Introduction: Epistemic Communities and International Policy Coordination Author(s): Peter M. Haas Source: International Organization, Vol. 46, No. 1, Knowledge, Power, and International Policy Coordination (Winter, 1992), pp. 1-35 Published by: The MIT Press Stable URL: http://www.jstor.org/stable/2706951 . Accessed: 05/01/2011 14:40

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at . http://www.jstor.org/action/showPublisher?publisherCode=mitpress. .

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

The MIT Press is collaborating with JSTOR to digitize, preserve and extend access to International Organization.

http://www.jstor.org Introduction:epistemic communities and internationalpolicy coordination Peter M. Haas

The growingtechnical uncertainties and complexitiesof problemsof global concern have made internationalpolicy coordinationnot only increasingly necessarybut also increasinglydifficult. If decisionmakers are unfamiliarwith the technicalaspects of a specificproblem, how do theydefine state interests and develop viable solutions? What factorsshape their behavior? Under conditionsof uncertainty,what are the originsof internationalinstitutions? And how can we best studythe processes throughwhich international policy coordinationand orderemerge? While a varietyof analytic approaches have been used to address the problems of internationalcooperation, the approaches have yielded only fragmentaryinsights. At itscore, the studyof policycoordination among states involvesarguments about determinismversus free will and about the ways in which the internationalsystem is maintainedand transformed.Among the overlappingtopics of debate are whethernational behavior is determinedor broadlyconditioned by system-level factors, unit-level factors, or some complex interplaybetween the two; whetherstate policymakerscan identifynational interestsand behave independentlyof pressuresfrom the social groupsthey nominallyrepresent; and whetherstates respond consistently to opportunities to create,defend, or expandtheir own wealth and power,to enhancecollective materialbenefits, or to promotenonmaterial values.' A related question of

For theircomments on earlier versionsof this article,I am gratefulto Pete Andrews,Peter Cowhey,Barbara Crane, George Hoberg,Raymond Hopkins, Ethan Kapstein,Peter Katzenstein, StephenKrasner, Craig Murphy,John Odell, Gail Osherenko,M. J.Peterson, Gene Rochlin,and RichardSclove. 1. See, for example, Alexander E. Wendt, "The Agent-StructureProblem," Intemational Organization41 (Summer 1987), pp. 335-70; Margaret S. Archer, "Morphogenesis Versus Structuration:On CombiningStructure and Action," BritishJoumal of Sociology33 (December 1982), pp. 455-83; David Dessler, "What's at Stake in theAgent-Structure Debate?" Intemational Organization43 (Summer1989), pp. 441-73; PeterGourevitch, "The Second Image Reversed:The InternationalSources of Domestic ,"Intemational Organization 32 (Autumn 1978), pp. 881-912; Peter J. Katzenstein,ed., BetweenPower and Plenty:Foreign Economic Policies of

IntemationalOrganization 46, 1, Winter1992 ? 1992 bythe WorldPeace Foundationand the MassachusettsInstitute of Technology 2 InternationalOrganization debate is the extentto whichstate actorsfully recognize and appreciate the anarchic nature of the systemand, consequently,whether rational choice, deductive-typeapproaches or interpretiveapproaches are mostappropriate for the studyof internationalcooperation.2 In focusingon the structureof internationalor domesticpower in their explanationsof policycoordination, many authors ignore the possibilitythat actors can learn new patternsof reasoningand may consequentlybegin to pursue new state interests.While othersmention this possibility, few investi- gate the conditionsthat foster a change in state interestsand the mechanisms throughwhich the new interestscan be realized.3 In this volume of articles,we acknowledgethat systemicconditions and domesticpressures impose constraintson state behavior,but we argue that thereis stilla wide degree of latitudefor state action. How statesidentify their interestsand recognizethe latitudeof actionsdeemed appropriatein specific issue-areasof policymakingare functionsof the mannerin whichthe problems are understoodby the policymakersor are representedby those to whomthey turn for advice under conditionsof uncertainty.Recognizing that human agency lies at the intersticesbetween systemicconditions, knowledge, and nationalactions, we offeran approach thatexamines the role thatnetworks of knowledge-basedexperts-epistemic communities-play in articulatingthe cause-and-effectrelationships of complex problems,helping states identify their interests,framing the issues for collective debate, proposingspecific policies,and identifyingsalient points for negotiation. We argue thatcontrol overknowledge and informationis an importantdimension of power and that

AdvancedIndustrial States (Madison: Universityof WisconsinPress, 1977); Peter J. Katzenstein, Small States in WorldMarkets (Ithaca, N.Y.: Cornell UniversityPress, 1986); Robert Putnam, "Diplomacyand Domestic Politics:The Logic ofTwo-Level Games," IntemationalOrganization 42 (Summer 1988), pp. 427-60; Peter B. Evans, DietrichRueschemeyer, and Theda Skocpol, eds., Bringingthe StateBack In (Cambridge: CambridgeUniversity Press, 1985); Eric A. Nordlinger, "Taking the State Seriously,"in MyronWeiner and Samuel P. Huntington,eds., Understanding PoliticalDevelopment (Boston: Little,Brown, 1987), pp. 353-90; Roger Benjamin and Raymond Duvall, "The CapitalistState in Context,"in Roger Benjamin and Stephen L. Elkin, eds., The DemocraticState (Lawrence: UniversityPress of Kansas, 1985), pp. 19-57; Stephen D. Krasner, "Approaches to the State," ComparativePolitics 16 (January1984), pp. 223-46; and Howard M. Lentner,"The Conceptof the State:A Response to Krasner,"Comparative Politics 16 (April 1984), pp. 367-77. 2. See Robert0. Keohane, "InternationalInstitutions: Two Approaches,"Intemational Studies Quarterly32 (December 1988), pp. 379-96. 3. Krasneracknowledges the importanceof sharedbeliefs in explainingthe Group of 77 (G-77) cooperation and also discusses the role of shared understandingin regime creation. See the followingworks of Stephen D. Krasner: StructuralConflict (Berkeley: Universityof California Press, 1985), p. 9; and "Regimes and the Limitsof Realism: Regimes as AutonomousVariables," in Stephen D. Krasner,ed., IntemationalRegimes (Ithaca, N.Y.: Cornell UniversityPress, 1983), p. 368. Keohane notesthe possibilitythat states may learn to recalculatetheir interests, and Gilpin also acknowledgesthat states occasionally"learn to be more enlightenedin theirdefinitions of theirinterests and can learn to be more cooperativein theirbehavior." See Robert 0. Keohane, AfterHegemony (Princeton, N.J.: Princeton University Press, 1984), pp. 131-32; and RobertGilpin, Warand Changein WorldPolitics (Cambridge: Cambridge University Press, 1981), p. 227. Introduction 3 thediffusion of new ideas and informationcan lead to newpatterns of behavior and proveto be an importantdeterminant of internationalpolicy coordination. An epistemic communityis a networkof professionalswith recognized expertiseand competencein a particulardomain and an authoritativeclaim to policy-relevantknowledge within that domain or issue-area.4Although an epistemiccommunity may consist of professionalsfrom a varietyof disciplines and backgrounds,they have (1) a shared set of normativeand principled beliefs,which provide a value-based rationalefor the social actionof commu- nitymembers; (2) sharedcausal beliefs,which are derivedfrom their analysis of practicesleading or contributingto a centralset of problemsin theirdomain and whichthen serve as the basis forelucidating the multiplelinkages between possible policyactions and desired outcomes;(3) shared notionsof validity- thatis, intersubjective,internally defined criteria for weighing and validating knowledge in the domain of their expertise; and (4) a common policy enterprise-thatis, a set of commonpractices associated with a set of problems to which their professionalcompetence is directed,presumably out of the convictionthat human welfare will be enhancedas a consequence.5 The causal logic of epistemic policy coordinationis simple. The major dynamicsare uncertainty,interpretation, and institutionalization.In interna- tional policy coordination,the formsof uncertaintythat tend to stimulate demandsfor information are thosewhich arise fromthe strongdependence of states on each other'spolicy choices forsuccess in obtaininggoals and those

4. The term "epistemic communities"has been definedor used in a varietyof ways, most frequentlyto referto scientificcommunities. In thisvolume, we stressthat epistemic communities need not be made up of naturalscientists or of professionalsapplying the same methodologythat naturalscientists do. Moreover,when referringto epistemiccommunities consisting primarily of naturalscientists, we adopt a stricterdefinition than do, forexample, Holzner and Marx,who use the term"epistemic community" in referenceto a sharedfaith in the scientificmethod as a wayof generatingtruth. This ignores that such faith can still bond togetherpeople with diverse interpretationsof ambiguous data. By our definition,what bonds membersof an epistemic communityis theirshared beliefor faithin the verityand the applicabilityof particularforms of knowledgeor specifictruths. Our notionof "epistemiccommunity" somewhat resembles Fleck's notionof a "thoughtcollective"-a sociologicalgroup with a commonstyle of thinking.It also somewhatresembles Kuhn's broader sociological definitionof a paradigm,which is "an entire constellationof beliefs,values, techniques,and so on shared by membersof a givencommunity" and whichgoverns "not a subjectmatter but a groupof practitioners."See BurkhartHolzner and JohnH. Marx,Knowledge Application: The KnowledgeSystem in Society(Boston: Allyn& Bacon, 1979), pp. 107-11; LudwigFleck, Genesisand Developmentof a ScientificFact (Chicago: University ofChicago Press,1979; translatedfrom the 1935 editionprinted in German); and Thomas S. Kuhn, The Structureof ScientificRevolutions, 2d ed. (Chicago: Universityof Chicago Press, 1970), pp. 174-210, with quotes drawn frompp. 175 and 180. Regardingscientific communities, see also Michael Polanyi,"The Republic of Science,"Minerva, vol. 1, 1962,pp. 54-73. 5. Othercharacteristics of epistemiccommunities that were mentionedor discussedduring the preparationof this volume included the following:members of an epistemiccommunity share intersubjectiveunderstandings; have a sharedway of knowing;have shared patternsof reasoning; have a policy project drawingon shared values, shared causal beliefs,and the use of shared discursivepractices; and have a shared commitmentto the application and productionof knowledge.These phrases were not incorporatedin the formaldefinition listed here; theyare simplyprovided to evoke additionalnotions that are associatedwith epistemic communities. 4 InternationalOrganization which involve multiple and only partlyestimable consequences of action. Examples drawn fromthe studies presentedhere include the uncertainties about strategiesto avertnuclear destruction and theuncertainties about howto respondto the hypothesizedthreats to an invisiblelayer of ozone located seven to fifteenmiles above the earth'ssurface. These formsof uncertaintygive rise to demands forparticular sorts of information.The informationneeded does notconsist of guesses about others'intentions, about theprobability of discrete eventsoccurring, or about a state'sown abilityto pursueunilaterally attainable goals that are amenable to treatmentby various political rules of thumb. Rather,it consistsof depictionsof social or physicalprocesses, their interrela- tionwith other processes, and the likelyconsequences of actionsthat require applicationof considerablescientific or technicalexpertise. The informationis thusneither guesses nor "raw" data; it is the productof humaninterpretations of social and physicalphenomena. Epistemiccommunities are one possibleprovider of thissort of information and advice. As demandsfor such informationarise, networksor communities of specialistscapable of producingand providingthe informationemerge and proliferate.The membersof a prevailingcommunity become strongactors at thenational and transnationallevel as decisionmakers solicit their information and delegateresponsibility to them.A community'sadvice, though, is informed by its own broaderworldview. To the extentto whichan epistemiccommunity consolidatesbureaucratic power withinnational administrations and interna- tional secretariats,it stands to institutionalizeits influenceand insinuateits viewsinto broader international politics. Membersof transnationalepistemic communities can influencestate inter- ests eitherby directlyidentifying them for decision makers or by illuminating the salient dimensionsof an issue fromwhich the decision makersmay then deduce theirinterests. The decisionmakers in one statemay, in turn,influence the interestsand behaviorof otherstates, thereby increasing the likelihoodof convergentstate behaviorand internationalpolicy coordination, informed by thecausal beliefsand policypreferences of the epistemiccommunity. Similarly, epistemiccommunities may contributeto the creation and maintenanceof social institutionsthat guide internationalbehavior. As a consequence of the continuedinfluence of theseinstitutions, established patterns of cooperationin a givenissue-area may persist even thoughsystemic power concentrations may no longerbe sufficientto compelcountries to coordinatetheir behavior. By focusingon the various ways in which new ideas and informationare diffusedand takeninto account by decision makers, the epistemiccommunities approach suggestsa nonsystemicorigin for state interestsand identifiesa dynamicfor persistent cooperation independent of the distributionof interna- tional power. It assumes that state actors are uncertaintyreducers as well as power and wealthpursuers. It also seeks to explainthe substantivenature of coordinatedpolicy arrangements, a subjecton whichmany structural analysts are notablysilent. Yet to some extent,the approach supplementsstructural Introduction 5 theoriesof internationalbehavior: in responseto newknowledge articulated by epistemiccommunities, a state mayelect to pursue entirelynew objectives,in whichcase outcomesmay be shaped by the distributionof informationas well as by the distributionof power capabilities.Table 1 presentsa schematized outline of the epistemiccommunities approach and compares it with other approaches to the study of policy change that have been advanced by internationalrelations scholars. Pursuingthe epistemiccommunities approach, contributors to thisvolume analyze the impactof fiveepistemic and two epistemic-likecommunities on decision makingin a varietyof issues concerningthe internationalpolitical economy, internationalsecurity, and the environment.In analyzing the processes leading to policycoordination in a specificissue-area, each author describesthe membershipand shared beliefsof an expertcommunity, traces thecommunity's actions, and discussesits impact. By comparingthe beliefs and behaviorof policymakersin one countryover timeand by comparingthem in countriesin whichexpert communities were activeversus those in whichthey were not, the authors tryto specifythe extent to which decision-making processeswere influencedby the communityas opposed to thepolitical factors and actorsemphasized in otherapproaches to internationalrelations. The articlesby William Drake and KalypsoNicolaidis, Emanuel Adler,M. J. Peterson,and Peter Haas investigatethe ways in whichepistemic communities initiallyframed the issues forcollective debate, thereby influencing subsequent negotiationsand bringingabout theirpreferred outcomes to the exclusionof othersin the cases involvingtrade in services,nuclear arms control,manage- ment of whaling,and protectionof stratosphericozone. In the whalingand ozone cases, the authors also outline the role that epistemiccommunities played in identifyingspecific policies fornational and collectiveadoption. In the studyregarding the principlesand practicesof food aid, RaymondHopkins traces the changes in the beliefs and understandingsof the epistemic communitythat had a hand in the food aid regimeand linksthese changesto regimereforms. Ethan Kapstein's analysisof bankingregulators and G. John Ikenberry'sanalysis of economistsinvolved in the Anglo-Americanpostwar economicsettlement both shed lighton the epistemiccommunities approach by discussingfactors that differentiate these expertgroups from the epistemic communitiesdiscussed in the other case studies included here. And James Sebenius adds an additionalviewpoint in hiscommentary on thecommonalities and differencesbetween the epistemiccommunities approach and negotiation analysis. While all of the case studiesin thisvolume consider the arrayof politicaland systemicconstraints within which expert communitiesoperate, Ikenberry focusesin particularon how politicalfactors can impede the applicationof the consensualviews of specialists.In his analysisof postwareconomic manage- ment,he thus offersa limitingcase, indicatingthat epistemicagreement was possible only in those areas removed fromthe political whirl.One of the (U

