From: AAAI Technical Report WS-93-03. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved. Artificial Intelligence and Organizational Learning Howcan AI contribute to Organizational Learning? ShigehisaTsuchiya Interdisciplinary Courseon Advanced Scienceand Technology TheUniversity of Tokyo 4-6-1 Komaba,Meguro-ku,Tokyo, 153 Japan e-mail: tsuchiya@ai, rcast.u-tokyo.ac.jp Abstract Informationtechnologyassisted by artificial intelligence can support organizational learning and enable organizations to form and executestrategy effectively. Organizationalstyle is changing. Realizingthe characteristics of social problems, moreand moreorganizations are now loosening couplingsso that their subsystems mayhave morechance to enact environment. As Weick (1976) suggested, loosely coupled organizations permitting considerableflexibility in the behaviorof their subsystemsare better able to adapt and survive. Traditional organization style --hierarchical tightly coupledsystem-- is becomingout of date. In formulatingstrategy, executives are nowbecomingaware of inevitable ambiguity in the real world.Traditional analytical strategy is based on two assumptions whichoften prove false: that the formulator is fully informed,and that the environmentis stable, or at least predictable. Due to the absenceof either condition these days, the importanceof emergentstrategy (Mintzberg, 1978) has becomewidely acknowledgedamongmanagers in the busi16 ness world. Becauseof these changes in organization and strategy, manyorganizations are havingdifficulties in makingstrategic decisions. Throughintensive case studies of a private college, a typical exampleof loose coupling, 1 have developedand verified following new hypotheses: information technologyassisted by artificial intelligence can help organizations form and implementappropriate strategies by improvingorganizational decision makingprocess through organizational learning. Keywords.information technology; artificial intelligence; organizationallearning; emergentstrategy; loosely coupled system; IBMhas drastically restructured its organization from tightly coupled hierarchy to loosely coupled federation. Soviet Union died and Commonwealthof Independent States was created. Thus traditional organizationstyle -- hierarchical tightly coupled system-- is becomingout of date and loosely coupled system seems to be the organizationstyle in the future. Faced with frequent and drastic environmental changes, more and more organizations are now loosening couplings so that I Strategy Formation The term strategy has been defined in a their subsystems may have more chance to variety of ways, but almost always with a enact environment. Sensitivity and flexibility to the environmentare essential for commontheme, that of a deliberate conscious set of guidelines that determines de- their survival. Loosely coupled system seems to be the cision into the future. A typical definition organization style in the future. However, there is a big problem: strategy formation is basic long-term goals and objectives of an enterprise, and the adoption of courses of difficult action and the allocation of resources nec- in a loosely coupled system. Weick of strategy is "...the determination of the (1976) suggests that loosely coupled orga- essary for carrying on these goals nizations permitting considerable flexibility in the behavior of their subsystems are bet- (Chandler, 1962)". All these definitions treat strategy as (a) explicit, (b) developed ter able to adapt and survive. Yet, he ad- consciously and purposefully, mits that, whenthe changes in the environ- in advance of the specific decisions to ment are drastic, the organizations have to tighten up their couplings to survive. which it applies. The basic assumption here is the environment is analyzable. Meanwhile,as for strategy, the impor- and (c) made However,deliberate and analytic strate- tance of emergent strategy (Mintzberg, gies such as PPMmodels, SBU, etc., 1978) has become widely acknowledged. Traditional planning theory postulates that failed to be effective in the "real world", because the environment surrounding orga- the strategy-maker "formulates" from on nizations is unanalyzable. By nature, high while the subordinates humanworld is not rational nor systematic. "implement" have lower down. This dichotomy is based on but ambiguous, equivocal and liquid. two assumptions which often prove false: "real world" is composed of innumerable that the formulator is fully informed, and facts and characteristics, that the environmentis stable, or at least ship amongthose components and causali- predictable. Due to the absence of either condition these days, strategy formation ty are ambiguous. Through