足疗对身体有什么好处:CONCEPTUAL FRAMEWORKS FOR NETWORK LEARNING ENVIRONMENTS: CONSTRUCTING PERSONAL AND SHARED KNOWLEDGE

Michael J. Jacobson
James A. Levin
Contact:Dr. Michael J. Jacobson
Learning and Performance Support Laboratory and Instructional Technology
611 Aderhold Hall
University of Georgia
Athens, GA 30602-7101
Email:mjacobso@coe.uga.edu
International Journal of Educational Telecommunications, 1(4), 367-388.
Permission to publish electronically granted by the publisherthe Association forthe Advancement of Computing in Education (AACE)
TABLE OF CONTENTS
ABSTRACT
INTRODUCTION
THE DISTRIBUTED NETWORK LEARNING FRAMEWORKNetwork Mediators and Flow of Information and Knowledge
Expected Value of Information
Information Optimization
Illustrating the Distributed Network Learning Framework
KNOWLEDGE SPACES, HYPERTEXT, AND NETWORK LEARNING ENVIRONMENTSKnowledge Spaces and the DNLF
Hypertextual Knowledge Spaces
DISTRIBUTED NETWORK LEARNING FRAMEWORK, KNOWLEDGE SPACES, AND NETWORK RESEARCH PROJECTSThe Message AssistantHyperlinks and Constructing Personal Knowledge Spaces
The Message Assistant and User-Defined Message Processing Rules
Message Assistant and the DNLF
Learning Resource Server
CONCLUSION
REFERENCES
ACKNOWLEDGMENTS
NOTES
Educational uses of networks are rapidly expanding as the problems of"ease-of-use" and "access" are gradually being solved. However, even as theseproblems are being solved, the solutions create second-order problems, such asstudents and teachers becoming overwhelmed with massive amounts of networkgenerated information. In this paper we present conceptual frameworks thatcharacterize some of the unique properties of network learning environmentswhich then can be used to provide systematic guidance to the design of networklearning activities and software tools. We illustrate these frameworks byshowing how they have helped us design two different tools for educational usesof networks: the Message Assistant and the Learning Resource Server. Theutilization of these frameworks to address general issues related to networklearning environments is also considered.
Considerable progress is being made in the development of a computer networkinfrastructure that can serve education. Research is also suggesting waysdistributed electronic networks may be employed in instructionally beneficialways (Bruce & Peyton, 1992; Hunter, 1992; Levin, Riel, Miyake, & Cohen,1987; Newman, Goldman, Brienne, Jackson, & Magzamen, 1989; Riel &Levin, 1990). Further, improved communication software allows easier access tonetwork-mediated resources such as electronic mail, electronic bulletin boardsystems, and information servers (e.g., Eudora, Mosaic, Gopher). But theincreasing power and ease-of-use of these network software programs is atwo-edged sword. On one hand, students and teachers can now easily communicateelectronically and access a wide range of information resources. On the otherhand, with a huge mega-network like the Internet (10 to 20 million users andover a million servers, and growing daily), teachers and students may beoverloaded with hundreds of messages a day or have trouble deciding on which ofthe millions of Internet servers to look for a particular piece of information.The very richness of these resources imposes not only logistical, but alsospecial cognitive, demands on the user that may unfortunately diminish thelearning potential of educational network use (Riel & Levin, 1990; Ruopp,Gal, Drayton, & Pfister, 1993).
A critical area for research involving educational networks is at this junctureof cognition and the network learning environment: the network software toolsand the human-computer interface. A new class of tools for network learningenvironments is needed which preserves the easy utilization of networkresources, yet also helps deal with the cognitive complexity associated withdistributed network-mediated learning activities. These tools must bebased on (a) an understanding of the special characteristics of distributednetwork learning environments, (b) the situated nature of human cognitivefunctioning, and (c) ways to support the attainment of substantive educationalgoals such as learning complex knowledge, problem solving, and independentthinking.
Whereas there has been considerable work dealing with these last two areas(Brown, Collins, & Duguid, 1989; Bruer, 1993; Clancey, 1993; Greeno &Moore, 1993; Lave & Wenger, 1991; Norman, 1993; Suchman, 1987), there hasbeen little consideration of the first. In this paper, we consider aperspective describing some of these "special characteristics" of educationalelectronic networks, the Distributed Network Learning Framework (Levin& Jacobson, 1992), and the links this framework has with currentunderstandings of human cognitive functioning and new views of classroomlearning. First, we review our earlier work on the Distributed NetworkLearning Framework (Levin & Jacobson, 1992) and then we discuss aKnowledge Spaces conceptual model based on this framework (Jacobson& Levin, 1993; Levin & Jacobson, 1993). We conclude with a discussionof prototype network software tools based on the Distributed Network LearningFramework and the Knowledge Spaces conceptual model.
