology accepted by a community of scholars will include both observable and unobservable entities. But the latter (quarks, energy quanta, trade regimes, leaders’ emotional states, etc.) are inferred from what is posited by our best theories, which are, in turn, supported entirely by our observations and chosen criteria. We have no other access to the existence of such things. All we know about the unobservable world is what our theories tell us, and the theories are justified by the experience we have of observing that which is observable. Any ontological criteria we have, aside from avoiding contradictory posits, would seem to be suspiciously and illegitimately a priori. Thus, as I have tried to show elsewhere, of those in IR who discuss metatheory, many seriously overstate the role of ontology. ¹⁷

Eclectic theorizing and pragmatism

Sil and Katzenstein endorse pragmatic IR research that produces “successful results.” They use pragmatism in the common parlance sense of the term as merely “an ethos” that requires some relationship to practice. ¹⁸ But they also use it in the philosophical sense of the doctrine of American pragmatism. ¹⁹ In my view, Sil and Katzenstein are compelled to deal with more developed versions of pragmatism, for two reasons. One is that it yields a philosophically coherent basis for the connection between scientific research and human practice. And the other is that Sil and Katzenstein advocate the criterion of “successful practice” but leave this notion undeveloped. Any attempt to clarify what “successful” means in this context, and how it is identified in real cases, requires a much more precise and rigorous operationalization of the term—a project that that American pragmatism can straightforwardly underpin.

Sil and Katzenstein disapprove of the search for abstract empirical principles and covering laws in IR. But philosophical accounts of pragmatism provide clear groundwork for covering laws and at the same time conceives of laws as general principles that can be both (a) accepted as known and (b) fallible, that is, as capable of revision or replacement as new evidence accrues. Charles Sanders Peirce, the founder of philosophical pragmatism, was a practising scientist and accepted general laws of science. But he understood the epistemic status of those laws in a revolutionary way, namely, that scientists develop genuine knowledge but that legitimate scientific laws are revisable and fallible.

Peirce’s notion of scientific knowledge rejected foundationalism, but did so in a way that escaped the skepticism that was believed to follow from any anti-foundationalist theory of knowledge. Peirce rejected the metaphor of a body of knowledge as a building that has to be constructed upwards on a rock-solid foundation—a view that gives a body of statements the strength to hold up a theory with covering laws. With such a view, every bit of support must have as much or more certainty as anything that stands on top of it. In contrast, Peirce argued in his paper “Some consequences of four incapacities” that science functioned more along the lines of a bridge that may be stable and able to withstand great stresses because it is held up by a powerful cable, even though the cable is composed entirely of thin strands of wire. In developing scientific knowledge, some evidentiary observations may be thrown out at various times. But the body of knowledge does not collapse, just as a strand of wire in a cable may break, but the bridge, held up by many cables composed of many strands of wire, securely retains its stable position. One set of observations may support two or more theories or may support one and disconfirm others. ²⁰

Peirce argued consistently, throughout his long career, that scientific knowledge can never achieve certainty. Scientific knowledge is truly knowledge, but it does not provide any final truth until all inquiry has, in an ideal sense, come to a final terminus. This offers a powerful argument for cross-theory combinations of variables without the limitations of Beyond Paradigms. It is quite striking that, in the late nineteenth century, when Newton’s laws had enjoyed three centuries of corroboration, Peirce viewed them, along with all other natural science, as subject to revision.

According to the pragmatist account of empirical inquiry, social sciences like IR may seek covering laws. They are, like physics and chemistry, always subject to further scrutiny of new experience. But as long as the laws are developed following the rigorous procedures of the discipline, they constitute knowledge. Sil and Katzenstein use their opposition to covering laws to criticize the search for control. In this connection, they cite Cochran, ²¹ who interprets Dewey, famously a follower of Peirce, as holding that “covering laws or general theories are little more than fleeting efforts to exert ‘control’ over the real world.” ²² There are two troubling aspects to this use of Dewey’s view. One is that “fleeting,” or at least something far short of “permanent,” is all we can get from covering laws, in either the natural or social sciences; Peirce and almost all philosophers since have seen knowledge as fallible and always subject to further revision. Moreover, the practical applications that Sil and Katzenstein desire from IR have the effect of allowing policy-makers to formulate policies that (one hopes) will produce desired outcomes. So, the second problem is that it is precisely better control that policy-makers are able to exert if they have solidly developed and practically applicable knowledge. That is, if IR researchers produce studies of world politics that enable policy-makers to learn what motivates states and leaders and what conditions in the world are probabilistically linked with what resulting outcomes, they are more likely to identify the policy tools that will allow them to succeed in policy aims, or at least succeed with greater frequency.

