From: AAAI Technical Report SS-93-01. Compilation copyright © 1993, AAAI (www.aaai.org). All rights reserved. SUPPORTING THE CREATMTY CYCLE THROUGH VISUAL LANGUAGES BRIAN R. GAINES and MILDRED L. G. SHAW KnowledgeScience Institute University of Calgary Calgary Alberta T2N 1N4 Canada Abstract: The application of visual language tools for knowledgeacquisition and modelingto the support of student’s creative processes has been investigated. Suchtools have potential application to both phases of the creativity cycle, the transition from loose divergent thinking to focusedconvergentthinking, and the converse transition wherebyreceived knowledgeis deconstructed through detecting anomalies and making overt the underlying tacit presuppositions. This paper outlines this modelof creative dynamics, gives an overview of past applications of visual languages, such as concept maps, in a range of disciplines, and illustrates their use by graduate research students developingtheir research frameworks. 1. Introduction This paper summarizesthe manydifferent forms of concept mapused in different disciplines, gives an overview of how This paper presents a progress report on ongoingresearch on we see them supporting creative processes, and reports some using knowledgeacquisition and modeling tools to support creative processes (Shaw and Gaines, 1982). At the AAAI empirical studies of students on a recent Cognitive Science course using KMapto develop their ideas. Workshopon Creativity in 1991 we presented some early studies of graduate students on a Cognitive Science course 2. KnowledgeModeling in the Creativity Cycle using repertory grid tools to develop their research ideas Barron, who has authored a number of major encyclopedia (Shaw and Gaines, 1992). We also gave these students articles on creativity, has described what he sees as the access to another knowledge modeling tool recently current consensusof the nature of creativity: developed at that time which provided graphic support of a "1. Creativity is an ability to respondadaptively to the needs visual language for KL-ONE-styleknowledgerepresentation for new approachesand newproducts. It is essentially the systems (Gaines, 1991). The students exported their ability to bring somethingnew into existence purlx~efuUy, repertory grids and derived rules to this tool, and were also though the process mayhave unconscious, or subliminally able to use it for direct knowledgeentry. One student in conscious, as well as fully conscious components. Novel particular developedelaborate knowledgestructures for his adaptation is seen to be in the service of increased PhDresearch in this tool and commentedthat he found it flexibility and increased powerto growand/or to survive. more effective in developing his research ideas than a 2. The "somethingnew"is usually a product resulting from a conventional text outliner--it was a visual idea outliner. We process initiated by a person. There are three modes in becameinterested in investigating this type of tool as a which creativity may easily be studied: as product, as support for creative thinking, and realized that the students’ process, as person. using it had not adopted the KL-ONE semantics. They had used the shapes and connections to create concept mapsthat 3. The defining properties of these new products, processes, were expressive and useful to them in thinking, but had no and persons, are their originality, their aptness, their formal semantics (or, at least, did not have those originally validity, their adequacyin meeting a need, and a rather intended). subtle additional property that may be called, simply, fitness---esthetic fitness, ecological fitness, optimum form, Concept maps, as used in education (Novak and Gowin, being "right" as well as original at the moment. The 1984), have already been used as knowledge acquisition emphasis is on whatever is fresh, novel, unusual, tools (McNeese, Zaff, Peio, Snyder, Duncanand McFarren, ingenious, clever, and apt. 1990). and are generated by Cognosys, a text analysis tool 4. Such creative products are quite various, of course: a developed for knowledge modeling (Woodward, 1990). novel solution to a problemin mathematics; an invention; the knowledge modeling research, we were concerned with the discovery of a new chemical process; the composition refining the maps into well-defined formal knowledge of a piece of music, or a poem,or a painting; the forming structures. In the creativity research, we becameinterested in of a new philosophical or