The Distributed Network Learning Framework (DNLF) consists of three mainelements: (a) network mediators and the flow of information and knowledge, (b)expected value of information, and (c) information optimization. We considerthese three elements in turn.
Fundamental to our view of electronic networks supporting learning activitiesis that there are a variety of mediators, both human and computer-based,at nodes on the network. These mediators control the flow of information. Oneimportant characteristic of electronic networks is the rapid flow ofinformation through the network. Given this rapid movement of networkinformation, decisions must be made at the network nodes about the nature andvalue of this information. One general result of these decisions is the moregradual flow of organized information or knowledge.
The DNLF is based on the following general principle: Information appearingat each network node is stored locally if the expected value of storing thatinformation is positive. The expected value of information storage is theexpected benefit minus the expected cost. For human mediators, theevaluation of expected value can be quite complex and situation specific. Incontrast, computer mediators (sometimes called agents) typicallydetermine the expected value more crudely using syntactically orientedalgorithms or rules checking for pre-specified text patterns.
The expected value of information is estimated from the probability of needingthe information again (i.e., a prediction problem). The simplestapproach to prediction is to assume that the future will be like the past, andthus to predict the likelihood of an event occurring in the future will be thesame as the occurrence of the event in the past. If the computer-basedmediator analyzed its log of past needs for a particular kind of information,it could, under this "the future will be like the past" assumption, predictthat information needed frequently in the past will be needed again in thefuture. The decision could thus be made to locally store information that wasfrequently accessed in the recent past. Human mediators may also employ other,more sophisticated semantic ways of predicting the future, such as usingbackground knowledge and mental models about a domain to make causal inferencesabout potential value of information. But in the absence of such knowledge,the "past frequency" rule can be used.
Related to the expected value of information is another element of the DNLF:Each node in the network attempts to optimize its functioning by storingthings that are likely to be used again. Local storage of information thusoccurs over time, essentially creating a local "mini-network" or database ofinformation. This local storage of information may also be internalized by themediator (i.e., resulting in individual learning).
There are several implications of this element of the DNLF. For example, in alearning environment, there would be added value to information that is usedrepeatedly at a particular node. There would be a tendency to increase thejudged probability of needing that information again, and thus the probabilitywould be raised that the mediator at the node would attempt to store or learnthe information. This rule also suggests that over time there would be agradual acquisition of expertise, as the local storage of neededinformation would also be expected to become increasingly richer and moreorganized.
One may also regard the optimization occurring at a node as analogous to theknowledge representation and learning processes of accretion, tuning, andrestructuring (Rumelhart & Norman, 1978). Initially, there would bethe mere accretion or accumulation of information (both in terms of thehuman representations in memory and the computer-based storage). Over time,this unordered amassing of locally stored information would become unwieldy andwould likely exceed the locally available storage. The mediator at that nodecould then be motivated to develop more differentiated and organized knowledgestructures. These new knowledge structures would then serve as the basis fromwhich the mediator would evaluate new network-based information and would betuned as new information is locally stored and used at that node.Finally, with the continued access of dynamically changing network informationand changing learning needs of the mediator, it will likely be necessary torestructure or develop new representational structures. For example,the human mediator might construct new schemas or mental models, while for thenetwork mediator might require new rules or hyperlink configurations to becreated.
To illustrate key components of the DNLF, let us first consider a non-computerexample: Why does someone learn something? (See Table 1.) From theperspective of this framework, one reason people learn something is because itis easier than regenerating the information the next time they need it. Peopledo not memorize every telephone number they call. They only do so if theyperceive the value of knowing the number by heart to be worth the effort tonecessary to commit the phone number to memory. If someone calls Joe‘s Pizza,she might initially look up the number in the phone book. The next time sheneeds to call for a pizza, she might need to look it up again. If she callsJoe‘s Pizza a lot, she may write the number on a piece of paper and stick itnext to the telephone. After a while, if she is a real Joe‘s Pizza fan, shewill learn the phone number by heart. From the perspective of the DNLF, thehuman mediator (the pizza lover) determined that the expected value of theinformation (phone number of Joe‘s Pizza) was greater than the cost or effortof storing the phone number externally. The phone number information wastherefore initially moved from the phone book to the piece of paper by thephone. Over time, the probability of needing the information again increased(because she was a real Joe‘s Pizza fan), and she restructured the storage ofthe information to optimize access to it by finally memorizing the phone numberand not needing the paper by the phone.[1]
Table 1
Selected Features of the Distributed Network Learning Framework.