empirical studies, has becomea learning process, whereby im- (1978) defined strategy as "a pattern in plementation feeds back to formulation and stream of decisions". Successful organiza- intentions get modified en route, resulting tions in Japan have never adopted analytical strategies. Assumingthe unanalyzable in an emergent strategy. Many organizations muchdifficulties The and the relation- Mintzberg environment, they experiment environment are now having in makingstrategic deci- sions because they are loosely coupled and adopt emergent strategies. by action and form effective strategies inductively. For them, strategy formation is a learning process, where emergent strategies play important roles. Emergentstrategies 17 are formedandimpleUNDIRECTEDVIEWING mentedfrom the middle interprststione. up, while deliberate Unenelizabla Constrained Nonroutine,informal data. Hunch, rumour, chance strategiesare formulated ASSUMPTIONSopportunities. and implementedfrom ABOUT ENVIRONMENT the top down. Loose CONDITIONEDVIEWING couplingis suited better Analizable Interprets within traditional for such "process type boundaries. Passive detection. Routine, formal strategies" than tight data. coupling.. ENACTING Experimentation, testing, coercion, invent environment. Learn by doing. DISCOVERING Formal torch. Questioning, surveys, datagathering.Active detection. Passive Active ORGANIZATIONALINTRUSIVENESS 0Neick8, Daft,1983; Daft & Weick,1984) Fig.1 Modelof Organizational Interpretation Modes 2 Organizational Decision Making Organizational decision makingis governed by the modeof its interpretative framework.Organizations must makeinterpretations. Interpretation is the processof translating events surroundingthem, of developing modelsfor understanding, of bringing out meaning,and of assembling conceptual schemes amongkey members (Daft and Weick, 1984). The modelproposed by Daft and Weickdescribes four interpretation modes:undirected viewing, conditioned viewing, discovering, and enacting (Fig. 1). Eachmodeis determined (I) management’s beliefs about the analyzability of the external environmentand (2) the extent to whichthe organization intrudes into the environmentto understand iL In order to cope with the newturbulent environment by loosening couplings and adopting emergentstrategies, organizations must changethe modeof its interpretative framework to ENACTING. 18 3 Organizational Learning ORGANIZATIONAL LEARNINGis defined as the process by which knowledge about action outcomerelationships between the organization and the environment is developed (Duncan& Weiss, 1979). It is the process of evolutionof interpretation modeof the organization. Organizational interpretative framework is formedthrough interpreting the outcomes of its DECISIONS and ACTIONS. Organizationalinterpretation is defined as the process of translating events and developing shared understanding and conceptual schemesamongmembersof organization. The organization makesdecisions and actions based on its organizational INFERENCE.Andorganizational inference is limited by the modeof its interpretative frameworkbecause the organization will not perceive any information inconsistent with its interpretation mode. Thusthe three stages -- inference, decision and action, and interpretative framework-- are interconnected through a feedback loop in Figure 2. Changingorganizational interpretation modethrough organi- "~’"l zational learning is a very I I difficult processbecause theorganization has tokl cut the loop. According "-~ NewInterpretative Framework Interpretation Interpretative "~"’i Mode Framework interpretation Dialogue Mode Decision~Action Inference i.~ Decision/ through Dialogue I" I Action Mode ] L-,,J / to mycasestudies, the Fig.2 OrganizationalLearning key stage in evolution of interpretation modeis INFERENCE. possible.It includesstorage,reproduction, transportation, andrearrangement. Man’s 4 Inference intellectual superiorityoverthe animalsis Organizations as well as humansmake almostentirely dueto the useof language. But, tacit knowing is still the fundamental tacit and explicit inferencein a simultanepowerof the mind,whichcreatesexplicit ous and integrated wayto cope with probknowing,lends meaning to it andcontrols lemsin the ambiguousreal world. Tacit inference is nonlinguistic and comprehensive its uses. thought,and explicit inferenceis linguistic and analytical thought. 5 Information Technology ad AI Informationtechnologies assisted by arKnowledgeis fundamentally tacit. We knowmorethan we can tell (Polanyi, tificial intelligence can support organiza1966). All knowledge is either tacit or roottional learning and makeinterpretative ed in tacit knowledge,because, while tacit frameworkto evolve by improving organiknowledgecan be possessed by itself, exzational inference. 