64 64 a) 0 0, r. t Cd 64 : 04 0. C14 0 ;- " Z 064 4;4 00 cl 7; u 41 o c7j Cld -4 cld 10 -0 C's 0 0 oo > Ct .4 64 a-, :Z: ,6-J. - , 'a >0 1-4 O 0 CZ Cld 4- 4-

cl 41 0 O Cd -10 Cld n > 4-4 0 0 CZ0 0 C'S oo C's E 0- Cld (A =3 C's C'd 0 0. -, 14.) CZ 41.1 Ln (1) 64 C'd C)- 0 to 0 > -0 ?, 1.4 E Z ct =3 = C'S 0 64 x 0 u- 0 C C) C') C) to +,co, v) 0.4 cts 64 n C) -CS 64 -, C'S 2: ct a) - - t4-4 0 cn r. 'n 0 0. 0 Q ";LI C's Cld cis ct 4) Cld O. C'sto u (A v) (A C4

t:s

ci r 0 Cl Cld (A 1.6 u C'S C13 cl 0 C:L., .- 3 E4 ;.,.r

C's C'S 4-4 C's 0 Cld 64 0 64 -st 0 '= O 0 ;-, U -a 4-4 .- 4- 4- ct cts C's :z CA E = 00 4-4

C'd (A 0 0. 0 64 E 0 tn C's C's 0 4-i OD Z -5 z CL4-; .0 cld N 0 00 C's c', = Z u = W u --I

oo col M M CN Cd 6-4

Cld to 0-1; C's Cld 0 Cld 0 In u C'S 1.4 64 14 ao cld M M cld C,* u C'd cld

t,4 cli 0 C'd Cld C's O cr CN q =$ ct -- C's " " E t4,;4 .- 0) a-, Cd = CZ - 64 ;-4 64 64 0 C13 Q 4-j pub 64 U 0 14) cld C:L, C-)

6, ct -4 0 > O64 =(A Cld C,* C's

0 0 O ct O 04 . . cn 0) " U W5 >, = C's 0 Cl >.,. C) (U 0 W 0 to - A) 'A - 64 O 75 O 0 04

04 C's 04 Cld 0 M C- . Z C', CN N C > 4- ct C's ci so r- -= a) C's C,* Z C; 0 U4 64 U4 ;.4 4-1 C13 rol cd 0 04 0-4 4-i 0 (A 0 04 0, 04 (A 6, 0 1.) cl 04 0 C's Cld 0 O., w z a) A4 (Z4 Introduction 7 conclusionsthat can be drawnfrom Ikenberry's study, as well as fromearlier studiesof epistemic-likecommunities presented elsewhere,6 is thatwhile the formof specificpolicy choices is influencedby transnational knowledge-based networks,the extentto whichstate behaviorreflects the preferencesof these networksremains strongly conditioned by the distribution of power internation- ally. Thus, the range of impact that we might expect of epistemic and epistemic-likecommunities remains conditioned and bounded byinternational and nationalstructural realities. The extentof thatconditioning-the amount offlexibility in theinternational system available for reflection and understand- ingin the face of powerand structure-isthe focusof thisvolume.

The internationalsetting for epistemic communities The modern administrativestate: expansion, professionalization, and deferenceto the "knowledge elite" Many of the major dimensionsof contemporaryinternational relations can be tracedto the late nineteenthcentury, when crafts and guildswere declining

6. A numberof earlier studies focusingon the interplaybetween expertise,technical issues, consensualknowledge, and statepower have consideredthe role of epistemic-likecommunities in the decision-makingprocess. At the level of internationalorganizations, such studieshave been undertakenwith regard to widevariety of issue-areasand have demonstratedthat webs of nonstate actorsprovided information and were involvedin the shapingof agendas and the definingof state interests.While all of these studiescannot be listedhere, a fewexamples show the rangeof areas analyzed: Robert W. Russell, "TransgovernmentalInteraction in the InternationalMonetary System,1960-1972," Intemational Organization 27 (Autumn 1973), pp. 431-64; WilliamAscher, "New DevelopmentApproaches and the Adaptabilityof InternationalAgencies: The Case of the ," IntemationalOrganization 37 (Summer 1983), pp. 415-39; Barbara B. Crane and JasonL. Finkle,"Population Policy and WorldPolitics," paper presentedat the FourteenthWorld Congress of the InternationalPolitical Science Association,Washington, D.C., 28 August to 1 September1988; PeterM. Haas, Savingthe Mediterranean: The Politics of Intemational Environmen- tal Protection(New York: Columbia UniversityPress, 1990); Barbara Johnson,"Technocrats and the Managementof InternationalFisheries," Intemational Organization 29 (Summer 1975), pp. 745-70; and Warren S. Wooster, "Interactions Between Intergovernmentaland Scientific Organizationsin Marine Affairs,"Intemational Organization 27 (Winter 1973), pp. 103-13. For examplesof studiesin comparativepolitics that discuss the role of epistemic-likecommunities in the developmentand enforcementof commonpolicies, see MargaretWeir and Theda Skocpol, "State Structuresand the Possibilitiesfor 'Keynesian' Responses to the Great Depression in Sweden,Britain, and the United States," in Evans, Rueschemeyer,and Skocpol,Bringing the State Back In, pp. 107-68;Peter A. Hall, Govemingthe Economy (Cambridge: Polity Press, 1986), pp. 275 ff.;and AnthonyKing, "Ideas, Institutions,and Policiesof Governments:A ComparativeAnalysis" (in 3 parts),British Joumal of Political Science 3 (Julyand October 1973), pp. 291-313 and 409-23. With respect to policycoordination, it is worthstressing that even if actors believe that their commonunderstandings will contributeto enhancingthe collectivegood, serious unanticipated consequencesare possible; see StephenVan Evera, "Cult of the Offensiveand the Originsof the First World War," IntemationalSecurity 9 (Summer 1984), pp. 58-107. For examples of purely national studies that discuss the role of epistemic-likecommunities in transformingstate preferences,see John Odell, U.S. IntemationalMonetary Policy (Princeton, N.J.: Princeton UniversityPress, 1982); Emanuel Adler, "Brazil's Domestic ComputerIndustry," International Organization40 (Summer1986), pp. 673-705; and Dennis Hodgson,"Orthodoxy and Revisionism in AmericanDemography," Population and DevelopmentReview 14 (December 1988), pp. 541-69. 8 InternationalOrganization and scientificand engineeringexpertise were increasinglyapplied to commer- cial research,development, and governance.7Scientific rationality began to prevail over alternativeparadigms of knowledge as a model for decision- makingscience as well,although it did not reach itspeak untilabout fiftyyears later,when logical positivism and theideas ofthe Vienna Circlewere embraced and the entryof white-coatedprofessionals into the process became morewidespread. As HarveyBrooks observedin 1965, "Much of the historyof social progressin the TwentiethCentury can be describedin termsof the transferof wider and wider areas of public policy from politics to expertise."8With the proliferationof governmentministries and agencies to coordinateand handle manynew tasks, has become an increasingly importantbureaucratic function,9 and the expertiserequired has extendedto a widerrange of disciplinesthan ever before. The domainof publicgovernance has also growncorrespondingly technical. Despite the fact that numerous ministriesestablished for conductingWar World II were decommissionedin subsequent years, the total number of ministriestripled duringthe period fromthe late 1940s to the mid-1970s. Around 1950, there were 70 independentcountries with 850 ministries,or roughly12 ministriesper country.By 1975, there were 140 independent countrieswith 2,500 ministries, or nearly18 ministriesper country,indicating a strong shifttoward more active social regulation.10The rapid growthof governmentagencies was particularlyevident in the United States,where two economic regulatoryagencies and fivemajor social regulatoryagencies were

7. Whilethe transferof authorityto thesphere of the secular and the rationalcan be tracedback to the eighteenthcentury and the grantingof Noblesse de la Robe in France, the integrationof scientistsand engineersinto a new rationalizedcorporate structure really began withthe second industrialrevolution of the 1880s. For backgroundinformation, see FranklinL. Ford, Robe and Sword(New York: Harper, 1953), pp. 248-52. Regardingthe accelerationof technicallygrounded forms of governance and decision making, see David C. Mowery and Nathan Rosenberg, Technologyand thePursuit of Economic Growth(Cambridge: Cambridge University Press, 1989); JoAnneYates, ControlThrough Communication (Baltimore, Md.: JohnsHopkins University Press, 1990); AlfredD. Chandler,Strategy and Structure:Chapters in theHistory of the American Industrial Enterprise(Garden ,N.Y.: Doubleday, 1966); Alfred D. Chandler, The VisibleHand: The ManagerialRevolution in AmericanBusiness (Cambridge,Mass.: Belknap Press, 1977); and A. Hunter Dupree, ed., Science and theEmergence of Modem America,1865-1916 (Chicago: Rand McNally,1963). 8. HarveyBrooks, "Scientific Concepts and CulturalChange," Daedalus 94 (Winter1965), p. 68. 9. See Ezra N. Suleiman, ed., Bureaucratsand PolicyMaking: A ComparativeOverview (New York: Holmes & Meier, 1984); Joel D. Aberbach, Robert D. Putnam,and Bert A. Rockman, Bureaucratsand Politiciansin WestemDemocracies (Cambridge, Mass.: HarvardUniversity Press, 1981); JamesQ. Wilson,ed., ThePolitics of Regulation (New York: Basic Books, 1980); and Terry M. Moe, "The Politicsof BureaucraticStructure," in JohnE. Chubb and Paul E. Patterson,eds., Can theGovemment Govem? (Washington, D.C.: BrookingsInstitution, 1989), pp. 267-328. 10. See Jean Blondell, The Organizationof Govemments(Beverly Hills, Calif.: Sage, 1982), pp. 195-96. For data on the professionalbackgrounds of ministersand individualsoccupying other ministerialposts, see JeanBlondell, Govemment Ministers in the Contemporary World (Beverly Hills, Calif.:Sage, 1982). Blondell notes that9.5 percentof the ministersserving between 1945 and 1981 could be considered"specialists," with most of thisgroup consisting of civilengineers, electrical engineers,and agronomists. Introduction 9 createdduring the five-yearperiod from 1970 to 1975,while the federalbudget allocationsfor economic and social regulationgrew by 157 percentand 193 percent,respectively.11 Governmentsof industrializedcountries also developed a greaterinterest in and began to establish futures-orientedresearch bodies."2With decolonization and the frequent emulation of the Western development models,the attitudesof thesegovernments spread to thoseof the ThirdWorld as well.13This was reflected,for example, in the factthat the governmentsof 118 countriesestablished agencies responsiblefor environmental and natural resourcesbetween 1972 and 1982. The processof professionalization accompanied the expansion of bureaucra- cies in many countries.In the United States, for example, the number of scientificand technicalpersonnel employed by the federalgovernment grew from123,927 in 1954 to 189,491in 1976 to 238,041in 1983.This mere doubling of the numberover nearlythree decades obscuresother pertinent changes in individualexpertise in U.S. governmentemployees. From 1973 to 1983 alone, the proportionof scientistsand engineerswith doctoral degrees grewby 51 percent,and the proportionwith masters degrees grew by 44 percent.During thesame period,the government was increasingits staff of scientists, engineers, and computerspecialists by 4 percentper year,while the increase for other personnelwas only 2 percent per year. By 1983, scientists,engineers, and computerspecialists comprised 15 percent of the governmentwhite-collar workforce, in contrastto 13 percentin 1973 and in contrastto 6 percentof the nongovernmentwork force in 1983.14