Distributed Network Learning Framework Feature Example Comments
Information flows through networks based on decisions made by mediators (students/teachers or a computer agent) at each node. Pizza lover (i.e., the mediator) decides initially to look up the number, then to write it down on a note, then to remember it.
Information and knowledge flows toward where the learners need it. The phone number moves from the phone book, to the note, to the pizza lover‘s memory.
Information appearing at a network node is stored locally if the mediators expect the information to be of value. Pizza lover makes different decisions about locally storing information, first looking it up, then writing it down, and finally learning it. The phone number would not have been stored if the pizza turned out to be bad and thus that piece of information judged to be of little value.
Over time, mediators at nodes would optimize the organization of locally stored information. The pizza lover gradually optimized her ability to access the pizza phone number, as it was initially slow access in the phone book, then faster access with the written note, and fastest when committed to memory.
Network-based information is not a static, fixed "thing," but rather is dynamic, fluid, and changing.
Joe‘s Pizza may get a new phone number.
But how are we to convey these distinctive--but unfortunately somewhat complexand abstract--characteristics of network learning environments described by theDNLF to students and teachers? How can this framework be used to inform thedesign and use of distributed network learning environments? Given research onthe value of an appropriate conceptual model to assist users in operating acomplex device or computer program (Norman, 1988; Norman & Draper, 1986),we have been developing network software tools based on the DNLF that employ aKnowledge Spaces conceptual model. The Knowledge Spaces conceptualmodel can function generatively to evoke multiple metaphors and analogiesrelevant to articulating different conceptual aspects of network learningenvironments.[2] At the core of thisconceptual model is a spatial metaphor that has useful target aspects fordescribing important characteristics of electronic networks (e.g., "objects canbe physically located in different places separated by space" maps to networklearning environments where "computers and people are physically located indifferent places and separated by space"). The Knowledge Spaces conceptualmodel also suggests other familiar systems that can serve as a coordinated setof analogies for describing network learning environments. For example,"highway networks that connect physically separated places and people fortransportation purposes" metaphorically maps to "electronic networks thatconnect distributed computers for information transmission purposes."
The Knowledge Spaces conceptual model can also be used to articulate abstractepistemic notions about "knowledge structure" and constructivist approaches tolearning that are relevant to instructional uses of network learningenvironments.[3] For example, Wittgenstein(Wittgenstein, 1953) employed the metaphor of knowledge-as-a-landscapein the preface to his Philosophical Investigations, while a"criss-crossing the knowledge landscape" metaphor (inspired by Wittgenstein)has been used for research into a constructivist conception of the nature oflearning (Jacobson et al., in press; Jacobson & Spiro, 1995; Spiro,Feltovich, Jacobson, & Coulson, 1992; Spiro, Vispoel, Schmitz,Samarapungavan, & Boerger, 1987). A network learning environment may thusbe "viewed" as a knowledge space that can be criss-crossed or explored fordifferent purposes and from different conceptual perspectives, with differentlearning possibilities afforded by each traversal.
Another aspect of "knowledge spaces" is that they exist along a continuum frompersonal to shared (Jacobson & Levin, 1993; Levin & Jacobson, 1993)(see Figure 1). Personal Knowledge spaces are constructed for one‘sindividual learning and knowledge utilization purposes (e.g., personalelectronic mail messages or a personal "knowledge-base" intended only for asingle person‘s reference or future use). In contrast, shared knowledgespaces are created for information and knowledge dissemination involvinglarger audiences (e.g., electronic mailing lists, bulletin board news groups,distributed information servers).

Figure 1. The spectrum of knowledge spaces and representative corresponding network resources.
