Differences in Knowledge Processing: Why Operations Researchers Do Not Understand Manufacturing Managers Roger Jenkins, Paul W. B. Hyland and Ross L. Chapman InCITe (Innovation and Continuous Improvement Technologies) Research Centre, School ofManagement, University ofWestern Sydney, Campbelltown Campus. Locked Bag 1797, Penrith South DC 1797, NSW, Australia (r.chapman@uws.edu.au) Abstract Manufacturing Management and Operations Research should be complementary disciplines. This work examines the importance of learning and knowledge capture to both groups as they seek to understand the manufacturing environment. The work discussed in this research is empirical, systematic, and compares a group of OR professionals to a group of manufacturing managers. A particular technology was selected to provide a focus for the detailed analysis: and this technology was discrete event simulation, and the managers involved ·in the research, were based in Australian companies. This paper develops an empirically based analysis of one aspect of manufacturing management of interest to both groups. From the data it is clear that manufacturing managers and operations researchers have different approaches to the collection, storage and codification of lmowledge, and the infonnation subjected to this process is different. These findings indicate that these two groups have significant difficulty in communicating concepts such as knowledge management because oftheir clearly divergent mindsets. Introduction The study of Operations Research (OR) as a distinct branch of management research had its genesis in the late 1930's, (Rosenhead 1992) in a world where the internal constraints of the organization were the dominant barriers standing between the organization and its goals. As markets became saturated, constraints moved outside the bOWldaries of the operations function, and often outside the boundaries of the organization. In this environment, the well-established abstractions of the operations research profession (diagnosis with a reductionist approach; analysis with quantitative techniques against single goals; and treatment of the technology of the process) were not enough. Companies needed more than capacity and volume to compete, issues such as product variation, service speed, reliability and flexibility and product cost and quality became the dominant sources of competitive advantage. Operations researchers have developed techniques that could, for example, optimise overall product mix. to maximize gross margin. Managers, however, are confronted with a market that requires satisfaction on multiple goals, contains unrevealed and unknown alternatives, and has constraints lying outside the domain of the organization, and behaviours in most of the organisation's systems that are mathematically intractable andret1exiv~.ret1exiv~. Research reported in this paper examines the potential mindset discordance between operations researchers and manufacturing managers. The research has used collocate analysis of discourse produced by two groups to develop a representation of the mindset of each group. This representation is tenned a frame (Carley 1997, Gephart 1997). The comparison of the frames of the managers and the OR professionals was used to identify the way in which the two groups approached the problem of knowledge creation and capture, and what sorts of concepts were included as part of the knowledge of each group. This problem has been considered through an examination of the adoption and use of Discrete Event Simulation (DES) by a group of Australian manufacturing managers and the discourse of Operations Researchers on DES at key Simulation conferences. DES is a software-based technology used to develop models of operational processes. These models can then be used to examine the impact of process change, or policy change, on the perfonnance of the process. The technology has received considerable attention in the Operations Research area and has been presented as an extremely valuable tool for practicing managers facing difficult decisions involving process and system restructure or improvement. Learning To remain internationally competitive finns must sustain a high level of learning that both refines current practices and adopts new ones. There is an emerging debate concerning what organisations should learn and how effective learning takes place and is translated into action (Dunphy, Twner, and Crawford, 1996). Clearly at the individual level each of us learns about the enviromnent in which we operate in a variety of ways and with varying degrees of skill and success. However how organisations learn is more difficult to understand as this can be achieved both through individual employees and collectively. Dunphy, et al (1996, 7) have defined "a learning organisation as one which develops and maintains competencies both to perfonn, and to change the organisation to maintain or improve perfonnance." Nonetheless, there is general consensus that organisations need to find better and smarter ways to learn (Argyris, 1994; Senge, 1990). Huber (1991) maintains