11. See GiandomenicoMajone, "RegulatoryPolicies in Transition,"Jahrbuch fur neue politische Okonomie(Tubingen: J. C. B. Mohr, 1984), p. 158. For discussionsof the progressiveexpansion and professionalizationof bureaucraciesin the United States,see StephenSkowronek, Building a NewAmerican State: The Expansionof National Administrative Capacities, 1877-1920 (Cambridge: Cambridge UniversityPress, 1982); Charles Maier, ed., ChangingBoundaries of the Political (Cambridge:Cambridge University Press, 1987); Louis Galambos, ed., The New AmericanState: Bureaucraciesand Policies Since WorldWar II (Baltimore,Md.: JohnsHopkins UniversityPress, 1987); Bruce L. R. Smith,American Since World War II (Washington,D.C.: BrookingsInstitution, 1990), pp. 28-35; RobertGilpin and ChristopherWright, eds., Scientistsand NationalPolicy Making (New York: ColumbiaUniversity Press, 1964); and George Kistiakowsky,A Scientistat theWhite House (Cambridge,Mass.: HarvardUniversity Press, 1976). 12. Yehezkel Dror, PolicymakingUnder Adversity (New Brunswick,N.J.: TransactionBooks, 1986). 13. JawaharlalNehru, arguingthat less developed countriesmust also turntoward science, offeredthe followingrationale: "It is science alone that can solve the problemsof hungerand poverty,of insanitationand illiteracy,of superstitionand deadeningcustom and tradition,of vast resourcesrunning to waste,of a richcountry inhabited by starvingpeople.... Who indeed could affordto ignorescience today?At everyturn we have to seek its aid.... The futurebelongs to science and thosewho make friendswith science." Nehru is quoted byMax F. Perutzin Is Science Necessary?(New York: E. P. Dutton,1989), p. vii. 14. See National Science Foundation,Federal Scientificand TechnicalWorkers: Numbers and Characteristics,1973 and 1983 (Washington,D.C.: National Science Foundation,1985), pp. 1-2. Duringthe periodfrom 1973 to 1978,the increasein scientists,engineers, and computerspecialists occurredlargely outside the Defense Department. 10 InternationalOrganization

These trendscontributed to the emergenceof what Dorothy Nelkin has called "the policy role of the knowledgeelite."'5 The proliferationof new agenciesand thepractice of staffingthem with professionals also contributedto the erosion of centralized control over public , which has occurreddespite widespread efforts since WorldWar II to curbthe discretion of bureaucraticadministrators. As Joel Aberbach,Robert Putnam,and Bert Rockman found in their surveyof public servantsin the major Western industrializedsocieties, the overwhelmingmajority of civil servantsregard themselvesas technicians,policymakers, and brokers,unlike elected officials, who primarilyregard themselves as advocates and partisans."6In the case of professionals,the degree to whichthey are sympatheticwith the missions of the agencies in whichthey work is influencedby a varietyof factors,including the extentof theirspecialized training,the fieldin whichthey were trained,and theirpersonal views."7In otherwords, "where theystand" is associated with factorsother than "where they sit." In internationalbureaucracies, such as the United Nations (UN), technical responsibilitieshave proliferatedsince the inaugurationof the International GeophysicalYear in 1957,yet the trainingof personnelwithin the UN system has not kept pace. Only 13 percentof the staffmembers have doctorates,and less than 50 percenthold more than a firstuniversity degree."8 In 1986,when the UN employed 54,000 people worldwide,about 18,000 were serving "professional" functions,4,000 to 5,000 of which were "substantive" in nature.19Nevertheless, the budgetingof fundsin the UN indicatesa shiftaway fromthe more traditionalpolitical and securityconsiderations of the General Assemblyand towardthe more technical concerns of specialized agencies.20

15. See Dorothy Nelkin, "ScientificKnowledge, Public Policy, and Democracy,"Knowledge Creation,Diffusion, Utilization 1 (September1979), p. 107. See also DorothyNelkin, "The Political Impact of Technical Expertise,"Social Studiesof Science 5 (February 1975), pp. 35-54. For a criticalview of the role of scientistsin decisionmaking, see Joel Primackand FrankVon Hippel, Adviceand Dissent(New York: Basic Books, 1974). 16. See Aberbach,Putnam, and Rockman,Bureaucrats and Politiciansin WestemDemocracies. 17. See Samuel P. Hays, Beauty,Health, and Permanence:Environmental Politics in the United States, 1955-1985 (Cambridge: Cambridge UniversityPress, 1987), pp. 357-59; William T. Gormley,Jr., "Professionalism Within Environmental Bureaucracies: The PolicyImplications of PersonnelChoices," La FolletteInstitute of Public Affairs, occasional paper no. 1, Madison,Wisc., December 1986; and Thomas M. Dietz and Robert Rycroft,The Risk Professionals(New York: Russell Sage, 1987). 18. See Peter Fromuthand Ruth Raymond,"U.N. PersonnelPolicy Issues," in UnitedNations Managementand Decision-MakingProject (New York: United Nations, 1987), p. 13. See also Douglas Williams,The SpecializedAgencies and theUnited Nations (London: C. Hurst,1987), p. 254. 19. See AnthonyMango, "The Role ofSecretariats of International Institutions," in Paul Taylor and A. J. R. Groom,eds., IntemationalInstitutions at Work(New York: St. Martin'sPress, 1986), pp. 40-43. Based on his surveyof 75 percentof the UN's professionalstaff, Mango concludedthat about 4,000 servedkey functions "in all areas of humanendeavor from peace and disarmamentto health,nutrition, industry, communications, and the environment."Thus, forthe full100 percent of the staff,the figuremay have reached5,000. 20. The percentageof the UN budgetallocated forspecialized agenciessteadily rose from45.1 percentin 1950 to 60.5 percentin 1985. With the adoption of the Kaasebaum amendment,the percentage has remained at the 1985 level. Two specialized areas involvingscience and Introduction 11

Thus, the expansion and professionalizationof bureaucracies and the growingtechnical nature of problems have fostered an increase in the deferencepaid to technicalexpertise and, in particular,to thatof scientists."In modernsocieties," Barry Barnes and David Edge have argued,"science is near to beingthe source of cognitiveauthority: anyone who would be widelybelieved and trustedas an interpreterof nature needs a license fromthe scientific community."21 As several studies have pointed out, policymakersand leaders typically expectto remainin controleven whendelegating authority.22 Questions arise, then,about the effectsthat the interactionof expertsand politicianshave on policychoices. Manyexpected that scientists, because of theircommon faith in the scientificmethod, would make policymakingmore rational.Yet even in cases involvingwhat is regardedas a technicalissue, policymakingdecisions generallyinvolve the weighing of a numberof complexand nontechnicalissues centeringaround who is to get what in societyand at what cost. Despite the veneerof objectivityand value neutralityachieved by pointingto the inputof scientists,policy choices remain highlypolitical in their allocative conse- quences.23Especially in cases in whichscientific evidence is ambiguousand the expertsthemselves are splitinto contending factions, issues have tendedto be resolved less on their technical merits than on their political ones. That scientistsworking within the bureaucracyhave a commonfaith in the scientific methoddoes not guaranteetheir solidarity, nor does it make themimmune to pressuresfrom the institutionsin whichthey work or frompolitical temptation. Studies of science policyand of scientists'effects on Americanpolicy and regulationhave been at best equivocal, findingonly slight and transitory

technology-thatof food and agricultureand thatof health-have come to controlover 25 percent of the resourcesof the UN system.See UN documentnos. A/1312, A!3023, A!6122, A!7608, A/42/683,and A/10360,UN, New York, 1951,1956, 1967,1971, 1976,and 1986,respectively. The highestpostwar rates of growthfor new internationalscientific and professionalassociations (ISPAs) was also in the areas of science and technology,followed by economicsand finance.See Diana Crane, "AlternativeModels of ISPAs," in WilliamM. Evan, ed., Knowledgeand Powerin a Global Society(Beverly Hills, Calif.: Sage, 1981), p. 30; and Werner Feld, "Nongovernmental Entitiesand the InternationalSystem," Orbis 15 (Fall 1971), pp. 879-922. 21. See BarryBarnes and David Edge, "General Introduction,"in BarryBarnes and David Edge, eds., Sciencein Context(Cambridge, Mass.: MIT Press,1982), p. 2. For an argumentthat the influenceof scientificspecialists often extends to areas beyondtheir formal training, see AlvinM. Weinberg,"Science and Trans-Science,"Minerva 10 (April 1972), pp. 209-22. 22. See TerryM. Moe, "The New Economics of Organization,"American Joumal of Political Science 28 (November 1984), pp. 739-77; and JonathanBendor, Serge Taylor,and Roland Van Gaalen, "Stackingthe Deck: BureaucraticMissions and PolicyDesign," American Political Science Review81 (September1987), pp. 873-96. 23. See Yaron Ezhrahi,"Utopian and PragmaticRationalism: The PoliticalContext of Scientific Advice," Minerva18 (Spring 1980), pp. 111-31; Robert F. Rich, "The Pursuitof Knowledge," KnowledgeCreation, Diffusion, Utilization 1 (September 1979), pp. 6-30; Robert H. Socolow, "Failures of ," in Harold A. Feiveson,Frank W. Sinden,and Robert H. Socolow, eds., BoundariesofAnalysis (Cambridge, Mass.: Ballinger,1976); and PeterdeLeon, Advice and Consent: TheDevelopment of the Policy Sciences (New York: Russell Sage, 1988). 12 InternationalOrganization influenceby scientists.24 Similarly,early studies of policy coordinationin technicalareas have demonstratedthat state decision makerswere no more willingto sacrificeautonomy in these areas than in issues of security;that as theirgovernments grew cognizant of the politicalcosts of technicalcoordina- tion, they grew more unwillingto coordinate their actions; and that many foreignministries proved resistant to anyencroachment by technical functional ministrieson theirsphere of responsibility.25Thus, in spite of the increasing involvementof technocratsin governmentinstitutions and contraryto the hopes of functionalistssuch as David Mitrany,outcomes in technicalissues provedlittle different from those of moreconventional high politics. Unlike the functionalists,who turnedtheir attention to the developmentof commonactivities and the transferof technocraticloyalty to a superordinate authority,the concern of the contributorsto this volume is with stylesof policymakingand changes in the patternsof policymakers'reasoning. As argued below, the increasinguncertainties associated with many modern responsibilitiesof internationalgovernance have led policymakersto turnto new and differentchannels of advice, oftenwith the resultthat international policycoordination is enhanced.