But how is it that the Knowledge Spaces conceptual model can help conveyaspects of the DNLF? This is done in three main ways. First, the spatialnature of this conceptual model presents the movement or "flow" of informationover a network as being metaphorically similar to the movement of objects fromone location to another in a physically distributed environment. Second, therapid nature of this movement is evoked by a different metaphor of electronsand electronics. Most students understand that electricity traverses physicalspace and does so very, very quickly. Third, the "personal and shared"continuum of the Knowledge Spaces model helps evoke the various ways networkknowledge may be organized, ranging from one‘s "personal space" to the moregeneral public "shared spaces." And the various types of mediators on thenetworks are also suggested. A personal knowledge space may primarily haveone‘s self as the mediator (and one‘s personal computer-mediator tools oragents) making the determination about the value of network information andwhether or not to store it, while the shared knowledge space will have otherhuman and computer-mediators helping to structure the public knowledge space.Our main point here is that the majority of educational network users will bemore comfortable with metaphors and analogies that are generatively derivedfrom a core knowledge spaces metaphor than they would be from the moretechnical DNLF.
Hypertextual tools are ideally suited for constructing the conceptualinterconnectedness that is central to our notion of personal and sharedknowledge spaces (Jacobson & Levin, 1993; Levin & Jacobson, 1993).[4] Powerful hypertext and hypermediatechnologies have been developed and are now available (Conklin, 1987). Adefining structural characteristic of hypertext and hypermedia is the use ofhyperlinks between nodes of information. Electronic networks mayalso be conceptualized in terms of a link and node structure. Notsurprisingly, we are now seeing a merging of hypertext/hypermedia and networktechnologies. In particular, network-based hypertext and hypermedia softwaretools, such as the World Wide Web servers and browsers (Netscape, Mosaic), arenow available for application in network learning environments, and provideflexible and nonlinear access to vast amounts of globally distributedinformation.
Also, research is beginning to emerge that suggests ways personal hypertext andhypermedia learning environments can help students to learn complex knowledge(Beeman et al., 1988; Jacobson et al., 1994; Jacobson & Spiro, 1995;Jonassen & Wang, 1993; Lehrer, 1993; Shapiro, 1994). Hopefully, futureresearch will document ways in which learning can be promoted through the useof hypertextual shared knowledge spaces as well.[5]
Conceiving of activities conducted over network learning environments from theperspectives of the DNLF and Knowledge Spaces suggests that software tools needto provide specific kinds of functionality beyond mere network access orgraphical user interfaces. In this section we consider the prescriptiveapplication of the Distributed Network Learning Framework and the KnowledgeSpaces conceptual model to the development of tools for network learningenvironments. We discuss in turn two of our ongoing research projects, theMessage Assistant and the Learning Resource Server.
The Message Assistant is an electronic mail system designed to assist a networkmediator in constructing personal Knowledge Spaces (Jacobson &Levin, 1992; Jacobson & Levin, 1993; Levin & Jacobson, 1993; Levin& Jacobson, 1992). The program offers standard electronic mail optionssuch as message creating, sending, receiving, forwarding, and replying (Figure2).[6] In addition, the Message Assistantprovides two sets of advanced features: hyperlinks between messages anda rule-based mediator for processing of messages. These are thespecific features that implement elements of the Knowledge Spaces conceptualmodel and thus are intended to address characteristics of electronic networksdescribed in the DNLF.


Figure 2. Zero-g World Design Project message first dealingwith magnetic shoes from a high school student (top screen)and the initial reply from a Lockhead engineer at theJohnson Space Center (with assigned Message Viewsdisplayed).
As discussed in the first section of the paper, hypertext and hypermedia aretechnologies well-suited for implementing a Knowledge Spaces metaphor fornetwork learning environments. In a large set of messages generated over aperiod of time (ranging from weeks to an entire school year) during adistributed network learning activity, there are many messages sent by theparticipants dealing with different issues, topics, or themes. Hyperlinks mayconnect related messages on these different issues, topics, or themes. Thishyperlinked web of interconnected messages thus would define a personalknowledge space for that network project message set. Furthermore, because ofthe unique situated contexts of the many participants in the network project(e.g., students of various ages in different classes and parts of the world,teachers, parents, university faculty, scientists, business partners), each onewould probably elect to organize or to interconnect the same set of messages indifferent ways.[7] With a hypertextualKnowledge Spaces tool, multiple dimensions of interconnectedness that exist inthe project messages could be specified with different hyperlink sets createdby the various users.