religious system; an innovation supporting concept maps in general and developed a tool, in law; a fresh way of thinking about social problems; a KMap,that allows the visual syntax of most commonforms breakthrough in ways of treating or preventing a disease; of concept mapto be emulated. the devising of new ways of controlling the minds of 155 others; the invention of mightynewarmaments,both of offense and defense; newwaysof taxing the citizens of a country by its government;or simplya changein manners for multitudes of people in a specific period of time." (Barren, 1988) be original. On the other hand, loose construing is characterizedby a person’s"preposterousthinking"(p. 529) whichis unformulatedand often preverbal. The structured alternation of these processes is what weperceive as the creativity underlyingintelligent behavior. Thereis priraafacievalidity of knowledge modelingtools to Support through support these creative processes from the literature on kingandSpedly Concepts scientific development, changeand revolution. For example, Torretti, in analyzingthe developmentof modernphysics, PrOCeSSeS considersthe dynamicsof conceptualschemato be the basic framework within whichall other activities can be located. Clarification Organization In particular, he analyzesin detail the essential limits to Convergence conceptualinnovationthat providecontinuity in scientific Loose Cons~uaion~ Creativity TightConstruction~ development, noting that: "C1. Someconcepts are immune to change,and they provide Cycle Associational Definitional Contextual Logical a stable referenceto decisivefacts. Informal Formal C2. The new concepts are arrived at through internal PrOCOSSeS criticism of the old, by virtue of which the facts Support through Supportthrough Deconstruction Hyperrnedia Representation Comparison Know~edge purportedlyreferred to by the earlier modeof thoughtare Systems Systems Criticism effectivelydissolved. C3. Referenceto facts does not dependon the conceptsby Support through Knowledge Modeling Systems usedto whichthey are grasped."(Torreui, 1990) DetectAnomalies andTacit Presuppositions Tools that provide meansto makethese relations between Figure l Creativity cycle and its support concepts moreovert and operational should have a role in the creative processof conceptualtransformation. Figure1 illustrates the natureof the creativity cycleandthe Barren, Torretti, and manyothers, describe the phenomena types of processinvolvedin it, together with the roles that of creativity, but what are its underlying psychological wehave designed the associated computer-basedtools to processes?Oneof the first psychologiststo consider the play in supporting these processes. Notethat wesee that phase of the cycle that is divergent and deconstructiveas relationship between the convergent and divergent knowledgeprocesses of creativity within a cognitive equal in importance to that which is convergent and constructive. It is as importantto be able to deconstrncta frameworkwas George Kelly whose personal construct throughcritical theorypresentsa personas exploringhis or her environment, well-establishedbase of receivedknowledge collecting data about the worldand constructinga personal processesthat can start a newcreativity cycle, as it is to be reality from which to view it (Kelly, 1955). Personal able to clarify and formalizeideas into theories. Becoming construct theory has been the basis of manyknowledge awareof anomaliesand tacit preconceptionsis a stimulusto acquisition and modeling methodologies. The theory the creation of newapproaches,experiments,interpretations encompassesboth the extensional, case-based methodology and theories. This is not only important to experienced of tools basedon the repertory grid, and the intensional, researchers but also to newgraduate students wherethe concept-based methodologyof tools supporting term- weight of received wisdomis often the most significant subsumptionknowledgerepresentation systems(Gaines and burden impedingtheir establishing their ownresearch directionsandidentities. Shaw,1993). Kelly analyzes the dynamics of change in a person’s construct system in terms of a creativity cycle in which convergentand divergent thinking alternate. A system of personal constructs is not fixed, but mayvary from one occasionto another: "Thereare typical shifts in the sequenceof construction which people employ in order to meet everyday situations .... TheCreativityCyclehas to do with the way in whicha person develops new ideas." (Kelly, 1955, pp.514-515) Kelly sees convergence/divergence as a sequence of loosening and tightening of constructions. A person who uses only light construingmaybe very productivebut cannot Creativityis part of a cyclic process.Eachphaseof the cycle is different in nature and requires its ownsupport. At one time one may be clarifying and substantiating one’s preconceptions, at another attempting to identify and demolish them. Different tools and techniques maybe appropriateto different parts of the cycle, andthe sametools maybe used in very different ways dependenton whether the objectiveis a transition fromdivergenceto convergence, or the converse. 