Decision-makingprocesses: complexity,uncertainty, and the turnto epistemic communitiesfor advice Among the factorsthat have contributedto the uncertaintiesfaced by decision makers are the increasinglycomplex and technical nature of the ever-wideningrange of issues consideredon the internationalagenda, includ- ing monetary,macroeconomic, technological, environmental, health, and populationissues; the growthin the complexityof the internationalpolitical systemin termsof the numberof actorsand the extentof interactions;and the expansionof theglobal economyand themodern administrative state.26 Forced

24. See DorothyNelkin, ed., Controversy:Politics of TechnicalDecisions (BeverlyHills, Calif.: Sage, 1979); Michael Mulkay,Science and the Sociologyof Knowledge(London: Allen & Unwin, 1979); WilliamKornhauser, Scientists in Industry(Berkeley: University of CaliforniaPress, 1962); and PeterWeingart, "The ScientificPower Elite: A Chimera,"in NorbertElias, HerminioMartins, and Richard Whitley,eds., ScientificEstablishments and Hierarchies(Dordrecht, Netherlands: Reidel, 1982), pp. 71-88. 25. See JohnG. Ruggie,"Collective Goods and FutureInternational Collaboration," American PoliticalScience Review66 (September 1972), pp. 874-93; Henry R. Nau, National Politicsand IntemationalTechnology (Baltimore, Md.: Johns Hopkins UniversityPress, 1974); and Roger Williams,European Technology:The Politics of Cooperation(New York: Wiley,1974). 26. For discussions of these changes and the increasingsocial, economic, and political interdependencethat accompanied them, see, forexample, Todd R. La Porte,ed., OrganizedSocial Complexity(Princeton, N.J.: Princeton University Press, 1975); Marion Levy,Modemization and the Structureof Societies(Princeton, N.J.: Princeton University Press, 1966); Alex Inkeles,"Emerging Social Structureof the World," WorldPolitics 27 (July1975), pp. 467-95; Karl Polyani,The Great Transformation(Boston: Beacon Press, 1944); RichardCooper, TheEconomics of Interdependence (New York: McGraw-Hill,1968); Robert 0. Keohane and JosephS. Nye,Power and Interdepen- dence:World Politics in Transition(Boston: Little,Brown, 1977); Edward Morse,Modemization and Introduction 13 to deal witha broaderrange of issues thanthey were traditionallyaccustomed to, decision makershave turnedto specialiststo amelioratethe uncertainties and help themunderstand the currentissues and anticipatefuture trends.27 Complexitytests the limitsof human understanding.Although knowledge maybe betterthan it was in thepast about thedynamics of anyof the individual issues, the natureof the interactionsbetween themis particularlydifficult to graspand deal witheffectively in thepolicymaking process. For example,to the extentthat economic interdependence and a globalizedeconomy require policy coordinationamong countries to pursuedomestic goals, the domesticagendas and internationalagendas have become increasinglylinked, yet decision makershave oftenfailed to comprehendthe complexlinkages. The result,as some analystshave complained,is that"to a fargreater extent than in thepast, the individualswho mustmake the difficulteconomic choices in Washington are in the dark."28 Similarly,in the case of internationalenvironmental issues, decision makers are seldom certainof the complexinterplay of componentsof the ecosystem and are thereforeunable to anticipatethe long-term consequences of measures designed to address one of the many environmentalissues under current consideration.Without the help of experts,they risk making choices thatnot only ignore the interlinkageswith other issues but also highlydiscount the uncertain future,with the result that a policy choice made now might jeopardize futurechoices and threatenfuture generations. Conditionsof uncertainty,as characterizedby Alexander George, are those underwhich actors must make choices without"adequate informationabout the situationat hand" or in the face of "the inadequacyof available general knowledgeneeded forassessing the expectedoutcomes of differentcourses of

the Transformationof IntemationalRelations (New York: Free Press, 1976); John G. Ruggie, "Continuityand Transformationin theWorld Polity," World Politics 35 (January1983), pp. 261-85; and Stephen Toulmin,Cosmopolis: The HiddenAgenda of Modemization(New York: Free Press, 1990). For discussionsof increasingecological interdependence, see W. C. Clark and R. E. Munn, eds., of the Biosphere (Cambridge: Cambridge University Press, 1986); and Organizationfor Economic Cooperation and Development (OECD), Economic and Ecological Interdependence(Paris: OECD, 1982). 27. Regardinguncertainty and the turnto specialistsfor advice, see Dror, PolicymakingUnder Adversity,pp. 60-61; Harold Wilensky,Organizational Intelligence (New York: Basic Books, 1967); Guy Benveniste,The Politicsof Expertise(San Francisco:Boyd & Fraser, 1977); WilliamAscher, "New DevelopmentApproaches and the Adaptabilityof InternationalAgencies"; J. Hirshleifer and JohnG. Riley,"The Analyticsof Uncertainty and Information:An ExpositorySurvey," Joumal of Economic Literature17 (December 1979), pp. 1375-1412; GeoffreyBrennan and James M. Buchanan,The Reason ofRules (Cambridge:Cambridge University Press, 1985), chap. 2; Zdenek J. Slouka, "InternationalLaw Making:A View fromTechnology," in Nicholas GreenwoodOnuf, ed., Law Makingin the Global Community(Durham, N.C.: Carolina Academic Press, 1982), p. 149; Langdon Winner,"Complexity and the Limitsof Human Understanding,"in La Porte,Organized Social Complexity,pp. 40-76; and Ina Spiegel-Rosingand Derek De Solla Price, eds., Science, Technologyand Society(Beverly Hills, Calif.: Sage, 1977). 28. C. Michael Aho and Marc Levinson,After Reagan: Confrontingthe Changed World Economy (New York: Council on ForeignRelations, 1988), p. 8. 14 InternationalOrganization action."29A growingnumber of issues and problemsfaced by decisionmakers fitthis description.That this is true indeed underminesthe utilityof many conventionalapproaches to internationalrelations, which presume that a state's self-interestsare clear and thatthe ways in whichits interestsmay be mostefficaciously pursued are equallyclear.30 As severalauthors have warned, however,misperceptions of the natureof the internationalsetting, as well as misperceptionsof others'intentions and actions,are mostlikely to occurunder conditionsof uncertainty.3" Decision makers do not always recognize that their understandingof complexissues and linkagesis limited,and it oftentakes a crisisor shock to overcomeinstitutional inertia and habit and spur themto seek help froman epistemiccommunity. In some cases, informationgenerated by an epistemic communitymay in factcreate a shock,as oftenoccurs with scientific advances or reportsthat make theirway into the news, simultaneouslycapturing the attentionof the public and policymakersand pressuringthem into action. In estimatingthe effectthat shocks or crises have on decision makers, the contributorsto thisvolume are influencedby two schools of thought.Those informedby organizationtheory presume that decision makers will seek informationand defer to actors who are able to provide credible technical advice. Those applyingthe politicalliterature presume that leaders will only deferto technicaladvice thatwill enable themto pursuepreexisting ends and to expand politicalcoalitions. This does not,however, rule out the possibility thatleaders would deferto specialistsunder circumstancesin whichthey are uncertainabout what course of action is in theirown politicalinterests, nor does it excludethe possibilitythat their delegation of authoritywill persist past the initialcrisis or shock: The conceptof uncertaintyis thusimportant in our analysisfor two reasons. First,in the face of uncertainty,and more so in the wake of a shock or crisis, manyof the conditionsfacilitating a focuson powerare absent.It is difficultfor leaders to identifytheir potential political allies and to be sure of what strategiesare most likelyto help them retain power. And, second, poorly understoodconditions may create enoughturbulence that established operat- ing procedures may break down, making institutionsunworkable. Neither powernor institutionalcues to behaviorwill be available,and new patternsof actionmay ensue.

29. Alexander George, PresidentialDecision Making in ForeignPolicy: The EffectiveUse of InformationandAdvice (Boulder,Colo.: WestviewPress, 1980), pp. 26-27. 30. Armen A. Alchian, "Uncertainty,Evolution, and Economic Theory,"Joumal of Political Economy,vol. 58, 1950,pp. 211-21. 31. See ArthurA. Stein,Why Nations Cooperate (Ithaca, N.Y.: CornellUniversity Press, 1990), chap. 3; Robert Jervis,Perception and Misperceptionin IntemationalPolitics (Princeton,N.J.: PrincetonUniversity Press, 1976); Glenn H. Snyderand Paul Diesing, ConflictAmong Nations (Princeton,N.J.: PrincetonUniversity Press, 1977); and Yaacov Y. I. Vertzberger,The Worldin TheirMinds (Stanford, Calif.: Press, 1990). Introduction 15

Under conditionsof uncertainty,then, decision makers have a varietyof incentivesand reasons forconsulting epistemic communities,32 some of them more politicallymotivated than others. First, followinga shock or crisis, epistemiccommunities can elucidate the cause-and-effectrelationships and provideadvice about the likelyresults of various courses of action. In some cases, theycan help decisionmakers gain a sense ofwho thewinners and losers would be as the result of a particularaction or event, as was the case in considerationsabout banning chlorofluorocarbonuse or facing a possible environmentaldisaster. Decision makersseldom apply the typesof decision- making heuristicsthat scientistsapply under conditions of uncertainty.33 Indeed, as JonElster argues,decision makers generally "are unable to assign numericalprobabilities to thevarious answers of what will happen. They can at most list the possible answers,not estimatetheir probabilities."34 While they may desire probabilitystatistics and similardata forpurposes of determining thegravity of a situation,they may also use the informationfor other purposes, such as justifyinga "wait and watch" policyand deferringresponsibility until thefuture, when other actors may be held responsible. Second, epistemiccommunities can shed lighton the natureof the complex interlinkagesbetween issues and on the chain of eventsthat mightproceed either fromfailure to take action or from institutinga particularpolicy. Informationis at a premiumin the face of possible systemicvolatility, when effortsto solve or curb a problem in one domain or issue-area may have unanticipatednegative feedback effects on others. Third,epistemic communities can help definethe self-interestsof a stateor factionswithin it. The processof elucidatingthe cause-and-effectrelationships of problemscan in factlead to the redefinitionof preconceivedinterests or to the identificationof new interests. Fourth,epistemic communities can help formulatepolicies. Their role in this regardwill depend on the reasons forwhich their advice is sought.In some cases, decisionmakers will seek advice to gain informationwhich will justify or legitimatea policythat theywish to pursue for political ends. An epistemic community'sefforts might thus be limitedto workingout the details of the policy,helping decision makersanticipate the conflictsof interestthat would emerge with respect to particularpoints, and then building coalitions in support of the policy. If the policy is institutedand problems ensue, the decisionmakers have theoption of pointing to theinformation given to themby

32. In Marketsand Hierarchies(New York: Free Press, 1975), Oliver Williamsonargues that underconditions of uncertainty,organizations are likelyto develop internalmethods to generate moreand betterinformation instead of turningto externalsources. 33. See Daniel Kahneman,Paul Slovic,and Amos Tversky,eds., JudgementUnder Uncertainty: Heuristicsand Biases (Cambridge:Cambridge University Press, 1982). 34. See Jon Elster, ExplainingTechnical Change (Cambridge: Cambridge UniversityPress, 1983), p. 185. See also JohnD. Steinbruner,The CyberneticTheory of Decision (Princeton,N.J.: PrincetonUniversity Press, 1974), pp. 17-18; and HerbertSimon, "Rationality as Process and as Productof Thought," American Economic Review 68 (May 1978), pp. 1-16. 16 InternationalOrganization expertsand spreadingthe blame.35 Again, however, it is importantto stressthat epistemiccommunities called in forpolitical reasons may succeed in imposing theirviews and movingtoward goals otherthan those initially envisioned by the decisionmakers. In less politicallymotivated cases, epistemiccommunities have a greater hand in thevarious stages of the policymakingprocess, including the introduc- tionof policyalternatives, the selectionof policies,and thebuilding of national and internationalcoalitions in supportof the policies. "The definitionof the alternatives,"as E. E. Schattschneidernoted, "is the supremeinstrument of power."36By pointingout whichalternatives are notviable on thebasis of their causal understandingof the problems to be addressed, the community memberscan limitthe range of alternativesunder consideration.While the actual choice of policiesremains the domainof the decisionmakers, it can also be influencedby community members. As HerbertSimon points out, almost all organizationsengage in some formof "satisficing"or proceduralrationality in theirconsideration of policyalternatives.37 If rationalityis bounded,epistemic communitiesmay be responsible for circumscribingthe boundaries and delimitingthe options.