The Message Assistant is a prototype of such a hypertextual Knowledge Spacestool for electronic mail messages. To create a knowledge space with theMessage Assistant, we have implemented two classes of hyperlink mechanisms:fixed hyperlinks and variable hyperlinks.[8]
Fixed hyperlinks. Fixed hyperlinks allow quick, nonlinear accessbetween related messages. These "traditional" hyperlinks may be manuallycreated between different messages by the user or automatically by the programwhen replying to or forwarding a message. Fixed links allow the user toorganize and access related messages in a nonlinear manner.
Variable hyperlinks. While fixed hyperlinks are quite useful,unfortunately they must be rigidly specified. For example, if a fixed link isset between two different messages, that link is always there, under allconditions. This means fixed hyperlinks cannot be readily used to representdifferent conceptual frameworks or information access conditions. For thisreason, we have also implemented another class of links in the MessageAssistant: variable hypertext links.
To help evoke the "Knowledge Spaces" metaphor in the program interface, werefer to variable hyperlinks as "views" of the locally stored messages.[9] Two default views of the messages are "InView" and "Out View" that correspond to received and sent (or to be sent)messages (Figure 5). The user may then create new views of the messages thatshare a common topic, theme, issue, or other conceptual/thematic relationship.Each view represents one possible set of hyperlinks between a subset of themessages. Switching to a different view thus re-configures the links betweenmessages, hence the term "variable hyperlinks."
The Message Assistant was used to organize multiple views of the Zero-g WorldDesign Project messages based on a number of topics and themes found in themessages, such as Magnetic Boots, Dishwashers in Space, or Dribbling (seeFigure 3). As many messages contained information that dealt with severalissues, it was quite common to find such messages in several views (i.e., aheterarchical structure). In addition, the program allows both for themediator to manually assign a message to a view and for the user-defined rulesto automatically assign messages to a view (see below). Thus the MessageAssistant allows the user to create multiple views (either manually orautomatically) and then to explore or criss-cross the messages from thesemultiple conceptual perspectives.


Figure 3. In View listing of Zero-g World DesignProject messages listed by date with "pop-up" list ofmessage views (top screen) and variable hypertext links ofthe Magnetic Boots view (bottom screen).
Unfortunately, the potential exists for educational users of national andinternational educational networks to become inundated with large amounts ofelectronic messages. Similar to earlier work on filtering mechanisms forelectronic mail (Malone, Grant, & Turbank, 1986), the Message Assistantpermits the user to have messages processed through a set of user-definedrules. These rules consist of conditions--Boolean testing for textstrings in the different message fields--that can then trigger several possibleactions. There are four Rule Actions: Auto Reply, AutoForward, Priority, and Auto Add Views (see Figure 4).


Figure 4. List of user-defined rules (top screen) and sample user-defined rule.
In the "real world" use of Message Assistant, user-defined rules mayautomatically check incoming messages and function as a message filter.Messages from a friend or on a specific project might be assigned a highpriority, while other messages might be assigned a lower initial priority. Theuser can also easily ignore the message prioritizations since an overviewlisting of all messages is available and the user can select any messages toread. But we anticipate that when the user is confronted with a large numberof new messages to read, the preprocessing and prioritizations of messagesbased on the user‘s own customized set of rules will prove helpful in decidingwhich messages to read immediately and which to read later.
Furthermore, as students will typically collect a large number of messages innetwork-based learning activities such as the Zero-g World Design Project, theuser-defined rules and the hyperlinking features of the Message Assistant areintended to assist them in constructing their own personal Knowledge Spacesfrom this information. A possible scenario of use for students in a networklearning project would be for an initial set of rules to preprocess and filterincoming messages. Then, over a period of weeks or months, students wouldevolve different rules to conceptually interconnect messages based on ideas,themes, issues, and so on. They would be thus creating their own personalknowledge spaces. Note that creating such a personal knowledge space out of alarge corpus of messages such as the Zero-g World Design project would be muchmore difficult to accomplish with a more traditional electronic mail programthat merely lists messages that have been received or sent. Indeed, it mighteven be impossible to create a personal knowledge space given the"read-and-delete strategy" that is common among many users because importantinformation about the project would be forgotten and the original messages lost.