Space does not permit a full explanation of the way in which at least some strands of pragmatism provide a basis for analytic eclecticism’s guidelines, especially the connection of theory with practical action. But a few comments may point the reader to further investigation. For American pragmatists, knowledge is intimately tied to experience. Many views of knowledge treat it as a matter of grasping transcendent truths, where what our minds grasp with true beliefs somehow corresponds to reality. For Peirce, too, knowledge is a form of belief. But the difference is that belief is itself action; specifically, is a “habit of action.” ²³ The experiences of policy-makers, their successes and failures in reaching intended goals of past policies they have enacted, are of the same epistemological material as the experiences of natural scientists. What separates scientific knowledge from other forms is the method by which the belief is fixed, namely, the method of science, in contrast to the methods of authority, tenacity, and pure a priori reason, the latter of which Peirce calls “agreeableness.” ²⁴ Sil and Katzenstein would be able to produce more powerful policy recommendations by recognizing that covering laws, always subject to revision pending new experiences, aid in policy-makers tools for reaching policy decisions.

Conclusion

Sil and Katzenstein make a valuable contribution to the study of IR methods by reaffirming the connection to practical applications of at least some IR scholarship and by showing that the metatheoretical principles that limit practical applicability are misplaced. They attempt to do this by downplaying the role of the philosophy of social science. But the attempt to sidestep philosophical issues is much more difficult than they acknowledge. Research in the natural and social sciences can only proceed after a range of philosophical matters are settled—even if only in a provisional way that is subject to revision by the scholarly community. While their effort to evade metatheory is underdeveloped, the line of philosophical support they allude to, American pragmatism, is indeed capable of providing answers that avoid the positivism, subjectivism, and skepticism they wish to avoid, but which space does not allow here. ²⁵ At the same time, the pragmatist alternative is capable of allowing rational debate and resolution of questions about criteria, and about the best theories among the competitors, while keeping connected to experience and practice.

In this paper I have criticized Sil and Katzenstein’s attempt to advance a methodological guideline that is not based on metatheory. I have tried, albeit briefly, to show that engaging philosophical questions can give us answers that allow us to provide a philosophical basis to move beyond the limitations on the factors that typically form the core of IR paradigms. For the Kuhnians in the social sciences, one has to acknowledge that the Kuhnian view has at least prima facie plausibility in that it helps to explain extremely durable disagreements between scholars in different schools of thought in Political Science. But by defining the term paradigm for their study as referring to things that are not inherently incommensurable and mutually exclusive, Sil and Katzenstein do not engage one of the most intractable sources of opposition to paradigm-marriages, namely, the Kuhnian-inspired view of social science. The metatheoretical element of the Kuhnian view of the social sciences, it seems, has to be addressed by means of metatheoretical critique.

Sil and Katzenstein’s argument for analytic eclecticism has value in part because it reminds IR researchers how important it is to connect their work to practice. Their approach is pragmatic in that they believe that IR research should relate to practice and also in that they at least allude to American pragmatism as providing insights. The latter offers a powerful, systematic set of solutions to the central philosophical and metatheoretical problems in IR research, including a more effective basis for the rejection of Kuhnian incommensurability and the possibility of comparing theories that might be viewed as coming from different paradigms. Overall, the conclusions Sil and Katzenstein reach about the importance of thinking beyond traditional factors at the core of research traditions is an important consideration to bear in mind when formulating and evaluating explanations in IR.