that "organizational adaptation and innovation, both critical in a rapidly changing world, could undoubtedly be improved if organizational designers and managers knew. more about how organizations learn and about how organizations might be guided to learn more effectively." The importance of knowledge creation and capture stems from the work of Nonaka (1991), where particular skills related to collecting, codifying, storing, and disseminating knowledge for future use is regarded as a competitive advantage of the organisation (Hamel and Prahalad, 1989; Grant, 1991; Leonard-Barton, 1992). In this paper we relate four activities; measuring, sensing, analysing and acting, to the process of collecting, codifying, storing, and disseminating knowledge for use within manufacturing organisations. A key role for all managers, but particularly manufacturing managers, is measuring. It is through measuring the outputs of the production process that they are able to understand how well their plant, equipment and people are performing. By analysing what they have measured they are able to identify areas for improvement, plant in need of maintenance and cost saving. In analysing the data, they are converting it to infolTIlation that they and others can use to create a competitive operation. Once managers have identified problems or areas for improvement they need to act. It is through action that leaming is put into practice. Actions involve the testing of what has been found in analysing the data and managers learn from both their success and their mistakes. It has been recognized that we can collect data explicitly through measuring but data is also collected in a more subtle way by sensing what is occurring in the environment. As mangers walk around their operation they see, hear and even gain a feel for what is· happening on the factory floor. In sensing the environment a manager creates knowledge ofhow well or poorly the operation is functioning. So organisations need managers who are competent at measuring, analysing, sensing and acting in ways that enhances the organisations effectiveness. The processes by which organisations create and improve such organisational competencies have gained increasing attention in studies of organisational learning (Dunphy, et aI, 1996; Grant, 1991; Barney, 1991). Within studies such as these, competencies are regarded as the embodiment of learning to do things better and differently (DlUlphy, et aI, 1996). The capabilities and competencies of a :finn are what it can do as a result of resources acting together. Competencies involve complex patterns of coordination between people, and between people and other resources that lead to differences in sustainable competitive advantage over a long period of time. Learning underpins the creation of competence. It provides quality and consistency, as well as replicability to the action: "The actions are the manifestation of the competence, which the learning has created and made possible. Action has to take place over time and involve improving current perfonnance and making effective change" (Dunphy, et al, 1996). Two groups that have had a dominant, although different role, in lmowledge generation in manufacturing organisations are manufacturing managers and operational research (OR) professionals. A major issue when considering how these two groups learn and interact is that OR professionals and managers use language differently, and they think differently. This is similar to the argument that it was the mindset of OR professionals, not simply communication (Ackoff: 1979) that was the problem when the two groups fail to understand one another. However, language reflects and contains the mindset of the manufacturing operation held by the OR profession and manufacturing managers. While communication is not·the problem, it is the arena in which the problem, as it relates to understanding the differences in knowledge bases of the two groups, is revealed. Language both reflects the illlderlying differences between the two groups, and also reinforces and entrenches the existing positions of each group. This paper examines the differences in how manufacturing managers and OR professionals collect, codify, store, and disseminate knowledge for future use, or, as we have tenned it in this research, how they measure, analyse, sense and act in order to learn. The research has used a construct tenned the frame to represent the mindset used by the groups as they make sense of issues such as measuring, analysing, sensing and acting in the manufacturing function. The construct of the frame will be used to demonstrate part of the reason why manufacturing managers and OR professionals have a different view ofleaming. Methodology The Frame As A Construct It is argued in this paper that it is possible to represent the cognitive structures used to make sense of issues, and this can be based on an analysis of the agent's discourse. This construct, tenned a frame, can then be regarded as a representation of part of what was in the mind of the infonnant at the time the discourse