3. Visualizationin the CreativityCycle It is no coincidencethat the wordsimageand imagination havethe sameroots. It has beenknownfromearly times that visual imageryplays a significant role in the creative 156 processes of manypeople. Hadamard’s studies in the 1940s showedthe significant role of imageryin creative thinking, andmanycase studies of geniusrecount vivid imagesas the precursorsto significant newideas (Hadamard, 1945). This has led to a specialist literature on the role of imageryin creative thought processes, such as Arnheim’s Visual Thinking(Arnheim,1969)andMiller" s lmageryin Scientific Thought(Miller, 1984). empirically to analyze organizational decision making (Eden, Jones and Sims, 1979), social systems (Banathy, 1991)and the policies of political leaders (Hart, 1977). linguistics, GraesserandClark havedevelopedan analysis of argumentformsin text in terms of structured conceptmaps with eight node types and four link types (Graesser and Clark, 1985), and Woodward has developedtools to extract such mapsfrom text (Woodward,1990). In the philosophy of science, Toulmindeveloped a theory of scientific argumentbased on typed concept maps(Toulmin,1958) that is regarded as one of the major themesof the rhetoric of western thought (Golden, Berquist and Coleman,1976). the history of science, the dynamicsof conceptmapshave beenusedto represent the processesof conceptualchangein scientific revolutions(Nersessian,1989;Thadgard,1992). The availability of low-cost, high-performancegraphic workstationsand personalcomputershas madeit possible to supportvisual thinkingprocessesin a variety of waysacross a wide range of disciplines, and again has led to a major literature on visualization such as the annualIEEEmeetings with this theme (Kaufman,1990; Nielson and Rosenblum, 1991).Thereis also a large andgrowingliterature on the use of computer-basedtools to support visual languageswith Hypertextsystemsthat display overviewsof nodesand links formalsemanticsfor the specification of a widevariety of provide a natural tool for the computersupport of concept structures(Glinert, 1990). mapdevelopment(Smolensky,Bell, Fox, King and Lewis, 1987), and specific tools to do so have been developed The studies reported here are concerned with the (Streitz, Hannemann and Thiiring, 1989; Bernstein, 1992). applicability of visual languagesdevelopedin artificial Frames, semantic networks and hypertext structures have intelligence research to the supportof creative processes. been compared(Travers, 1989), and Jonassen has shown Semanticnetworks began as qualitative modelsof human howthe elicitation of semantic networks maybe used to memoryprocesses (Quillian, 1968) and have been refined structure hypertext(Jonassen,1990). Avariety of techniques over this period to becomehighly formal visual languages representing knowledgestructures based on sound and and computer-basedtools have also been developedfor the analysis of concept maps.Whenthey represent the strict completedenotational or intensional semantics (Gaines, semanticsof a visual languagefor knowledge representation, 1991). Examplesinclude SNePs(Shapiro, 1979), Conceptual it is appropriateto use deductiveinferenceenginesto draw Graphs(Sowa,1984), and various visual languagesfor KLlogical conclusionsfromthe maps(Sowa,1991). Leishman’s ONE(Brachman and Schmolze, 1985). More pragmatic computationof analogical relations as constrainedpartial examplesinclude the graphic user interfaces for knowledge entry, editing andvisualization in commercial tools such as correspondences mayalso be used with such knowledge structures (Leishman,1989). Manycognitive mapsrepresent KEE, ARTand NEXPERT.Nosek and Roth have shown influencesandcausal relations and maybe treated as visual that users comprehendvisual languagerepresentations of languagesspecifyingqualitative simulationswhichmaythen knowledge structures better than they