Distinguishingepistemic communities from other groups

As outlined earlier, members of epistemic communitiesnot only hold in commona set of principledand causal beliefsbut also have shared notionsof validityand a shared policy enterprise.Their authoritativeclaim to policy- relevant knowledge in a particular domain is based on their recognized expertisewithin that domain. These featuresdistinguish epistemic communi- tiesfrom other groups often involved in policycoordination. Epistemiccommunities need not be made up of naturalscientists; they can consistof social scientistsor individualsfrom any discipline or professionwho have a sufficientlystrong claim to a bodyof knowledgethat is valued bysociety. Nor need an epistemiccommunity's causal beliefsand notionsof validitybe based on the methodologyemployed in the naturalsciences; they can originate fromshared knowledge about the natureof social or otherprocesses, based on analyticmethods or techniques deemed appropriate to the disciplines or professionsthey pursue. In this volume of articles,for example, while the communityinvolved in effortsto protectthe ozone layer claimed authority

35. See LauristonR. Kingand PhilipH. Melanson,"Knowledge and Politics:Some Experiences fromthe 1960s,"Public Policy 20 (Winter1972), p. 84. For similarobservations, see MartinL. Perl, "The ScientificAdvisory System: Some Observations,"Science 173 (September1971), pp. 1211-15. 36. E. E. Schattschneider,The Semisovereign People (Hinsdale, Ill.: DrydenPress, 1975), p. 66. 37. See the followingworks by HerbertA. Simon: Reason in Human Affairs(Stanford, Calif.: StanfordUniversity Press, 1983); and "Human Naturein Politics:The Dialogue of Psychologywith PoliticalScience," TheAmericanPolitical Science Review 79 (June1985), pp. 293-304. Introduction 17 based on knowledgeabout atmosphericscience, communities involved in other effortshad expertiserelated to disciplinesand professionssuch as economics and engineering. Whilenational epistemic communities may emerge and directtheir activities largelytoward a singlecountry, as in the case of the U.S. communityand the Soviet communitydescribed by Adler, they may in some cases become transnationalover time as a resultof the diffusionof communityideas through conferences,journals, research collaboration, and a variety of informal communicationsand contacts.But epistemiccommunities need notbe transna- tional,nor need theirmembers meet regularlyin a formalmanner. Collabora- tion in the absence of materialinterests binding together actors in different countrieswith common policy agendas would stronglysuggest the existenceof an epistemic communitywith transnationalmembership. A transnational community'sideas maytake rootin an internationalorganization or in various state bodies, afterwhich they are diffusedto other states via the decision makerswho have been influencedby the ideas. As a result,the communitycan have a systemicimpact. Because of itslarger diffusion network, a transnational community'sinfluence is likelyto be much more sustainedand intensethan thatof a nationalcommunity. The epistemic communitymembers' professionaltraining, prestige, and reputationfor expertisein an area highlyvalued by societyor elite decision makersaccord themaccess to the politicalsystem and legitimizeor authorize their activities.Similarly, their claims to knowledge,supported by tests of validity,accord theminfluence over policydebates and serve as theirprimary social power resource.38At the same time, the professionalpedigrees and validitytests set the communitymembers apart fromother social actors or groups39and not onlyserve as a barrierto theirentry into the communitybut also limitthe influencethat these other actors or groups mighthave in the

38. See WolfgangSchluchter, "Modes of Authorityand DemocraticControl," in Volker Meja, Dieter Misgeld,and Nico Stehr,eds., Modem GermanSociology (New York: Columbia University Press,1987), p. 297. "It seems thatin the case of functionalauthority," writes Schluchter, "it is the 'trust'institutionalized in the internalrelations between 'experts' that communicates to outsiders faithin thevalue of specializedknowledge." 39. Accordingto the definitionof epistemiccommunities employed in thisvolume, community membershave intersubjective,internally defined validity tests. This contrastswith Ernst Haas's usage of the concept of epistemic communities,in which he explicitlymentions that such communities"profess beliefs in extracommunityreality tests." See ErnstB. Haas, WhenKnowledge Is Power (Berkeley: Universityof California Press, 1990), p. 41. Although there are other differencesbetween his and our usage,they are fairlyminor. I believethat this particular difference in emphasison intracommunityversus extracommunity truth tests springs primarily from differing overarchinghistorical visions. Ernst Haas seeks to demonstratethe evolutionof rationalityover time,possibly through the gradual intercessionof epistemiccommunities into collectivedecision making.For such a normativeclaim to be sustained,the epistemiccommunity must share a commonbasis forvalidation of itsunderstanding with the broader policy community. Conversely, I am muchmore skepticalabout such universalvalidity claims and am contentto settlefor the less ambitiousinternal truth tests. While in mostcases membersoutside the epistemiccommunity may concurthat validity claims exist, it is less clear thatthey would be able to identifyor evaluatethem. 18 InternationalOrganization

Causal beliefs

Shared Unshared

a ; Epistemiccommunities Interestgroupsand social movements

bureaucratic .= z Disciplinesanda Legislators, v Disciplins agencies,and bureaucratic professions coalitions

Knowledgebase

Consensual Disputedor absent

Interestgroups, social E Epistemiccommunities movements,and bureaucraticcoalitions

i

FIGURE 1. Distinguishingepistemic communities from other groups policydebate. In response to new informationgenerated in theirdomain of expertise,epistemic community members may still engage in internaland often intensedebates leading to a refinementof theirideas and the generationof a new consensusabout the knowledgebase. As Figure 1 indicates,it is the combinationof havinga shared set of causal and principled(analytic and normative)beliefs, a consensualknowledge base, and a commonpolicy enterprise (common interests) that distinguishes epistemic communitiesfrom various other groups. They differfrom interest groups in that the epistemic communitymembers have shared causal beliefs and cause-and-effectunderstandings. If confrontedwith anomalies that under- minedtheir causal beliefs,they would withdraw from the policydebate, unlike interestgroups. Peterson's case regardingthe managementof whaling,for example,stresses the differencebetween the epistemiccommunity of cetolo- gists, the economic interestgroup of whaling industrymanagers, and the issue-orientedlobbying coalition of environmentalists. Introduction 19

Epistemiccommunities must also be distinguishedfrom the broaderscien- tific communityas well as from professionsand disciplines.40Although members of a given professionor discipline may share a set of causal approaches or orientationsand have a consensual knowledgebase, theylack the shared normativecommitments of membersof an epistemiccommunity. An epistemiccommunity's ethical standards arise fromits principled approach to theissue at hand,rather than from a professionalcode. Unlikemembers of a professionor discipline,who seldom limitthemselves to workthat is closely congruentwith their principled values,41 members of an epistemiccommunity tend to pursue activitiesthat closely reflect the community'sprincipled beliefs and tendto affiliateand identifythemselves with groups that likewise reflect or seek to promotethese beliefs. In practice,however, short-term alliances based on commonresearch and concernsoften exist between members of epistemic communitiesand professions.42 The pointto be stressedhere is thatwhile economists as a whole constitutea profession,members of a particularsubgroup of economists-for example, Keynesiansor followersof one of the schoolsof developmenteconomics-may constitutean epistemiccommunity of theirown and systematicallycontribute to a concreteset of projectsinformed by theirpreferred views, beliefs, and ideas. The beliefs and goals of epistemic communitiesdiffer from those of bureaucraticbodies, but the approaches to analyzingepistemic communities and bureaucraticpolitics share a focus on administrativeempowerment of specializedknowledge groups. Bureaucratic bodies operate largelyto preserve theirmissions and budgets,43whereas epistemic communities apply their causal knowledge to a policy enterprise subject to their normativeobjectives. Consequently,although members of epistemic communitiesmay use the bureaucraticleverage they are able to acquire throughobtaining key personnel

40. Accordingto A. M. Carr-Saunders,"What we nowcall a professionemerges when a number of persons are foundto be practicinga definitetechnique foundedupon specialized training." Carr-Saunders'sclassic formulationis cited by Howard M. Vollmer and Donald L. Mills in Professionalization(Englewood Cliffs,N.J.: Prentice-Hall, 1966), p. 3. Subsequentsociologists have formulateda fullerdefinition that includes a reputationfor authority, society's sanction, barriers to entry,a regulativecode of conduct,and a serviceorientation. See Harold L. Wilensky,"The Professionalizationof Everyone,"American Joumal of Sociology70 (September1964), pp. 137-58; and Schluchter,"Modes of Authorityand DemocraticControl." 41. See Charles Derber, WilliamA. Schwartz,and Yale Magrass,Power in theHighest Degree (New York: OxfordUniversity Press, 1990), p. 136. 42. This occurred in the contextof effortsto controlpollution in the Mediterranean,when several groupsof naturalscientists allied withthe ecological epistemiccommunity. While these scientistsshared some of the causal beliefsand policyconcerns of the epistemiccommunity, they did not share its full arrayof normativeand causal beliefs. See Peter M. Haas, "Do Regimes Matter? EpistemicCommunities and MediterraneanPollution Control," Intemational Organiza- tion43 (Summer1989), pp. 386-87. 43. See GrahamT. Allison,Essence of Decision (Boston: Little,Brown, 1971); and RobertJ. Art, "BureaucraticPolitics and American ,"Policy Sciences 4 (December 1973), pp. 467-90. 20 InternationalOrganization slots within bureaucracies, their behavior is differentfrom that of the individualstypically analyzed in termsof theirbureaucratic constraints. Such a normativecomponent means that epistemic communitymembers are not merelypolicy entrepreneurs. Because the behaviorwithin and by an epistemiccommunity is guided by various kinds of normativeand causal beliefsas well as circumstance,it will differfrom the behaviortypically analyzed and predictedby rationalchoice theoristsand principal-agenttheorists. The combinationof shared causal beliefs and shared principledbeliefs held by epistemiccommunity members would informthe advice theyoffer and would offsetor outweighthe pressures for them to offer alternativeadvice which is more consistentwith the preexistingpolitical interestsor preferencesof high-levelpolicymakers or whichmight further their individual careers.' SociologistJoseph Ben-David, writingabout scientificcommunities with tightly shared beliefs, notes thatthey provide "an example of an extreme case of effectivesocial control by a minimumof informalsanctions" and demonstrate"one of the interesting instanceswhere a groupof people is held togetherby a commonpurpose and shared normswithout the need of reinforcementby familial,ecological, or politicalties."45 The solidarityof epistemiccommunity members derives not onlyfrom their shared interests,which are based on cosmopolitan beliefs of promoting collectivebetterment, but also fromtheir shared aversions, which are based on their reluctance to deal with policy agendas outside their common policy enterpriseor invokepolicies based on explanationsthat theydo not accept. The members' institutionalties, informalnetworks, and collectivepolitical practicesalso contributeto the persistenceand solidarityof the communityin severalways. They providemembers with a valuable institutionalstructure in whichto compare informationand to findmoral supportfor their sometimes sociallyand politicallymarginalized beliefs. They also strengthenthe commit- mentsof individualsand inhibitthem from subsequently recanting the beliefs sharedwith and reinforcedby their fellow community members.

Cognateliterature Numerous bodies of literatureshed light on the three major dynamics- uncertainty,interpretation, and institutionalization-thatare exploredin the epistemiccommunities approach to the studyof internationalpolicy coordina- tionpresented here. Insightsgained fromwork in variousdisciplines appear to supportour argumentsthat epistemiccommunities are not epiphenomenal;

44. See Michael Hechter,Principles of Group Solidarity (Berkeley: University of CaliforniaPress, 1987); and Mancur Olson, The Logic of CollectiveAction (Cambridge, Mass.: HarvardUniversity Press,1965). 45. Joseph Ben-David, The Scientist'sRole in Society(Chicago: Universityof Chicago Press, 1984),pp. 5-6. Introduction 21 thatpolicy is not merelydetermined by a consistentset of deeper economic, political,or social structuresthat in some waygenerate a preconditionedset of outcomes;and thatwhile some politicaland social conditionssurely penetrate all technicaladvice and theoutlooks of specialists,all specialistsare notsubject to the same set of conditioningforces. While internationalrelations scholars have introducedmany variables and concepts to help us understandpolicy outcomesand coordination(see Table 2), we argue thatepistemic communi- ties,as objectsof study, are distinctfrom these concepts in thatthey may convey new patternsof reasoningto decision makersand encouragethem to pursue new paths of policymaking,which may in turn lead to unpredicted or unpredictableoutcomes.

Reality is socially constructed. Decision makers are most likelyto turn to epistemiccommunities under conditionsof uncertainty.While theirgoal is ostensiblyto obtain"knowledge" thatwill ameliorate the uncertainty and givethem some handleon the "reality" or "truth"of the situationat hand,the specialistscalled upon foradvice bring withthem their interpretations of the knowledge,which are in turnbased on theircausally informed vision of realityand theirnotions of validity. As numeroussocial and culturaltheorists have argued, realityis socially constructed.Epistemologically, the worldand our representationof it are not isomorphic;our conceptof realityis mediatedby priorassumptions, expecta- tions,and experience.46Even knowledge"cannot mean the 'grasping'of reality itself,"Burkhart Holzner and JohnMarx argue. "In fact,"they add, "philosoph- ical progresshas produced the conclusiveinsight that there can be no such thingas the direct and 'true' apprehensionof 'reality'itself. More strictly speaking, we are compelled to define 'knowledge' as the communicable mapping of some aspect of experiencedreality by an observerin symbolic terms."'' In a similarvein, philosophers and sociologistsof sciencehave pointedto the epistemologicaldifficulty of verifyingour collective visions of the world. Radical constructivists,for example, contend that since the verylanguage we use to describethe world is sociallyconstructed, there is no "objective"basis for identifyingmaterial realityand all claims for objectivityare therefore suspect.48In other words, subject and object are mutuallyconstitutive; no

46. See Peter L. Bergerand Thomas Luckman,The Social Constructionof Reality (Garden City, N.Y.: Doubleday Anchor, 1967). See also Stephen Toulmin, ed., PhysicalReality (New York: Harper & Row, 1970). Toulmin's book includes a turn-of-the-centuryexchange between Max Planckand ErnstMach regardingwhether quantum mechanics occurs in the mindor in the world. 47. Holzner and Marx,Knowledge Application, p. 93. 48. See Steve Woolgar, Science: The VeryIdea (London: Tavistock, 1988); and Karin D. Knorr-Cetinaand Michael Mulkay,eds., Science Observed:Perspectives on the Social Studyof Science (London: Sage, 1983). For a balanced presentationof the radical and more moderate constructivistviews, see M. Hollis and S. Lukes,eds., Rationality and Relativism(Cambridge, Mass.: MIT Press,1982). U 4--4