was created (Carley 1997). This definition of a frame is based on that used by Entman (1993). This definition is widely used in the communication literature, and is similar to a range of implicit definitions in the management literature. Entman (1993) proposed that a frame ·would identify concepts, identify their salience and expected associations, and finally, represent these . concepts within a nonnative context. This paper uses the term frame to refer to a construct developed through some fonn of discourse analysis which represents the abstract notions of the world used .by managers as they work to make sense of issues, make decisions and control systems for which they are responsible. The frame sensitises the agent to particular elements of the environment, making those elements prominent for the agent. Concepts in the frame will be given salience, and placed in a framework of associations. Agents operating with a frame will be sensitive to the presence of salient concepts, and will seek out concepts that have not been observed if the frame suggests they should be present. Selecting the discourse Two sources of discourse were used to study the mindsets of OR professionals and manufacturing managers; text for OR professionals and transcripts of interviews for manufacturing managers. Professions are defined loosely as "exclusive occupational groups applying somewhat abstract knowledge to particular cases." (Abbot 1988, p. 8) This body of knowledge will be developed within a community of practice (Brown and Duguid 1991), and will be expressed in internal communication processes which the profession must engage in (Abbott 1988, Dery, Landry and Banville 1993). For an academic community, written text is the mode of communication, and the structures and organization of this community cannot be lUlderstood without reference to the written text (Bazennan 1988). Spoken discourse was chosen as the mode of communication for the managerial community for this paper. This reflects the importance of such activities as spokesperson, negotiation and liaising (Mintzberg 1990) for the manager. Managerial talk is the media through which much of the structures of organizations are enacted and maintained (Mangham 1986), and thus talk is the media which will reflect the mindset of rmnagers, and though which the managerial mindset will evolve. The Manufacturing managers participating in this research were all manufacturing managers based in Australia, employed in a group of companies with international and domestic ownership. Eleven companies wereinvolve~ranging in size from sales of $20 million to $700 million. Semi stroctured interviews were conducted with these eleven managers, the interviews were transcribed, and the transcriptions were analysed using the LINGO software. The OR Group frame was developed based on analysis of a nwnber of papers selected from a major international conference specializing in the area of DES. This conference was the Winter Simulation Conference (WSC), which is held in the US once per· year. Papers were considered appropriate if they were directed at the exploration of applying DES to manufacturing, in a non-hypothetical manufacturing environment. Papers of this nature could be seen to be an attempt by OR professionals to make sense, and develop knowledge about the manufacturing function. A total of 21 papers were selected from the Manufacturing Applications Stream of WSC 1998 using the above selection criteria. The text-based content of these papers was then analysed using the same methodology as that used for the interview transcripts ofthe manufacturing managers. Mapping methodology Mapping of the construct of a frame is not restricted to any particular representation. Measures of association and prominence of concepts can be represented in matrix fonn, as a graphic map or as tabulated data of comparisons. A graphic representation is an effective approach to demonstrate relationships between concepts. In this representation it is possible to report concept prominence and associations simultaneously and extensively. Associations between concepts can be represented in at least three ways, causal, taxonomic or collocate. LINGO, the software developed for this research, uses the collocate method to develop associations. Collocation occurs when two concepts are found in a specified sub text, or window of a specified size, in the whole text. Carley (1997) described the process as moving a window of a specified size throughout the full text. If two concepts are seen at any stage in the window then that is counted as a collocate. The process, tenned collocate analysis in this research, is based on tagging the co-occurrence of words in the text. The concept of control has been studied using this technique (Barley, et. aI., 1988) and in the context of that study the technique was placed in the field of pragmatics. The notion expressed in the field of pragmatics, that the sense of individual words is dependent on context, also Wlderlies the approach followed in .the work reported in this paper, and it the work of Gephart (1997) ·and Carley (1997); the works which strongly influenced the approach embedded in the LINGO software. The analyses undertaken within LINGO are explained in detail in the following section. involve~ LINGO: Content and Collocate Analysis The first stage of the analysis .carried out by LINGO is commonly tenned content analysis. A group of key functions have been identified as desirable in software used to conduct content analysis and these have been incorporated in LINGO. Stone (1997) listed the first four of the following group, the fifth ftmction is not commonly identified as part of the content analysis approach, but is required to develop data on association (or' collocation) between concepts. It perfonns the function identified by Carly (1997) and Gephart (1997). The five functions ofLINGO are thus: • Search for specified text strings. • Count word and concept frequencies (Content analysis). • Standardize the text in the source document, the 'Raw Text', into a smaller subset of concepts that are relevant to the research framework. (ConceptTags) • Confinn the sense of words through the display of local and document level context through a Key Word in Context (KWIC) module. • Be able to search for c9mbmations using the text windowing teclnrique (Collocates) The first four ftmctions are common in much software used to analyse text. The fifth ftmction, collocate analysis, is not nonnally found in content analysis, but enables a much richer representation of the discourse, which will indicate the context of word use as well as a simple frequency. This teclnrique moves a window (of variable size, ten for this paper) through the :full text and accumulates all combinations of words (ConceptTags for this paper) in the window as it progresses through the ext. It creates an array, which represents the way in which words are collocated in the discourse. This provides the analyst with one perspective on the way in which different concepts are associated in the discourse. LINGO represents the collocations between concepts in both numerical and graphical fonnat. Analysis oftext Text, either in the fonn of conference papers or interviews transcripts, was pre-fonnatted using a small Microsoft Word macro before entry to the LINGO system. This pre-fonnatting was used to strip non analysed punctuation and to convert a number of double barrel words, such as 'trade off, into single strings. The text was then converted to a single column of words and periods. This column of text is referred to as RawText. All further analysis is carried out in Microsoft Excel, using a worksheet and set ofVisual Basic modules referred to as LINGO. In building a concordance, the RawText is analysed to produce a list of words found in the text that are not recognized by the existing concordance in LINGO. The researcher then examines this list for words which are significant but which have not been previously encountered by the system. If required, words are selected from this list, tagged with a ConceptTag and then entered. into the system concordance. All subsequent analysis by LINGO is done at the ConceptTag level and all text analysed for this paper was processed with the same concordance ofLINGO. During RawText coding once the RawText has been reviewed for unrecognised vords it is then tagged. In this process each word in the RawText column is tagged with its ConceptTag as dictated in the Concordance. WoreIs that are not successfully tagged are excluded from the list at this stage. Each word is also tagged with its Author. The tagging process is simple. Words that have a sense of 'strategy' can occur as for example; 'strategic', 'strategically', 'SWOT', or 'tactics'. The program searches for 'strateg*', where '*' has the nonnal Boolean tnmcation function. Key word in context analysis (KWIC) is carried out once the RawText Coding is complete using the KWIC module. The KWIC module is a crucial step in achieving validity in the content and collocate analysis. The RawText associated with each occurrence of the ConceptTag is reported in context. This tabulated data is used to confirm that the search process has located words with the correct sense. If for example a ConceptTag was based on the occurrence of the RawText teffi1 'bank', then it is possible to find the word 'bank' used in the sense of a place where financial transactions occur, as the edge of a river, or as a flight manoeuvre in an aircraft. LINGO does not include logic to distinguish sense in the occurrence of a word. This analysis is carried out by the user in the KWIC module. Next content analysis is carried out, where the frequency of occurrence of each ConceptTag in the full document is computed and reported. For collocate analysis LINGO calculates and records the number of collocates for every pair of ConceptTags in the ConceptTag listing. The collocate is the most important entity for the analysis used in this research. Gephart (1997) codes the significance of the collocate using a Z-score, which is a function embedded mthe software used in his study (Gephart 1997, p. 594). This function is based