5.1 Substitution of Human Activities plicit knowledgemustrely on being tacitly Traditional view of the contribution of understood and applied. All explicit informationtechnologyin inference is that knowledge,howevercrystallized in the forinformation technology can substitute a malismsof words,pictures, or other articulate devices, relies on the grasp of meaning part of humanactivities so that we may throughits articulate forms: on the comprespendmoretime in creative activities which is innate in humanbeings. hensionthat is its tacit root. However,formalization of tacit knowing Usinginformationtechnologyis diffiimmenselyexpands the power of the mind, cult. Dueto the characteristics of managerial problems,flexibility is essential for the by creating a machineryof precise thought. system, which makesbuilding an effective It also opensup newpaths to intuition. systemdifficult and expensive. In addition, Whenknowledgeis expressed in language becausethe benefits of the systemare inin a broad meaningthrough articulation, tangible, a high initial investmentis difficult sign manipulation of knowledgebecomes 19 to be justified. However,in spite of those difficulties, several companiesare now using informationtechnologies to let employeeshave moretime for creative work. S.2 Support of inference However,my case studies have madeit clear that information technologyassisted by artificial intelligence can directly support and improve inference. Whenthinking, humansmaketacit and explicit inference simultaneouslyand interactively. It is an integrated process. Although knowledgeis fundamentally tacit, formalizationof tacit knowing immensely expands the power of the mind. Sign manipulation of knowledgeimmensely improves humanthought. Human intellectual superiority to animals comes fromthe ability of articulation and sign manipulation. As long as sign manipulation is concerned, computers surpass humansby far. Human beings are strong at intuitive thinking, but weakat calculation and logical reasoning. Therefore, information technology can support inference through effectively replacingthe explicit side of humanthinking process. Individual tacit knowledge,whichis the fundamentalsource of organizational knowledge,needs to be shared and legitimated before it becomesorganizational knowledge. Transmission of knowledge becomespossible whenknowledgeis expressed in language (in a broad meaning) through sense-giving. Communicationis an interactive process of sense-giving and sense- reading: endowingones ownutterances with meaningand attributing mean2O ing to the utterances of others. As the language used in communicationincludes facial expression, gesture, atmosphere,etc., nothing is better than face-to-face communication. However,it has limitations regarding time, space, and organization. Information technology can overcomesuch problemsand supplementface to face communication,because articulate knowledge in communicationcan be processed by information technology. AI plays a very importantrole here. Explicit inference including sign manipulation creates knowledgeonly whentacit inference gives sense to explicit inference. Therefore, decision makersshould control their owndevelopmentand operation of information systems. Whatis most important in inference using informationtechnology is intuitive interaction betweenthe computer and the end-user. Consequently, a simplebut flexible user interface is indispensable. AI can contribute to improve organizational inference by providing user friendly interface. 5.2.1 Interpretation Tools Byproviding interpretation tools, information technologies can help proper manipulation and reading of sign. They improvethe process of internalization of articulate knowledge,and assist sense-reading of events surroundingindividuals, groups, and organizations. For instance, raw data can hardly meananything, but statistical analysesusing informationtechnologies can endowdata with meaning. 5.2.2 Metaphors Individual knowledgeis fundamentally tacit and can only be expressed and transferred indirectly by metaphor. Information technologies can assist sensegiving or articulation, whichis the other important process of thinking, by providing effective metaphorssuch as words, data, graphs and images. These metaphors also assist sharing knowledgewhichis the essential process of inference at grouplevel. 