03 C13

44

Ct 4-4

4-1 CA 42 E. Ct a)> 0 a) ct CZ ct ct CZ

Ct 4) $. 4-- 4-4 CA 0 4.1 C) cn c:) ct ct cn C) Ct

0 C) 0 "4 0 M 0-4

E4 x x x

ZS

Ln "A - (Z3 C'3 = CZ C'n .0 ct ct 64 > 4-1

Cl

u Ct 4- E = 0 C)

CI 4-1 C) C) os to C* u Introduction 23 descriptioncan exist independentlyof the social circumstancesunder which that descriptionis made. Science, theyargue, is no differentfrom any other formof knowledgecreation, and there is no basis forprivileging "scientific" knowledge. Alternatively,those witha more essentialistor materialistview argue that the worldis a real and separate object of inquirythat exists independently of the analystand thatalthough the categoriesin whichit is identifiedare socially constructed,consensus about thenature of the world is possiblein thelong run. This limitedconstructivist view informsthe analysespresented by mostof the authorscontributing to thisvolume. It also has implicationsfor evaluating the validityof a given body of knowledge,pointing to the need for a consensus theoryof a finiteand temporallybounded notion of truth,rather than a correspondencetheory.49 Although knowledge is only accepted belief, not correctbelief, correctbeliefs may evolve over time, as progressivelymore accurate characterizationsof the world are consensuallyformulated.50 By referenceto internallyformulated truth tests, contending groups may collec- tivelyvalidate their conclusions and theirbeliefs may converge intersubjectively in the mediumrun. The epistemiccommunities approach focuses on thisprocess through which consensusis reachedwithin a givendomain of expertiseand throughwhich the consensual knowledgeis diffusedto and carriedforward by other actors.Its primaryconcern is the political influencethat an epistemiccommunity can have on collectivepolicymaking, rather the correctnessof the advice given. While epistemic communitiesprovide consensual knowledge,they do not necessarilygenerate truth.The epistemologicalimpossibility of confirming access to reality means that the group responsible for articulatingthe dimensionsof realityhas greatsocial and politicalinfluence. It can identifyand representwhat is of public concern,particularly in cases in whichthe physical manifestationsof a problemare themselvesunclear, such as the case involving threatsto the stratosphericozone layerexplored in thisvolume of articles. Pursuingontological lines of inquiry,some scholarsin the fieldsof critical theoryand thesociology of science have takena moreinstrumental approach to analyzinghow the nature of realityis elucidated by groups. Rather than addressingquestions about contendingindividual access to reality,they have

49. See Joseph Rouse, Knowledgeand Power (Ithaca, N.Y.: Cornell UniversityPress, 1987); Richard Rorty,Philosophy and theMirror of Nature(Princeton, N.J.: PrincetonUniversity Press, 1979); and Peter Munz, Our Knowledgeof theGrowth of Knowledge (London: Routledge & Kegan Paul, 1985). 50. See Donald T. Campbell,"Evolutionary Epistemology," in P. A. Schilpp,ed., ThePhilosophy of Karl Popper(LaSalle, Ill.: Open Court Publishing,1974), pp. 413-63; Donald T. Campbell and Bonnie T. Paller, "ExtendingEvolutionary Epistemology to 'Justifying'Scientific Beliefs," in Kai Halweg and C. A. Hooker, eds., Issues in EvolutionaryEpistemology (Albany: State Universityof New York Press, 1989), pp. 231-57; Stephen Toulmin,Human Understanding.The CollectiveUse and Evolutionof Concepts(Princeton, N.J.: PrincetonUniversity Press, 1972); Larry Laudan, Progressand Its Problems(Berkeley: Universityof CaliforniaPress, 1977); and LarryLaudan, Scienceand Values(Berkeley: University of CaliforniaPress, 1984). 24 InternationalOrganization investigatedthe effectsof specificsocial institutionson specialists'research. Specialists' choices of research programs,they argue, are influencedby a varietyof factors,ranging from the predispositionsof fundingagencies and the possibilitiesfor career advancementto the deeper social forcesthat identify fruitfulresearch questions and facilitatetheir investigation while impeding the investigationof others.51Moreover, scientists' observations are themselvesthe functionof prior interestsand are influencedby factorssuch as language usage.52This is emphasized by Joseph Rouse, who argues that "scientific observationof theworld is theoreticallyselected and interpretedand functions onlywithin a networkof presupposedtheories. Observation is veryfar from giving us an independent check on the accuracy of our theoretical representations."53Technical advice that fixesattention on specificproblems thusreflects more fundamental social and economicfunctional needs. Many authorshave been quick to pointout thatthe increasinginfluence of specialized groupssuch as epistemiccommunities may have serious negative implicationsfor such deep-seatedpolitical values as democracyand participa- tion. The transferof decision-makingauthority to a groupof elite specialists, theyargue, can furtherlimit access to power by the public.54The Frankfurt school of scholarsand manycritics of modern society and technology,including Ivan Illich,Jacques Ellul, and ,saw the turnto elite specialists as the firstwave of the victoryof instrumentalreason over fundamental interests.Others have warned that privilegingthe advice of specialistsin a particulardomain, such as engineering,may result in the generationof "bad" decisions,either because it leads to a neglectof potentiallyvaluable interdisci- plinaryinsights or ignoresthe social ends to whichdecisions regarding specific issues are directed.55

51. See SylviaNoble Tesh, HiddenArguments: Political Ideology and Disease PreventionPolicy (New Brunswick,N.J.: RutgersUniversity Press, 1988); and JurgenHabermas, Knowledgeand Human Interests(Boston: Beacon Press, 1971). Mannheim's approach was a precursorof this approach,although his was moredirected toward social sciencesthan toward natural sciences. See Karl Mannheim,Ideology and Utopia(New York: Harcourt,Brace & World,1936). 52. See Hannah Pitkin,Wittgenstein and Justice(Berkeley: University of CaliforniaPress, 1972); David Laitin,Politics, Language, and Thought(Chicago: Universityof Chicago Press, 1977); and Benjamin Whorf,Language, Thought,and Reality(Cambridge, Mass.: MIT Press, 1956). Pitkin, followingWittgenstein, argues that language is a sociallycreated artifact.Others argue forthe stronginfluence of externalreality on language, based on evidence of the correlationbetween languages withrespect to particularperceptual concepts. See Steven Lukes, "Relativismin Its Place," in Hollis and Lukes, Rationalityand Relativism,pp. 261-305; Michael Cole and Sylvia Scribner,Culture and Thought(New York: Wiley, 1974); and Marshal H. Segall, Donald T. Campbell,and MelvilleJ. Herskovits,The Influenceof Cultureon VisualPerception (Indianapolis: Bobbs-Merrill,1966). 53. See Rouse, Knowledgeand Power, pp. 3-4. Rouse also notes that "even the simplest concepts,such as 'yellow'or 'ball,' have been said to involvefar-reaching theoretic assumptions." 54. See David Dickson,"Limiting Democracy: Technocrats and the Liberal State,"Democracy 1 (January1981), pp. 61-79; FrankN. Laird,"Limiting Democracy: Participation, Competence, and EnergyPolicy," Ph.D. diss.,Massachusetts Institute of Technology,Cambridge, 1985; and Richard Evan Sclove, Technologyand Freedom(Chicago: Universityof Chicago Press,forthcoming). 55. See Duncan Macrae, Jr.,"Technical Communitiesand PoliticalChoice," Minerva 14 (Spring 1976),pp. 169-90; David F. Noble,America by Design: Science, Technology, and theRise of Corporate Introduction 25

These approacheslead us to ask whatunspoken societal assumptions experts transmit.Is therea dominantsocial culturethat influences the ideas developed and disseminatedby scholars?Potentially, even Karl Popper's argumentthat theoriescan be evaluated accordingto theirinternal coherence ratherthan theircorrespondence to empiricalreality must rest upon some social set of conditions which dictates the value of logic-deductivethought and the impossibilityof "a" and "not a" being true simultaneously.56That new conditionsbring about shiftsin our own beliefswould appear to be supported bythe fact that what Popper refersto as the "thirdworld" of past lore,the body of worksstored in libraries,is continuallyreinterpreted and evokes different responsesin subsequentgenerations of scholars. Additionalquestions remain. First, to whatextent are specialists'theoretical edificessocially conditioned?And, second, does this conditioningreflect a systemicbias? That is, can it be foundin all technicaladvice, or is it merely another social factorthat must be considered in specificcircumstances? If thereis a systemicbias, thenthe analysisof beliefsand perceptionsomits some of the other fundamentalforces in internationalpolitics and focuses on epiphenomena.Specialists will not have a salutaryeffect on policycoordina- tion,since theyonly mask deeper-seatedeconomic forces. Yet thisargument goes too far. Few would denythat technical advice reflects some priorsocial conditioning. I would not go so faras to argue,however, that all technicaladvice sharesthe same conditioning,that such conditioningis irrevocableover the mediumto longterm, or thatall disciplinesare equallyburdened. Although the specialists' claimto privilegedknowledge is certainlysuspect, it is notirrevocably flawed. If the consensus theoryof truthis valid, then the fundamentaldistortions to collectiveunderstanding may over time be reducedor at least be detectableby informedstudy. It is by no means clear thatthe same sets of constraintsand censors operate in everyinstance of specialization and interpretation.For instance,although Jurgen Habermas argued that social needs dictate that centralizedand instrumentalbodies of knowledge are deployed for policy purposes,57the growingapplication of ecologicallyinformed views is much more integrativeand open in its orientation.Moreover,

Capitalism(Oxford: OxfordUniversity Press, 1977); and Nelkin, "ScientificKnowledge, Public Policy,and Democracy."See also AlvinW. Gouldner,The Future of Intellectuals and theRise ofthe New Class (New York: Seabury Press, 1979), in which Gouldner argues that the social consequencesof conferringsteering authority on particulargroups remain unclear. 56. See Karl Popper, ObjectiveKnowledge (Oxford: OxfordUniversity Press, 1972). See also Imre Lakatos, "Falsificationand the Methodologyof ScientificResearch Programmes,"in Imre Lakatos and Alan Musgrove,eds., Criticismand theGrowth of Knowledge (Cambridge: Cambridge UniversityPress, 1970), pp. 91-196. Lakatos offersa sophisticatedextension of thisargument, with normativesuggestions for how to do scientificresearch and positivesuggestions for how to evaluate truth claims from contending research programs. For an alternativeviewpoint, see Paul Feyeraband,Against Method (New York: Schocken,1978). 57. See Habermas,Knowledge and Human Interests. 26 InternationalOrganization ultimatelyfailed to demonstratea consistentsource of social influencesthat operatedon the developmentof discipliningbeliefs and practices.58To take an examplepresented in thisvolume of articles,corporate atmospheric scientists stuckto theirscientific beliefs, counter to the immediateeconomic needs of theircompany.59 That therehave been pathologicalinstances of thewidespread joining of political norms and scientificmeans is reflectedin Nazi medical science and in Lysenko'sevolutionary studies in the Soviet Union. But social pressuresin thesecases servedas a formof controland led to theirhalt.