on the standard statistical test used with proportions developed from samples drawn from a nonnal distribution. LINGO counts the number of collocates found in the text for all pairs of ConceptTags. The Z-score is a measure of the probability of finding that number of collocates by chance, given the frequency of occurrence of the separate ConceptTags in the text. A high Z-score indicates that the collocates, which were found, were not likely to have occurred by chance. Consequently the associations reflect some purposeful combination by the author ofthe text. Collocates are included in the analysis if they satisfied the criteria used by Gephart (1997). These criteria required the collocate to have an expected, or observed, frequency of at least three and a Zscore of at least seven. The requirement that at least three collocates are expected, or observed, eliminates collocates with high Z-scores which arose from single (or potentially two) occurrences of low frequency pairs. A Zscore of at least seven indicates that the collocate could not have occurred by chance in that population. Collocates are referenced in three ways in this paper; strong collocates have Z scores which rank them in the 90 th or higher percentile, secondary collocates have Z scores which rank them between the 70 th and 90 th percentile, and significant collocates have·Z scores greater than seven and a frequency ofoccurrence ofthree or more. Results By selecting the ConceptTags analyse, sense act and measure and comparing their collocates as found in the discourse of the two groups we· can get a clear picture of the mindsets of the Manufacturing Managers versus the OR researchers concerning the key aspects of knowledge management and learning. Analyse The analyse ConceptTag coded text such as analyse, evaluate, analytical. It is intended to code text that referred to a process ofanalysis. OR Group Analyse is ranked at the 81 st percentile ranking based on the median ranking of the collocates. The mapping is shown in Fig. 1. Figure 1. OR group concepts collocated with Analyse Case =fall981 Concepts connected to the Pivot concept - analyse Light line indicates 70 percentile collocates. Heavy line indicates 90 percentile collocates. Width of Concept Node indicates frequency of use. The mapping shows strong collocates with design, techmat, cooperation, model, causal, performance, and choice and indicates a very dense network for analyse. Techmat, ranked at the 97th percentile, and model, ranked at the 100 th percentile, are two of the most highly ranked collocated ConceptTags. Market, organize, strategy and finance are disconnected at the mapping level and this supports the view that the focus of the discourse is driven by the need to model and analyse, or codify and store knowledge related to the petfonnance oftechnological systems, independent ofthe business environment. MMGroup Analyse is Janked at the 30 th percentile ranking based on the median ranking of the collocates. There are no strong collocates indicated in the mapping and only one secondary collocate with people.. Analyse is a very weak concept in the frame of the MM group and consequently the mapping is not presented in diagrammatic fonn. Sense The sense ConceptTag coded text such as 'saw, felt, or heard'. It is intended to code text that referred to some process that involved an identification ofa sensory reception. OR group Sense is ranked at the 48 th percentile ranking based on the median ranking of the collocates. The collocate mapping, shown in Fig. 2, indicates no strong collocates and secondary collocates with techmat, analyse, uncertain, certify, model and choice. The mapping indicates that sense was a weakly prominent ConceptTag in the discourse of the OR group. When it is used, it is associated mostly with the prominent set ofConceptTags oftechmat, model and choice. [a1l981 looth performance people. form. MM group Sense is ranked at the 74 th percentile ranking based on the median ranking ofthe collocates. The collocate mapping, shown in Fig. 3, indicates strong collocates with people, act, communicate, market, problem, choice, and enquire. Secondary collocates are mapped with design, techmat, finance, organize, uncertain, model, causal, manage, change, and knowing. Sense is a prominent concept in a broad context. It is collocated with important concepts oftechmat, people, market and organize, but not with strategy, control or improve. This group clearly are much more dependent than the OR group on this means ofcollecting infonnatiolt Figure 2. OR group concepts collocated with Sense Case = fall981 Concepts connected to the Pivot concept - sense ~ ~ ~ ~ Light line indicates 70 percentile collocates.