do textual be analyzedusing constraint-based reasoning, or logics of representations(NosekandRoth, 1990). uncertain reasoning (Kosko, 1986; Zhang,Chenand King, Suchdevelopments in artificial intelligenceare paralleled in 1992). Evengeneral conceptmapscan be analyzedin terms other disciplines by the development of visual languagesfor of their structure of their link relations (Cropper,Edenand ’concept maps’, ’cognitive maps’and argumentforms that Ackermann,1990). are visually remarkablysimilar to those cited above,but do 4 SomeExperimentswith KMap not have the formal semantics. In education, Ausubel’s theories of the role of conceptual structures in learning KMap is a generalizedvisual languagetool that providesa (Ausubel and Robinson, 1969) led Novakto develop user interface and data structures for creating, editing, systemof concept maps(Novak,1977) that has been widely analyzing and exporting typed graphs with directed and appliedin the evaluationof students’ learningin the school undirected links. The user interface allows KMapto be system (Novakand Gowin, 1984). There is nowa wide tailored to particular requirementsby specifying the node variety of different forms of concept mapthat have been types required, and associating with visual interface applied in education (Lambiotte, Dansereau, Cross and characteristics suchas shape,color andtext attributes. KMap Reynolds, 1989). They have also been used as tools to recognizesa variety of user interface items, suchas menus support the interviewing process in knowledgeacquisition andbuttons, that neednot be specified in advance,andthis from experts, for example in the Wright-Patterson enablesits interface to be tailored by specifyingdifferent development of the pilot’s associate (McNeese et al., 1990). interface resourcesrather than modifying code. Thespecified In management,Axelrod proposed cognitive maps as a types and interface then allow graphsto be generatedusing meansof representing the conceptualstructures underlying these node types together with lines and arrows as links. decision making(Axelrod, 1976), and these have beenused Linksmaybe labeled usingspecifiedlink label nodetypes. 157 Objective Pr~ of motivation underlyingthe aim Aim J Overallaimofl carrying out J research J I / Objective /Determination of re(1ulres ~ current state of the art relatingto the aim leads to Objective Critical analysis of the current state of theart t p t I / q~- A.ctlvlty. J requires l Pvaluation or ~ _j statea3fthej Activity ~1~ Developmentof detailed research objectives Activity I requires I Activ"YJ J Attendance at J -(H seminars, / workshops & J / conferences J I Activity J I .... I ~ P, nalysls.oT J ~ j - J r~.Ssearcn I I Activity J Synthesis of "7 frameworkfor J research issues Activity Clear statement of objectivesof existing research Activity Clear statement of achievements & problems ~_~ uires leads to Objective Evaluationof research implementation readingof J literatureJ ~J Detailed J "I discussion with ~ requires specialist J J researchers / J Activity J /I / I Development of a / requires ---114 conceptual ~ requires J frameworkfor J researcharea ~.J ¯ Activity Discussions with other researchers research "-l"requires l Activity leads to Objective Research implementation basedon the design requires / J Close reading J /J of relevant J / J literature leads to Objective Research design basedon the analysis ~, ~J Definition of an -I area of ~ requires 1[ / / Activity Intense personal research activity Activity Projectionof next steps Activity Documentation of researchactivities and outcomes requires ~ t Activity Critical analysisof ownresearch and comparisonwith stateof art leads to requires - ;[ Activity ’ ObJelJrctive J JConclusions and ~ requires - - J Imaginative ~ projection of Jfuture directions I J worthwhile J of research l Jdeveiopmants Figure 2 Conceptmapon carrying out graduate research 158 Activity Clear statement of basisof evaluation KMap itself supportsall the interactivegraphicalcapabilities maps. Examplesof Toulmin’s and Gowin’sconcept maps for editing, printing and so on. It supports an interwere discussed and madeavailable, but the tool was application communication protocol enabling other programs otherwisepresented as an open-endedgrapher with links to to makeuse of the graphs, and to interact with the users HyperCardso that nodes could be annotated. A HyperCard throughthe graphs. Henceit maybe used as a front end to stack supportingsuch annotation was providedso that the manyother systems such as