Ideas informpolicies. Prevailingideas may be an importantdeterminant of policy choice and persistence.For instance,under the swayof economicliberalism, open trade policies emerged and remained prevalentin the nineteenthand twentieth centuriesdespite strong pressures toward protectionism. And in thecase of the repeal of the Corn Laws, the popularityof economic liberalism helped overcomeor preemptthe politicalresistance of agriculturalinterest groups.60 Pointingto thepersistent power of economicorthodoxy, John Maynard Keynes observedthat "practical men, who believe themselvesto be quite exemptfrom any intellectualinfluences, are usuallythe slave of some defuncteconomist. Madmen in authoritywho hear voices in the air are distillingtheir frenzy from some academic scribblerof years back."'61 JohnRuggie offerssimilar arguments with respect to the power of broader visionsof reality, or ,that provide the assumptions from which policies followand shape the patternof politicsover the long run.He also introduces the term"epistemic communities" in keepingwith this package of dominant worldviews: Institutionalizationinvolves not onlythe institutionalgrid of the stateand the internationalpolitical order, through which behavior is acted out,but also theepistemes through which political relationships are visualized.I have borrowedthis term from Michel Foucault,to referto a dominantway

58. Discipliningbeliefs and practicesare discussedby Michel Foucault in the followingworks, forexample: Discipline and Punish:The Birthof thePrison (New York: VintageBooks, 1979); The Historyof Sexuality(New York: Pantheon Books, 1978); and The Birthof the Clinic (New York: PantheonBooks, 1973). 59. Two other examples are noteworthy.Although U.S. Surgeon General C. Everett Koop opposed abortionon personalgrounds, he did notsuccumb to politicalpressures to publiclyoppose it on medical grounds.A Catholic priest trainedin carbon datingtechniques offered evidence againstthe claimsthat the shroudfound in Turinwas Christ'sshroud. 60. See Charles Kindleberger,"The Rise of Free Trade in Western Europe," Journalof EconomicHistory 35 (March 1975), pp. 20-55. See also JudithGoldstein, "Ideas, Institutions,and Trade Policy,"International Organization 42 (Winter1988), pp. 179-217; JudithGoldstein, "The Impact of Ideas on Trade Policy,"International Organization 43 (Winter 1989), pp. 31-71; and Odell, U.S. InternationalMonetary Policy. 61. JohnMaynard Keynes, The GeneralTheory of Employment, Interest and Money(New York: Harcourt,Brace & World,1936), p. 383. Introduction 27

of lookingat social reality,a set of sharedsymbols and references,mutual expectationsand a mutualpredictability of intention.Epistemic communi- ties maybe said to consistof interrelatedroles which grow up aroundan ;they delimit, for their members, the proper construction of social reality.62 Our usage of the term "epistemic community"is at a lower level of abstractionthan Ruggie's usage. We use the term to referto a concrete collectionof individualswho share the same worldview(or episteme)and in particularshare the four aspects of it that were outlined earlier. While membersof an epistemiccommunity by definitionshare an epistemewith each other,they do not necessarilyshare it with other groups or individuals.In practice,the numberof membersin the communitieswe describeis relatively small. It is the politicalinfiltration of an epistemiccommunity into governing institutionswhich lays the groundworkfor a broader acceptance of the community'sbeliefs and ideas about the properconstruction of social reality. The resultin turnmay be the creationof the proper construction of realitywith respectto a particularissue-area as well as mutualexpectations and a mutual predictabilityof intention.The intentof the articlesin thisvolume is to analyze thisprocess in numerousconcrete cases and discernthe extentto whichthe substantivecontent of policieswas shaped bycommunity views and the extent to whichother actors and politicalforces played a role. While the notionthat ideas informpolicies is provocative,it leaves a number of questions unanswered.Are ideas themselvessocially conditioned,or do social conditionsmerely affect which ideas gain acceptance? How are ideas disseminated?Why do some prevailover others? What is thelife cycle of ideas? How do theyevolve? Even ifscholars resort to a naturalselection model for the evolutionof ideas, as Donald Campbell does,63then theymust identifythe mechanismsor agents of selection.A fruitfulapplication of ideas to policy choice,at least overtime, requires a greaterspecification of how ideas emerge and change(or evolveand are selectedfor). Without compelling answers to the questionsthat remain in thisregard, it is difficultto supportthe argumentthat ideas are independentvariables and notjust interveningvariables. The view presented in this volume is that epistemic communitiesare channelsthrough which new ideas circulatefrom societies to governmentsas well as fromcountry to country.However, an epistemiccommunity cannot be reduced to the ideas it embodiesor purveys,since these ideas are transmitted in tandemwith a set of causal and principledbeliefs and reflecta particular politicalvision. The ideas would be sterilewithout carriers, who functionmore or less as cognitivebaggage handlers as well as gatekeepersgoverning the entry of new ideas into institutions.The influencethat an epistemiccommunity has

62. JohnGerard Ruggie,"International Responses to Technology,"International Organization 29 (Summer1975), pp. 569-70. 63. See Campbell,"Evolutionary Epistemology." 28 InternationalOrganization and the ideas thatit transmitsmay well be mutuallyreinforcing. In the articles presentedin thisvolume, the precise dynamics by which epistemic communities generatednew ideas and chose between alternativesin particularcases are presented, and the various social and political conditionsbearing on the developmentand disseminationof theirideas and viewsare discussedin detail.

Actors' understandingof the world and theformulation of alternativeactions are shaped by belief systems,operational codes, and cognitivemaps. While we can draw on epistemologicalarguments about perceptionsof and access to reality,we can also draw on the insightsof cognitivepsychologists, who stressthe conditioningrole that priorbeliefs and establishedoperating proceduresplay in determininghow individualswill respond to new situations or events and choose a course of action when confrontedwith uncertainty. Faced witha new situation,we identifyand interpretproblems within existing frameworksand according to past protocols and then tryto manage the problemsaccording to operatingprocedures that we have applied in analogous cases. Aspectsof the situationthat cannot be dealt within establishedways are only incompletelyperceived and processed, with the result that salient dimensionsof a problemor issue at hand are oftenignored.'M Investigatingcrises during which decisions affecting survival had to be made quicklyby heads of state and high-leveladvisers, scholars have found that informationprocessing was at best incrementaland that decision makers tended to apply simplifiedimages of realitywhich were highlyresistant to modification.65Examining noncrisis cases as well, other analystshave noted thatdecision makers are not alwaysaware of the possibleimpact of the signals theysend, since theytend to presumethat the receiversof these signalshave a worldviewwhich mirrors their own.66 Similarly, decision makers' understanding of others'behavior is shaped bytheir own beliefs,motives, and intentions,and thissometimes leads themto misinterpretthe signalsthey receive from others. Even the internationalcontext in whichproblems are to be resolved is not equally transparentto all actors.67Factors such as these can contributeto the breakdownof cooperation.

64. See Janice Gross Stein, "InternationalNegotiation: A MultidisciplinaryPerspective," NegotiationJournal, July 1988, pp. 221-31; and Deborah Welch Larson, "The Psychologyof Reciprocityin InternationalRelations," Negotiation Journal, July 1988, pp. 281-301. 65. See Ole Holsti,"Crisis Decision Making,"in PhilipE. Tetlocket al., eds., Behavior,Society, and NuclearWar, vol. 1 (New York: OxfordUniversity Press, 1989), pp. 8-84; JeanLave, Cognition in Practice(Cambridge: Cambridge University Press, 1988); and R. Nisbettand L. Ross, Human Inference(Englewood Cliffs,N.J.: Prentice-Hall,1980). See also ErvingGoffman, Frame Analysis (Cambridge,Mass.: HarvardUniversity Press, 1974). 66. See Jervis,Perception and MisperceptioninInternational Politics; and RobertJervis, "Realism, Game Theory,and Cooperation,"World Politics 40 (April 1988), pp. 317-49. 67. See Snyderand Diesing,ConflictAmong Nations. Introduction 29

As Ole Holsti has argued, belief systemsimpose "cognitiverestraints on rationality."68More broadly,the combinationof priorbelief systems,opera- tionalcodes, and cognitivemaps shapes decisionmakers' responses not only by influencingthe ways in which theyinterpret the world but also by erecting barriersto the typesof informationthat they consider valuable.69 Yet policyresponses to uncertaintycannot be reduced to cognitivepsychol- ogy.Belief systemsmay be conferredby epistemiccommunities. As arguedin an earliersection, whether decision makers turn to epistemiccommunities for advice depends on the level of theiruncertainty about an issue-area. Failed policies,crises, and unanticipatedevents that call intoquestion their understand- ing of an issue-area are likelyto precipitatesearches fornew information,as are the increasingcomplexity and technicalnature of problems.If decision makershave no strongpreconceived views and beliefsabout an issue-area in whichregulation is to be undertakenfor the first time, an epistemiccommunity can have an even greaterimpact in shapingtheir interpretations and actionsin thiscase and in establishingthe patternsof behaviorthat theywill followin subsequentcases regardingthe issue-area.

Consensual knowledge may contributeto policy coordination and to more comprehensivepolicies. Before states can agree on whetherand how to deal collectivelywith a specificproblem, they must reach some consensusabout the natureand scope ofthe problem and also about the mannerin whichthe problem relates to other concernsin the same and additionalissue-areas. Ernst B. Haas has arguedthat "interrelatednessmay also become interdependencein the sense that new scientificknowledge will create a consensualbasis forthe recognitionof new cause-effectlinks which had not been recognizedbefore."70 As the scientific

68. See Ole Holsti," 'The OperationalCode' as an Approachto theAnalysis of Belief Systems," finalreport to the National Science Foundation,grant no. SOC75-15368,December 1977,p. 2. 69. See Kenneth Boulding, The Image (Ann Arbor: Universityof Michigan Press, 1956); Kenneth Boulding, Conflictand Defense (New York: Harper, 1962), chap. 14; Alexander L. George, "The Causal Nexus Between CognitiveBeliefs and Decision-MakingBehavior: The 'Operational Code' Belief System," in Lawrence S. Falkowski, ed., PsychologicalModels in InternationalPolitics (Boulder, Colo.: WestviewPress, 1979), pp. 95-124; and RobertAxelrod, ed., Structureof Decision: The CognitiveMaps of PoliticalElites (Princeton,N.J.: PrincetonUniversity Press, 1976). For discussionsabout artificialintelligence modeling of thoughtpatterns, see Dwain Mefford,"Analogical Reasoningand the Definitionof the Situation:Back to Snyderfor Concepts and Forwardto ArtificialIntelligence for Method," in CharlesF. Hermann,Charles W. Kegley,Jr., and JamesN. Rosenau, eds.,New Directionsin theStudy of Foreign Policy (Boston: Allen & Unwin, 1987); and Jaime Carbonell,Subjective Understanding: Computer Models of BeliefSystems (Ann Arbor:University of MichiganPress, 1981). 70. See ErnstB. Haas, "Knowledge,Technology, Interdependence," International Organization 29 (Summer 1975), pp. 858-59. See also Ernst B. Haas, "Why Collaborate? Issue-Linkageand InternationalRegimes," WorldPolitics 32 (April 1980), pp. 357-405; and ErnstB. Haas, MaryPat Williams,and Don Babai, Scientistsand WorldOrder (Berkeley: University of CaliforniaPress, 1977). 30 InternationalOrganization consensus becomes the collectiveconsensus of decision makers and as the natureof theproblem is collectivelyredefined in broaderand moreinterlinked terms,the need formore comprehensivepatterns of policycoordination may also be recognizedand pursued. Whethercollective behavior becomes more comprehensiverather than merely ad hoc and incrementalwill in turndepend on the extentto whichthe scientists'and the decision makers'views coincide and the extent to which the negotiationsreflect the pursuit of politically motivatedlinkages and the strugglefor control among states. In general, governmentsand organizationsmay be said to learn throughthe evolutionof consensualknowledge. While the role of consensualknowledge in policycoordination has been the focus of numerousstudies, the process by whichthe views of specialistsare accepted and acted upon bydecision makers are poorlyspecified. In particular, studies have not addressed the question of how recalcitrantstates can be persuadedto accept new causal understandingsthat point to policieswhich are contraryto their conceptions of self-interest.Moreover, given the many examplesof differentstates reacting in differentways to the same consensual evidence provided by specialists,71it is unclear how effectiveconsensual knowledgeis, as an independentvariable, at explainingor predictingstate behavior.The organizationalstructures through which consensual knowledge is diffusedmay be equally important.As the studiespresented in thisvolume demonstrate,epistemic communities can insinuatetheir views and influence nationalgovernments and internationalorganizations by occupyingniches in advisoryand regulatorybodies. This suggeststhat the applicationof consensual knowledgeto policymakingdepends on the abilityof the groupstransmitting this knowledge to gain and exercise bureaucraticpower. Moreover,while governmentsand organizationsmay learn throughthe evolutionof knowledge, the learningdoes not necessarilylead to policycoordination. In the case of internationalcommodity arrangements, for example, the consensusemerging fromnew knowledgewas that cooperation would not result in joint gains; hence,the effortsat policycoordination collapsed.72 This suggeststhat whether epistemicinfluence leads to policycoordination is a functionof whetherthe causal beliefsof epistemiccommunities demonstrate the need forit.