~ Heavy line indicates 90 percentile collocates. Width of Concept Node indicates frequency of use. choice Gh;Qu ~ G;9:) ~ ~ GO~ed~ ~ ~ Act The at.t ConceptTag coded text such as did, do or action. It is intended to code text that has some action sense. OR group Act is ranked at the 65 th percentile ranking based on the median ranking of the collocates. The collocate mapping, shown in Fig. 4, indicates one strong collocate with causal. A wide range of secondary collocates were mapped and these included techmat, people, analyse, model and choice. These ConceptTags were frequently mapped for the OR group. Secondary collocates were also mapped for enquire and plan that were less frequently mapped for the OR group. Strategy, finance, market and organize were disconnected on the mapping,indic~tingan internal focus for this ConceptTag. MMgroup Act is ranked at the 92 nd percentile ranking based on the median ranking of the collocates. The collocate mapping, shown in Fig. 5, indicates a very dense network with strong collocates with the majority (twenty one out of a possible thirty four) of ConceptTags. It has secondary collocates with strategy, improve, finance, analyse, organize, certify, abstract, decide, and knowledge. It is disconnected from model, rules, predict and techit. These four ConceptTags were not prominent in the discourse of the MM group. This network indicates that the ConceptTag act is prominent in the :frame ofthe MM group. infonnation. rall981 ~ ~~~=~::::::::::::~~~~~~~GnqUire.::> ~ ~ ~Cch~llll:> ~~ ~err;;;:> ~ act :frequently indicating Figure 3. MM group concepts collocated with Sense Light line indicates 70 percentile collocates. Heavy line indicates 90 percentile collocates. Width of Concept Node indicates frequency of use. Figure 4. OR group concepts collocated with Act Case = [a1l981 Concepts connected to the Pivot concept - act GncertaV ~ ~ model Light line indicates 70 percentile collocates. Heavy line indicates 90 percentile collocates. Width of Concept Node indicates frequency of use. ~ G1o;eV ~ Figure 5. MM group concepts collocated with Act Light line indicates 70 percentile collocates. Heavy line indicates 90 percentile collocates. Width of Concept Node indicates frequency of use. Measure The measure ConceptTag coded text such as measure and account. It is intended to code text that referred to a numerical measure ofsomething OR Group Measure is ranked at the 54 th percentile ranking based on the median ranking of the collocates. Numeric measures and the mapping indicate a moderate prominence for this PivotConcept. The collocate' mapping indicates a context with strong collocates with techmat, performance, and choice. Secondary collocates are indicated with analyse, prescribe, communicate, uncertain, certain, certify, model, and causal. People, finance and control are disconnected from the mapping, indicating that measure is used within a context set by model, techmat and analyse. This pattern of associations indicates that there is a focus on that part of reality that can be objectively measured, and a strong presence for uncertain, certain and certify indicates that there is a preoccupation with finding a truth and acceptance of this truth by other people. The presence of causal and analyse is also consistent with the mechanistic ontology Furthennore, the absence of people suggests that this mechanistic ontology is being applied to a mechanistic world, people, organizations and markets are only peripheral actors in this world. MMGroup Measure is ranked at the 65 th percentile ranking based on the median ranking of the collocates. Numeric measures and the mapping indicate a moderate prominence for this PivotConcept. The collocate mapping indicates strong collocates with act, communicate, performance, and enquire. Secondary collocates are mapped with control, techmat, people, uncertain, certain, causal, performance, and choice. People are more prominent in this network than techmat, and the network is disconnected from finance, strategy and market. The prominence or extent of measure is very similar for both groups and the frames are not presented here. The context however has changed.. The bias of the MM group for act has seen a change from model and analyse (for the OR group) to act, and the inclusion of control, people and enquire. The MM group have a more prominent position for people in their mapping, and they appear to be more focused than the OR group on control and the concept of enquire. coliocates. gO account PivotCon:ept. collocate world PivotConcept changed. Conclusion The results indicate clear differences in the processes of collecting, codifying, storing and dissemination of infonnation. The OR group collect their information through a process of measurement, the managers rely more so on sense, a process which is more likely to be associated .with tacit lmowledge. The codifying which .takes place is a visible and central activity for the OR group, in the conventional sense of analysis. This process is not visible in the discourse of the managers. The mangers manifestly could not act in the workplace if this codifying· process were not taking place, but it clearly is not discussed in the same way between the two groups. The increased focus of the mangers on the concept of action suggests that the lmowledge is codified within the tacit dimension, and stored in the action. The OR group is codifying infonnation, through a process of analysis, and storing the knowledge as explicit