conceptual graph and KL-ONE studentsdid not haveto programan appropriateprotocol. deductive systems, qualitative simulation systems, and KMapwas fully utilized by all the students, both Petrinet andcatego~’y-theoretic proofsystems. individually and in conjunction with others discussing, KMap mayalso be used moreinformally to specify concept amending,or developingalternatives to, the primaryconcept mappingenvironments,and to developconcept mapswithin maps developed. The students all developed their own them. It is this type of use that wehave explored with systemsof nodetypes and styles, taking little note of the graduate students on a cognitive science course. The examplesprovided. The node types introduced seem to be objective wasnot only to see howthe availability of such very domainspecific, legitimately so, and this is itself a tools affected their researchactivities, but also to see what lesson for future design. Onelimitation of previousconcept use they madeof the open-endedconcept mapenvironment mapsis that even whennodetyping is supported, it is not development tools. This contrasts with other applicationsin dynamicandthere is no explicit relation betweentypes. The ¯ education wherethe form of concept mapsis specified in open endedstudent use suggests that node types should be advance. treated as a full type lattice. This does not appearto be an artifact of the computingbackground of the students, since Figure 2 showsan exampleconceptmapon writing a thesis the types used are morethose suggested by common sense which we developed and supplied to incoming graduate than data structure theory, but the conclusion should be students. Three node types have been defined, Aim, tested with users with other backgrounds. ObjectiveandActivity, togetherwith twolink types, requires and leads to. A rectangular shapehas beenused to identify Figure 3 showsa typical exampleof a large concept map nodes, and an unboxedshape to identify links. The type producedby a student involved in groupwaredevelopment. nameof a node is shownat the top--this is an optional The students used two pagemonitors, and someof the maps interfacefeature. producedwere large, requiring several pages of printout whenput into their reports. Figure 4 showsthe right hand In the Fall of 1992, werepeated the previously reported side of the samemapso that the content maybe examined. experiment(Shawand Gaines, 1992) in whichstudents in cognitive science course used knowledgeacquisition and This student hadused roundedcorner box to indicate fairly modeling tools to investigate their ownresearchideas. Eight concreteexemplars.Thecluster of exemplarsof hypertexton students again used repertory grids and visual languages the right of Figure 4 forms a natural set of elementsfor individually and in conjunctionwith supervisors and other repertory grid elicitation, and the concept mapsproveda researchers. This time the visual languageediting tool was very natural wayof dividing up an overall domaininto KMap and we did not imposeany semantics on the concept coherentsub-domains that couldbe studiedin moredetail. Figure3 Overallstructureof a conceptmap on Issuesrelatingto collaborative hypertext development 159 platform ( PtontWorks e.g interactive experilentotion ( Intermedia) Hyper :ard ) e.g .~ Showcase ~~ Alternate Reality 1 ) e.g K~t C L.E.G.O ) HyperTIES Aids phonetic ~ e.g Type and # of rinks L ~~/ei ~{E’ulB~otk°r } C Referentio, y {Bidt~eCntk~°n°t} l Organizationat Figure4 Righthandside of conceptmaponIssuesrelatingto collaborative hypertextdevelopment Figure 5 showsan alternative approachadoptedby another Multimediadigital publication supports ’active documents’ studentin expressinghis ideas aboutthe legal protectionof containing not only text and pictures but also operational computersoftware. Hedevelopedfour separate maps,the top modelsfor visualization, animation,simulation, deduction, level one shownas an overall view of the issues in the experimentationand alternative analyses. This will lead to domain,and three linked ones about patents, copyrightsand major changesin scholarly publication in the next decade trade secrets. This student is developinga knowledge-based(Gaines, 1992, 1993). Weare already seeing parallel system to modelsoftware protection issues, and saw the publicationof booksandjournal articles as passiveprint on concept maps as a natural first stage in knowledge paperto satisfy existing publicationoutlets, and as active elicitation. Fromone perspective, a knowledge-based system electronic documents on CD-ROM to support improved developeris not a typical user andthis type of approachmay scientific communication (Claerbout and Dellinger, 1991; haveno general validity. However, this type of application Gainesand Shaw,1992). Fromthis perspective, one can see raises the intriguing questionas to whetherall researchers the tools that weare providingto studentsas the forerunners might not eventually becomeinterested in makingtheir of electronic publication tools providinga moreexpressive research results operationalas knowledge-based systems. medium for the disseminationof knowledge. 