71. See, forexample, B. Gillespie,D. Eva, and R. Johnston,"Carcinogenic Risk Assessmentin the United States and Great Britain:The Case of Aldrin/Dieldrin,"Social Studiesof Science 9 (August 1979), pp. 265-301; and George Hoberg, Jr., "Risk, Science and Politics: Alachlor Regulationin Canada and theUnited States,"Canadian Journalof Political Science 23 (June1990), pp. 257-77. For a broader discussion of additional factors influencingregulatory agencies' acceptance of consensual knowledge,see Sheila Jasanoff,Risk and PoliticalCulture (New York: Russell Sage, 1986). 72. See Robert L. Rothstein, "Consensual Knowledge and InternationalCollaboration," InternationalOrganization 38 (Autumn1984), pp. 733-62. Introduction 31

Policy choices are oftenmade by discrete networks of actors. Numerousscholars have argued thatdomestic regulation in cases involving complex and highlytechnical issues is often the result of collusion among interestedparties. Decision making,rather than being centralized,occurs withinan amorphousset of subgovernments.Whether the partiesinvolved are characterizedas interestgroups, iron triangles,advocacy coalitions, issue networks,or policynetworks,73 the pointis the same: small networksof policy specialistscongregate to discuss specificissues, set agendas, and formulate policy alternativesoutside the formalbureaucratic channels, and they also serve as brokers for admittingnew ideas into decision-makingcircles of bureaucratsand elected officials.74While muchof the literaturein thisregard focuseson policymakingin the United States,similar technocratic subgovern- mentselsewhere have been discussed.75Unfortunately, however, most of the literaturehas remaineddescriptive rather than analytic. It does not identifyor explorethe commoncausal beliefsthat participants may carry with them, nor does it indicate the degree to which such groups actuallyinfluence policy outcomes. As withthe transgovernmentalliterature discussed below, this literature on domesticnetworks supports several of the argumentspursued in the epistemic communitiesapproach, among them the argumentthat a nonsystemiclevel of analysisis usefulfor considering decisions made in responseto systemicstimuli and the argumentthat networks of specialistscan become strongactors at the national level. Key locationsfrom which members of epistemiccommunities could gain significantleverage over policy choices include think tanks, regulatoryagencies, and the typeof governmentalpolicy research bodies that are morecommon outside the United States.Allied throughtransnational and

73. See Hugh Heclo, "Issue Networksand the ExecutiveEstablishment," in AnthonyKing, ed., The NewAmerican Political System (Washington, D.C.: AmericanEnterprise Institute, 1978), pp. 87-124; JohnW. Kingdon,Agendas, Alternatives, and PublicPolicies (Boston: Little,Brown, 1984); Thomas L. Gais, Mark A. Peterson,and JackL. Walker,"Interest Groups, Iron Triangles,and RepresentativeInstitutions in AmericanNational Government,"British Journal of Political Science 14 (April 1984), pp. 161-85; JackL. Walker,"The Diffusionof Knowledge,Policy Communities, and Agenda Setting:The Relationshipof Knowledgeand Power," in JohnE. Tropman,Milan J. Dluhy,and Roger M. Lind, eds., New StrategicPerspectives on (New York: Pergamon, 1981), pp. 75-96; Paul A. Sabatier,"Knowledge, Policy-Oriented Learning, and PolicyChange," KnowledgeCreation, Diffusion, Utilization 8 (June 1987), pp. 649-92; and John Mark Hansen, "Creatinga New Politics:The Evolutionof Networks in Congress,1919-1980," Ph.D. diss.,Yale University,New Haven, Conn., 1987. 74. See JamesL. Sundquist,"Research Brokerage:The Weak Link," in LawrenceE. Lynn,Jr., ed., Knowledgeand Policy:The UncertainConnection, vol. 5 (Washington,D.C.: NationalAcademy of Science, 1978), pp. 126-44. 75. See Suleiman,Bureaucrats and PolicyMaking; and Merilee S. Grindle,ed., Politicsand Policy Implementationin theThird World (Princeton, N.J.: Princeton University Press, 1980). 32 InternationalOrganization transgovernmentalchannels, the specialistscould have an impacton interna- tionalpolicy coordination.

Coalitions are built transgovernmentally and transnationally. Internationalrelations scholars have also identifiedand pointed to the significanceof transgovernmentaland transnationalchannels throughwhich political alliances are forgedand informationregarding technical issues is transmittedbetween government officials, international secretariats, nongovern- mentalbodies, and nongovernmentalactors, including communities of profes- sional scientists.76Many of these scholarshave argued thatmanagement tasks at the internationallevel have to some extentbeen usurped by groups of functionallyequivalent nonstate actors who act relativelyindependently of the policies of top leaders of theirgovernments. The membersof these groups, when operatingin tandem throughtacit alliances, can concurrentlypromote their ideas and specific policy objectives within their own countries and governments. This approach describesthe coordinatingrole of membersof international secretariatsand of governmentaland nongovernmentalbodies and the chan- nels throughwhich theyinteract, but it is unclear about what outcomes are likelyto occur otherthan the formationof short-termpolicy coalitions among individualswho occupysimilar positions or levelsof responsibilityand interact on a regularbasis. Such channels could be used just as well by higher-level foreignpolicy officials to extendtheir own view of stateinterests. With respect to transgovernmentalalliances, for example, the approachdoes notinvestigate the originof the interestsof the membersinvolved. Do theirinterests stem fromtheir common bureaucratic roles withintheir own governments,or are theybased on preexistingbeliefs and interestswhich they brought to theirjobs and whichare likelyto be pursuedeven afterthey leave theircurrent posts? In the absence of attentionpaid to anycommon causal beliefsand understandings amongthe members,we maywell tend to conclude thatsuch alliances willbe

76. See Robert 0. Keohane and JosephS. Nye,eds., TransnationalRelations and WorldPolitics (Cambridge,Mass.: Harvard UniversityPress, 1971); Robert 0. Keohane and Joseph S. Nye, "TransgovernmentalRelations and InternationalOrganizations," World Politics 27 (October 1974), pp. 39-62; RaymondF. Hopkins,"Global ManagementNetworks: The Internationalization of Domestic Bureaucracies," InternationalSocial Science Journal30 (June 1978), pp. 31-46; WilliamM. Evan, ed., Knowledgeand Powerin a Global Society(Beverly Hills, Calif.: Sage, 1981); and Peter Willets,ed., PressureGroups in theGlobal System(New York: St. Martin'sPress, 1982). For more recent effortsin a similar vein, see ChristerJonsson, "IntegrationTheory and InternationalOrganization," International Studies Quarterly 30 (March 1986), pp. 39-57; Christer Jonssonand StaffanBolin, "IAEA's Role in the InternationalPolitics of Atomic Energy,"in LawrenceS. Finkelstein,ed., Politicsin theUnited Nations System (Durham, N.C.: Duke University Press, 1988), pp. 303-23; and Michael M. Cernea, "NongovernmentalOrganizations and Local Development,"World Bank discussionpaper no. 40, Washington,D.C., 1988. Introduction 33 short-lived.77Indeed, while those pursuing studies of transgovernmental alliances clearlyidentified channels of diplomacyand interstateinteraction whichhad been neglectedpreviously, they were unable to demonstratethat activitiesemanating from these channels had any independentinfluence on outcomes.Moreover, they found that as issues gained in saliency,they came to assume the characteristicsof highpolitics, with the resultthat the transgovern- mentallinkages waned.78 As Diana Crane found in her study of transnationalscientific groups, however,the sharedcausal beliefsof the individualsin these groupsproved to be more importantdeterminants of outcomesthan did the channelsthrough whichthey operated: "These studies show that it is not necessarilyan ISPA [internationalscientific and professionalassociation] which exerts political influencebut the expertcommittee which may or maynot be affiliatedwith an ISPA. The invisiblecollege which cuts across all the organizationsinvolved, both IGOs [internationalgovernmental organizations] and INGOs [interna- tional nongovernmentalorganizations], plays an importantrole in integrating the fragmentedIGO programs."79Crane's findingslend support to our argumentthat epistemiccommunities operating through transnationally ap- plied policynetworks can proveinfluential in policycoordination.

Organizations are not always captured. The recentliterature on the "new institutionalism"brings together many of the argumentsconcerning the process of decisionmaking.80 Those pursuingthe institutionalistapproach emphasize the relativeautonomy of politicalinstitu- tions,which may mediate the pressureson decisionmakers from international structuresand domesticforces. Ultimately, they argue, institutional choices are influencedto a greaterextent by historicallyinherited preferences and styles thanby externalstructural factors. This means thatthe initialidentification of interestsand decision-makingprocedures will have a major influenceon subsequent policy choices, alternativesdeemed possible, and actual state behavior-as was evident,for example, in the choice of the Norwegianoil

77. See Barbara B. Crane, "Policy Coordinationby Major WesternPowers in Bargainingwith the Third World: Debt Relief and the Common Fund," InternationalOrganization 38 (Summer 1984), p. 426. 78. See, for example, Harold K. Jacobson,"WHO: Medicine, Regionalism,and Managed Politics,"in RobertW. Cox and Harold K. Jacobson,eds., TheAnatomy of Influence (New Haven, Conn.: Yale UniversityPress, 1974), p. 214. 79. Diana Crane, "AlternativeModels of ISPAs," p. 39. 80. See the followingworks by James G. March and JohanP. Olsen: "The New Institutionalism: OrganizationalFactors in PoliticalLife," American Political Science Review 78 (September1984), pp. 734-49; and RediscoveringInstitutions (New York: Macmillan,1989). The actual definitionof institutionsin thisliterature appears remarkablyfluid. Institutions may be anythingfrom formal organizationsto social forces(including capitalism) to culture. 34 InternationalOrganization

ministryto applythe same safetyregulations to offshoreoil rigsin the 1970sas it had applied earlierto oil tankers.8" In a similar vein, studies based on organization theory indicate that institutionsdo not continuallymonitor their surroundings and reevaluateprior choices. Indeed, in the absence of crises,there will be littlereconsideration of past choices, subject to what Peter Blau termsthe elasticityof demand for advice.82Yet when prior understandingslead to unexpectedoutcomes and uncertainty,organizations will take on the functionof informationseekers and solicitorsof advice.As JamesWilson notes in thisregard, large institutions are more likelyto generatenew ideas but are less likelyto adopt themthan are smallerinstitutions.83 In conjunction,the bodies of literatureon organizationtheory and the new institutionalismshed lighton the circumstancesunder which groups with new ideas are likelyto emerge,offer an explanationfor their enduring influence, and point to the conclusion that, once in place, a group will persistuntil subsequentcrises challenge its ability to provideadvice.

Methodologyand guidancefor further research

The epistemiccommunities approach pursued in thisvolume is distinctfrom the approachesreviewed above, althoughit rigorouslyintegrates and buildson the insightsgained frommany of them.It distinguishesepistemic communities fromother groups that seek to exert influenceon decision makers,and it specifiesin greaterdetail both the factors that lead knowledge-basedgroups to cohere and the mechanismsby which theygain and retain influencein the policymakingprocess. The researchtechniques for demonstrating the impactof epistemiccommu- nitieson the policymakingprocess are straightforwardbut painstaking.With respectto a specificcommunity, they involve identifying community member- ship, determiningthe communitymembers' principledand causal beliefs, tracingtheir activities and demonstratingtheir influence on decisionmakers at various points in time, identifyingalternative credible outcomes that were foreclosedas a resultof theirinfluence, and exploringalternative explanations forthe actionsof decisionmakers.84 The use of counterfactualsmay be helpful in thisregard.

81. March and Olsen,Rediscovering Institutions. 82. Peter Blau, Exchangeand Powerin Social Life (New York: Wiley,1967), chap. 7. See also James G. March, "Footnotes to OrganizationalChange," AdministrativeScience Quarterly26 (December 1981),pp. 563-77. 83. James Q. Wilson, "Innovation in Organization:Notes Toward a Theory," in James D. Thompson,ed., Approachesto OrganizationalDesign (Pittsburgh:University of PittsburghPress, 1966),pp. 197-218. 84. See AlexanderGeorge, "Case Studyand TheoryDevelopment: The Method of Structured, Focused Comparison,"in Paul Gordon, ed., Diplomacy:New Approachesin History,Theory, and Policy(New York: Free Press, 1979), pp. 43-68; and AlexanderGeorge and TimothyJ. McKeown, "Case Studies and Theories of OrganizationalDecision Making," in Advances in Information Processingin Organizations,vol. 2, 1985,pp. 21-58. Introduction 35

One of the pointsemphasized in the earliersection about distinguishingan epistemiccommunity from another type of knowledge-basedgroup (a disci- pline,a profession,a coalitionof bureaucrats,and so forth)was thatwhile the membersof any knowledge-basedgroup may share criteriaof validityand a policy enterprise,members of an epistemic communityin addition share principled(normative) and causal beliefs.Individuals in thecommunity may be foundamong the respectedexperts whose names recuron delegationlists to intergovernmentalmeetings or among those responsiblefor draftingback- groundreports or briefingdiplomats. Identifying the beliefsof a community calls for a detailed study of materials such as the early publications of communitymembers, testimonies before legislative bodies, speeches, biograph- ical accounts,and interviews.The processof tracingcausal beliefsis obviously easier if members'backgrounds are in disciplinesthat make copious use of equations and models.Operationally, epistemic community members could be distinguishedfrom nonmembers on the basis of whetherthe keyvariables and transformationequations incorporated in theirmodels agree. Their beliefs and spread of networkscould then be depicted by causal mappingand network analysis.85The extentto whichepistemic beliefs mask social conditioningcan be assessed througha judicious use of the secondaryliterature regarding the intellectualhistory of the disciplinesfrom which the epistemiccommunity derivesits understanding of theworld. A robuststudy of an epistemiccommunity's influence calls forcomparative studiesof countriesand organizationsin whichthe communityhas been active and those in whichit has not. Moreover,it calls foran analysisof policies and practicespursued by governmentsand organizationsnot only in the period duringwhich a communityis active but also the periods before and afterin orderto determineboth the emergenceand the persistenceof influence. As the studies in this volume demonstrate,epistemic communities have exertedtheir influence on decision makersin a wide varietyof issue-areas. Generallycalled upon for advice under conditionsof uncertainty,they have oftenproved to be significantactors in shapingpatterns of internationalpolicy coordination.

85. See Axelrod,Structure of Decision.