Whereas the previous section considered the Message Assistant from a KnowledgeSpaces perspective, there are also several ways in which these featuresinstantiate components of the DNLF. The network mediator and flow ofinformation and knowledge component holds that probabilistic evaluations ofnetwork information are made by a mediator at a particular node. The MessageAssistant rules function as a computer-based mediator that assists a humanmediator in making an initial determination of the expected value of theinformation. The message rules form a user-defined expert system which isincrementally specified and tested over time and thus gradually increases itsexpertise with use. Such an expert system serves as a computer-mediator thattakes over repetitive or low-level evaluations of the expected value of certaintypes of network information and then automatically initiate actions whichfilter, route, or store that information. Information that does not match anypre-specified rules is passed on to the human mediator to be evaluated. If thenew information is regarded as being of value, then the human mediator willoperate on that information.
The variable hyperlinks feature also helps to instantiate the DNLF principle ofoptimization (i.e., network nodes attempt to optimize their functioningby locally storing potentially useful information). The message Views providethe user with a mechanism to create a personal knowledge space that is alsoflexible in terms of organizing and accessing the information in the messages.Over time as the human mediator works with the dynamic and changing learningenvironment, the specification of message views and rules would be expected toprogress from gradual accretion, to the tuning use of a body of accumulatedviews and rules, to the restructuring of views and rules.
Tools for creating hyperlinks are ideally suited for structuring multipleorganizing frameworks for any given set of information. This type offunctionality is especially important when a wide range of people need toflexibly access and use information, as is the case with network-basedinformation servers. However, merely providing flexible access to informationwith hyperlinks is not sufficient to ensure substantive learning will occur(Jacobson, 1994; Jacobson et al., in press; Jacobson & Spiro, 1995). As wenoted above, research is beginning to identify the theoretical and designcharacteristics of effective hypertextual learning environments, such as:active, nonlinear student exploration of hyperlinked information nodes, studentcreation of new hyperlinks in existing materials, explicit depiction ofimportant interrelationships between surface and structural knowledgecomponents across multiple case examples, and student authorship of hyperlinkedmaterials (Beeman et al., 1988; Jacobson et al., 1994; Jacobson et al., inpress; Jacobson & Spiro, 1995; Jonassen & Wang, 1993; Lehrer, 1993;Shapiro, 1994).
We are investigating the application of research findings such as these to thedevelopment of a network-based Learning Resource Server. The server, currentlyunder development, contains a large hyperlinked knowledge-base of educationprojects, curriculum units, research reports and papers, and links to othernetwork servers with documents relevant to education. Figure 5 shows a samplescreen from the UIUC Learning Resource Server. Overall, the UIUC LearningResource Server is intended to provide a shared knowledge space that can beutilized by a wide range of learners and researchers. We are also exploringhow tools for constructing hyperlinked personal knowledge spaces, such as theMessage Assistant, can be used by students in conjunction with the sharedknowledge space of the Learning Resource Server.

Figure 5. Home page of the University of Illinois atUrbana-Champaign Learning Resource Server (URLhttp://www.ed.uiuc.edu/lrs/).
In this paper, we have discussed conceptual frameworks for systematicallydeveloping software tools for network-based learning environments that can helpstudents and teachers construct personal and shared Knowledge Spaces. We havedescribed a software tool based on these conceptual frameworks, the MessageAssistant, that allows a user to construct a personal knowledge space. We havealso sketched out our initial efforts at applying similar concepts and tools tothe design of Learning Resource Servers that can be used to constructhyperlinked shared Knowledge Spaces.
For network learning environments to have a substantial and positive impact oneducation, "ease-of-use" and "universal access" are not enough. Students andteachers need conceptual frameworks to help organize their activity, they needtools that are consistent with such frameworks, and they need mediators toenable the activity. This paper describes some theoretical and research stepstoward these goals.
Beeman, W. O., Anderson, K. T., Bader, G., Larkin, J., McClard, A. P.,McQuillan, P. J., & Shields, M. (1988). Intermedia: A case study ofinnovation in higher education (Final report to the Annenberg/CPB Project).Providence, RI: Brown University, Office of Program Analysis, Institute forResearch in Information and Scholarship.
Brown, J. S., Collins, A., & Duguid, P. (1989). Situated cognition and theculture of learning. Educational Researcher, 18(1), 32-42.
Bruce, B. C., & Peyton, J. K. (1992). A situated evaluation of computernetworking to teach writing (Technical Report No. 565). Urbana: Universityof Illinois, Center for the Study of Reading.