160 protected by Figure5 Toplevel conceptmapon the protectionof computer software 5. Conclusions Acknowledgments This paper has provided a progress report on a research activity concernedwith the use of knowledgeacquisition and modeling tools to support creative processes in scholarly research. It would be inappropriate to attempt to draw significant conclusions from the informal studies reported. Wedo not knowwhether our students’ creative processes are being affected by the tools used. Moreover,we do not know how to measure this in a meaningful way. Wehave to rely on the primafacie validity of the approach and on subjective evaluation. This is not particularly satisfactory from a research perspective, and it wouldbe useful to develop and test morespecific hypothesesrelated to theories of creativity such as those of Kelly (1955) and Kim(1991). This work was funded in part by the Natural Sciences and EngineeringResearch Council of Canada. Wewish to thank the 1992 class of CPSC679, Cognitive Processes in Artificial Intelligence. Meanwhile, to tools builders, the support of scholarly research processes offers a very interesting application of knowledgeacquisition and modeling tools. Once scholarly publications are made available through interactive electronic mediait will becomepossible to actively support the creative processes of those reading them. That is, what we nowsee as specialist tools for knowledgeacquisition and modeling, when embeddedin interactive documents, will become part of the routine publication process. This application of artificial intelligence research to the general processes of knowledgedissemination is very attractive. References Arnheim,R. (1969). Visual Thlnk|ng. Berkeley, California, Universityof CaliforniaPress. Ausubel,D.P. and Robinson,F.G. (1969). SchoolLearning. New York,Holt, Rinehart and Winston. Axelrod,R. (1976). Structure of Decision.Princeton,NewJersey, PrincetonUniversityPress. Banathy,B.H. (1991). Cognitive mappingof educational systems for future generations. WorldFutures30(1) 5-17. Barton, F. (1988). Putting creativity to work. The Nature of Creativity. pp.76-98. Cambridge,UK,CambridgeUniversity Press. Bernstein, B. (1992). Euclid: Supporting collaborative argumentation with hypertext. Department of Computer Science,Universityof Coloradoat Boulder. Brachman,R.J. and Schmolze,J. (1985). Anoverviewof the KLONEknowledgerepresentation system. Cognitive Science9(2) 171-216. Claerbout, J.F. and DeUinger,J.A. (1991). StanfordExploration Project: SEP-CD-I: (1) Earth Soundings Analysism Processing versus Inversion; (2) Anlsotropic Seismic Wave 161 Propagation; (3) Progress Reports SEP-72 and SEP-70. Palo Alto, California, Stanford University Press. Cropper, S., Eden, C. and Ackermann,F. (1990). Keeping sense accounts using computer-based cognitive maps. Social Science Computer Review 8(3) 345-366. Eden, C., Jones, S. and Sims, D. (1979). Thinking in Organizations. London, Macmillan. Gaines, B.R. (1991). An interactive visual language for term subsumption visual languages. IJCAI’91: Proceedings of the Twelfth International Joint Conference on Artificial Intelligence. pp.817-823. San Mateo, California, Morgan Kaufmann. Gaines, B.R. (1992). Social dynamics of scholarly publication. AAAI-92 Workshop on Communicating Scientific and Technical Knowledge.pp.l-8. Menlo Park, California, AAAI. Gaines, B.R. (1993). Anagenda for digital journals: the sociotechnical infTastrucmre of knowledgedissemination. Journal of Organizational Computing to appear. Gaines, B.R. and Shaw, M.L.G. (1992). Documents as expert systems. Research and Development in Expert Systems IX. Proceedings of British Computer Society Expert Systems Conference. pp.331-349. Cambridge, UK, Cambridge University Press. Gaines, B.R. and Shaw, M.L.G. (1993). Basing knowledge acquisition tools in personal construct psychology. Knowledge Engineering Review8(1) to appear. Glinert’ I.P., Ed. (1990). Visual