5.2.3 SharedField Informationtechnologiescan assist inference at group level by providing shared field. In discussion, group membersoften use a blackboardor a piece of paper because they knowvisual expression complementinglanguageis effective in sharing knowledgethrough sense-giving and sense reading process. A large display with many terminals in a conference roomwill be much moreefficient than a blackboard. Althoughnothing is better than face-toface communication to have a shared field, it is often not possible. Informationtechnologies such as electronic mail and electronic blackboard can overcomelimitations inherent in face-to-face sharedfield, regarding time, space and organization. 5.2.4 Appropriate Process Informationtechnologies can assist inference at organizational level. Group knowledgebecomes organizational knowledge only whenit is regardedto be legitimate. Legitimacy often depends as much on the appropriatenessof the process as it does on the outcomes (March and Olsen 1986). Information technologies help make knowledgelegitimate by facilitating forecast, simulation,analysis, etc. whichare regarded as appropriate processes. 21 6 AI and Organizational Learning Informationtechnologyassisted by arlificial intelligence can improveorganizational decision makingprocess through organizational learning and assist organizations to form and implementappropriate strategies. Organizationallearning is a difficult process for all organizations, becausethe feedback loop amonginference, decision and action, and interpretative frameworkhas to be cut. Especiallyin a loosely coupledsystem,it is extremelydifficult becauseorganizational learning process is inherently incomplete. However,information technologies can assist organizational learning by helping organizational inference as follows: (1) Informationtechnologies are able improvethe process of inference and make dialogue mode evolve. (2) As a result, a divergenceis generated between dialogue modeand decision/action mode, which makesdecision/ action mode change. (3) As the decision~action mode changes, existing interpretative framework becomes incapable of making sense of the environment. Andthe interpretative frameworkof the organization shifts to a newmodethrough reinterpretation of experiences. (4) The improvedinterpretation mode makesit possible for the organization and its membersto perceive what they haveso far beenunable to see. Consequently, tacit knowledgein the organization is enlarged and reconstructed. (5) It makesdialogue modechange, generating a discrepancy between dialogue modeand information technologies. This discrepancy urges information technologies to develop. (6) Such developmentof information technology in return helps dialogue modeimprovefurther. Throughsuch spiral dynamism,interpretation modechanges and strategy formation process evolves as shownin Fig.3. Themost successful corporation of the 1990swill be somethingcalled a learning organization (Fortune Magazine). The organizationsthat will truly excel in the future will be the organization that discover howto makeand keep its interpretation modeenacting through organizational learning. Reference Chandler, A.D., Strategy and Structure, Cambridge,Mass.: MITPress. Daft, R. L. and K. E. Weick(1984) "Towarda modelof organizations as interpretation systems," Academyof ManagementReview 9, 284-295. Duncan, R.B. and A. Weiss (1979) "Organizationallearning: Implications for organizational design," in B. Staw (Ed.), Researchin organizational behavior(Vol. 1), Greenwich,Conn.: JAI Press, 75-123 March,J. G. and J. P. Olsen(1986) "Garbage can modelsof decision mak ing in organizations," in J. P. March and R. Weissinger-Baylon(Eds.), Ambiguity and Command:Organizational Perspectives on Military Decision Making, NewYork: Harper & Row. Mintzberg,H. (1978) "Patterns in strategy New Interpretative Interpretation Framework ~----~ Interpretation Mode e Interpretative Framework Interpretation ,~ Inference t....... t~ Decision/Action s # e ...,"Dialogue Mode Inference through Dlalogue Decision~Action Decision / Action Information Technology Flg.3 Mode } Information Technology and Organizational Learning 22 formation," Management Science 24-9, 934-948. Polanyi, M. (1958) Personal Knowledge, London: Routledge & Kegan Paul Ltd. Polanyi, M. (1966) The Tacit Dimension, London: Routledge & Kegan Paul. Polanyi, M. (1969) Knowing and Being: Essays by Michael Polanyi, (Grene, M. Ed), London: Routledge & Kegan Paul Ltd. Tsuchiya, S (1992) "Strategy Formation Universities: ChangingStrategic Deci- sion Process of loosely Coupled Systems Through Information Technologies," in D. Crookall and H. Arai (Eds.), Global Interdependence: Simulation and Gaming Perspectives, Springer-Verlag. Tsuchiya, S. (1993) "Strategic Learning a loosely coupled system," in Teramoto, Y. et al. Learning Organization, Tokyo: Dobunkan Weick, K. E. (1976) "Educational organizations as loosely coupled systems," Administrative Science Quarterly 21, 1-19. 23