knowledge in models of manufacturing situations. Models, and discussion of models, then becomes an important process of dissemination for OR professionals. The managers are acting on tacit knowledge, using systems of codification which are not expressed in the discourse, storing the lrnowledge in actions. The stories told by managers, the war stories, based on action (Brown and Duguid 1991) perform the role filled by models for the OR professionals. The war stories are the representations of the learning which has taken place for managers, and which is embodied in the mindsets ofthe managers. Strong support was found for the view that OR professionals have a strong bias to an analytical approach, low levels of dependency on tacit knowledge, and used explicit models to store the knowledge created in their work. This is a different perspective to that adopted by managers. Managers were more likely to use tacit means of collecting infonnation, and to apply it in ways that did not require codification. These differences are consistent with arguments proposed in the literature that have attempted to explain the low levels of use of DES by manufacturing managers. To the extent that these differences are valid for wider groups of professionals, then problems will exist in .relationships between managers and professionals in those domains also. The two groups use different sets of knowledge, and so, while the OR professionals and manufacturing managers knowledge bases should be complementary, unless one group can comprehend (though not necessarily be immersed within) the domain of the other, then there will be substantial barriers to effective communications between the two groups. This work indicates that OR professionals process and store lmowledge differently to manufacturing managers, and this is understandable. What is problematic, however, is the observation that the infonnation used is lUlduly focused in areas which are delineated by the technology of modelling, and this leaves extensive domains of the managers area 00 referenced in the discourse of the OR professionals. If OR professionals are to be more relevant to practicing managers then their knowledge domain will need to reference the wider domain of managerial action. to. information References Abbott, Andrew. 1988, The System ofProfessions, The University ofChicago Press, London. Ackort: Russell L. 1979, "The Future of Operational Research Is Past", Journal of Operational Research Society, vol. 30,00.2, pp. 93-104. Argyris, C. 1994, "Good Communication that Blocks Learning", Harvard Business Review, July August. Barley, Stephen R., Gordon W. Meyer, and Debra C. Gash. 1988, "Cultures of Culture: Academics, Practitioners and the Pragmatics of Normative Control", Administrative Science Quarterly, no. 33, pp.24-60. Barney, J.B, (1991), "Finn Resources and Sustained Competitive Advantage', Journal of Management, Vo1.17,No.l,pp90-120. Bazennan, Charles. 1988, "Shaping Written Knowledge", The University ofWisconsen Press, Madison. Brown, John S. and Paul Duguid. 1991, "Organizational Learning and Communities-o±:Practice: Towards a Unified View of Working, Learning, and Innovation", Organization Science, vol. 2, no. 1, pp.40-57. Carley, Kathleen M. 1997 "Extracting team mental models through textual analysis." Journal of Organizational Behavior. pp553-558. Dery, Richard, 'Landry, Maurice and Banville, Claude. 1993, "Revisiting' the Issue of Model Validation in OR: An Epistemological View", European Jotrnal ofOperational Research. no. 66, pp. 168-83. Dunphy, D. Turner, D. Crawford, (1996), "Organisational Learning as the Creation of Corporate Competencies". paper No 060, Centre for Corporate Change, Australian Graduate School of Management. Entman, Robert M. 1993, "Framing: Towards Clarification of a Fractured Paradigm", Journal of Communication, vol. L1-3, no. 4, pp. 51-58. Gephart, Robert. 1997, "Hazardous Measures: an Interpretive Textual Analysis of Quantitative Sensemaking During Crisis", Journal ofOrganizational Behaviour, vol. 18, pp. 583-622. Grant, R.M. 1991, "The Resource-based theory of Competitive Advantage: Implications for Strategy Fonnulation". California Management Review, vol.33 no.3, ppI14-135. Hamel, G. Prahalad, C.K. 1989 "Strategic Intent". Harvard Business Review, vol.89, no. 3, pp. 63-76. Huber, G.P. 1991, "Organizational Learning: The Contributing Processes and the Literatures". Organization Science, Vol. 2, No 1. Leonard-Barton, D. 1992, "The Factory as a Learning Laboratory". Sloan Management Review, Fall Mangham, lain L. 1986, "Power and Perfonnance in Organizations", Basil Blackwell, Oxford. Mintzberg, Henry. 1990, "The Manager's Job: Folklore and Fact", Harvard Business Review, March April, pp. 163-76. Nonaka, I. 1991, "The Knowledge-Creating Company", Harvard Business Review, Nov-Dec, pp. 96 104. Rosenhead, Jonathan. 1992, "Into the Swamp: the Analysis of Social Issues", Journal of Operational Research Society, vol. 43, no. 4, pp. 293-305. Senge, P. 1990, "The Leader's New Work: Building Leaming Organizations". Sloan Management Review, Fall. 30, 00. 2,