Programming Environments: Paradigms and Systems. Los Alamitus, California, IEEE ComputerSociety Press. Golden, J.L., Berquist, G.F. and Coleman, W.E. (1976). The Rhethoric of Western Thought. Dubuque, Idaho, Kendall/Hunt. Graesser, A.C. and Clark, L.F. (1985). Structures and Procedures of Implicit Knowledge.NewJersey, Ablex. Hadamard,J. (1945). An Essay on the Psychology of Invention the Mathematical Field. Princeton, NewJersey, Princeton University Press. Hart, J.A. (1977). Cognitive mapsof three latin american policy makers. World Polltks 30(1) 115-140. Jonassen, D.H. (1990). Semantic net elicitation: tools for strucmrin8 hypertext. Hypertext: State of the Art. pp.142-152. Oxford, UK,Intellect. Kaufman,A., Ed. (1990). Visualization’90: Proceedings of the First IEEE Conference on Visualization. Los Alamitos, California, IEEEComputerSociety press. Kelly, G.A. (1955). The Psychology of Personal Constructs. New York. Norton. Kim, S.H. (1991). A formal model of creative decision making. Robotics and Computer-Integrated Manufacturing 8(1) 5365. Kosko, B. (1986). Fuzzy cognitive maps. International Journal Man-MachineStudies 24(1) 65-75. Lambiotte, J.G., Dansereau, D.F., Cross, D.R. and Reynolds, S.B. (1989). Multirelational semantic maps. Educational Psychology Review 1(4) 331-367. Leishman, D. (1989). Analogy as constrained partial correspondence over conceptual graphs. Proceedings of KR’89: First International Conference on Principles of Knowledge Representation and Reasoning. pp.223-234. San Mateo, California, MorganKauffman. McNeese,M.D., Zaff, B.S., Peio, K.J., Snyder, D.E., Duncan, J.C. and McFarren, M.R. (1990). An Advanced Knowledge and Design Acquisition Methodology for the Pilot’s Associate. Harry G Armstrong Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio. Miner, A.I. (1984). Imagery in Scientific Thought. Cambridge, Massachusetts, M1Tpress. Nersessian, N.J. (1989). Conceptual change in science and science education. Synthese 80(1) 163-184. Nielson, G.M. and Rosenblum, L., Ed. (1991). Visualization’91: Proceedings of the Second IEEE Conference on Visualization. Los Alamitos, California, IEEE Computer Society Press. Nosek, J.T. and Roth, I. (1990). A comparisonof formal knowledge representations as communicationtools: predicate logic vs semantic network. International Journal of Man-Machine Studies 33(227-239) Novak, J.D. (1977). A Theory of Education. Ithaca, Illinois, Cornell University Press. Novak, J.D. and Gowin, D.B. (1984). Learning HowTo Learn. NewYork, CambridgeUniversity Press. Quillian, M.R. (1968). Semantic memory.Semantic Information Processing. pp.216-270. Cambridge, Massachusetts, MIT Press. Shapiro, S.C. (1979). The SNePSsemantic network processing system. Associative Networks: Representation and Use of Knowledge by Computers. pp.179-203. NewYork, Academic Press. Shaw, M.L.G. and Gaines, B.R. (1982). Tracking the Creativity Cycle with a Microcomputer. International Journal of ManMachine Studies 17(1) 75-85. Shaw, M.L.G. and Gaines, B.R. (1992). Mapping creativity with knowledge support tools. AAAI-91Workshopon Creativity: Models, Methods and Tools. pp.32-45. Menlo Park, California, AAAI. Smolensky, P., Bell, B., Fox, B., King, R. and Lewis, C. (1987). Constraint-based hypertext for argumentation. Proceedings of Hypertext’87. pp.215-245. NewYork, ACM. Sown, J.F. (1984). Conceptual Structures: Information Processing in Mindand Machine. Reading, Massachusetts, Adison-Wesley. Sown" J.F., Ed. (1991). Principles of Semantic Networks: Explorations in the Representation of Knowledge. San Mateo, California, Morgan-Kauf~nan. Streitz, N.A., Hannemann"J. and Th0ring, M. (1989). Fromideas and arguments to hyperdocuments:travelling through activity spaces. Proceedings of Hypertext’89. pp.343-364. NewYork, ACM. Thadgard, P. (1992). Conceptual Revolutions. Princeton, New Jersey, Princeton University press. Torretti, R. (1990). Creative Understanding: Philosophical Reflections on Physics. Chicago, University of Chicagopress. Toulmin" S. (1958). The Uses of Argument. Cambridge, UK, CambridgeUniversity press. Travers, M. (1989). A visual representation for knowledge structures. Proceedings of Hypertext’89. pp.147-158. New York, ACM. Woodward,B. (1990). Knowledgeengineering at the front-end: defining the domain. KnowledgeAcquisition 2(1) 73-94. Zhang, W.R., Chert, S.S. and King, R.S. (1992). A cognitive-mapbased approach to the coordination of distributed cooperative agents. IEEE Transactions on Systems, Man and Cybernetics SMC-22(1) 103-114. 162