Bruer, J. T. (1993). Schools for thought: A science of learning in theclassroom. Cambridge, MA: The MIT Press.
Clancey, W. J. (1993). Situated action: A neuropsychological interpretationresponse to Vera and Simon. Cognitive Science, 17, 87-116.
Conklin, J. (1987). Hypertext: An introduction and survey. IEEEComputer, 20(9), 17-41.
Edelman, G. (1992). Bright air, brilliant fire: On the matter of themind. New York: Basic Books.
Feltovich, P. J., Spiro, R. J., & Coulson, R. L. (1989). The nature ofconceptual understanding in biomedicine: The deep structure of complex ideasand the development of misconceptions. In Evans, D., & Patel, V. (Eds.),The cognitive sciences in medicine (pp. 113-172). Cambridge, MA: MITPress (Bradford Books).
Greeno, J. G., & Moore, J. L. (1993). Situativity and symbols: Response toVera and Simon. Cognitive Science, 17, 49-59.
Hunter, B. (1992). Linking for learning: Computer-and-communications networksupport for nationwide innovation in education. Journal of Science Educationand Technology, 1(1), 23-34.
Jacobson, M. J. (1994). Issues in hypertext and hypermedia research: Toward aframework for linking theory-to-design. Journal of Educational Multimediaand Hypermedia, 3(2), 141-154.
Jacobson, M. J., Maouri, C. , Mishra, P., & Kolar, C. (in press). Learningwith hypertext learning environments: Theory, design, and research. Journalof Educational Multimedia and Hypermedia.
Jacobson, M. J., & Levin, J. A. (1992). A rule-based electronic mailprocessor for collaborative electronic learning environments. Paperpresented at the annual meeting of the American Educational ResearchAssociation, San Francisco, CA.
Jacobson, M. J., & Levin, J. A. (1993, December). Hypertext andnetwork-based learning environments: Technology for the construction ofpersonal and shared knowledge spaces. Paper presented at the InternationalConference on Computers in Education: Applications of Intelligent ComputerTechnologies, Taipai, Taiwan.
Jacobson, M. J., & Levin, J. A. (1993). Network learning environments andhypertext: Constructing personal and shared knowledge spaces. In Foster, D.,& Jolly, D. V. (Eds.), Proceedings of Tel-Ed `93 (pp. 190-197).Dallas, Texas: International Society for Technology in Education.
Jacobson, M. J., & Spiro, R. J. (1995). Hypertext learning environments,cognitive flexibility, and the transfer of complex knowledge: An empiricalinvestigation. Journal of Educational Computing Research, 12(5),301-333.
Jonassen, D. H., & Wang, S. (1993). Acquiring structural knowledge fromsemantically structured hypertext. Journal of Computer-basedInstruction, 20(1), 1-8.
Lave, J., & Wenger, E. (1991). Situated learning: Legitimate peripheralparticipation. Cambridge: Cambridge University Press.
Lehrer, R. (1993). Authors of knowledge: Patterns of hypermedia design. InLajoie, S. P. & Derry, S. J. (Eds.), Computers as cognitive tools(pp. 197-227). Hillsdale, NJ: Lawrence Erlbaum.
Levin, J., & Jacobson, M. (1993). Educational electronic networks andhypertext: Constructing personal and shared knowledge spaces. Paperpresented at the annual meeting of the American Educational ResearchAssociation Conference, Atlanta, Georgia.
Levin, J. A., & Jacobson, M. J. (1992). Towards a distributed networklearning framework: Theory and technology to support educational electroniclearning environments. Proceedings of the Fourteenth Annual Conference ofthe Cognitive Science Society (pp. 927-932). Hillsdale, NJ: Erlbaum.
Levin, J. A., Riel, M., Miyake, N., & Cohen, M. (1987). Education on theelectronic frontier: Teleapprentices in globally distributed educationalcontexts. Contemporary Educational Psychology, 12, 254-260.
Malone, T. W., Grant, K. R., & Turbank, F. A. (1986). The informationlens: An intelligent system for information sharing in organizations (CISCWP Report No. 133). Boston, MA: Center for Information Systems Research.
Newman, D., Goldman, S. V., Brienne, Jackson, I., & Magzamen, S. (1989).Peer collaboration in computer-mediated science investigations. Journal ofEducational Computing Research, 5(2), 151-166.
Norman, D. (1993). Cognition in the head and in the world: An introduction tothe special issue on situated action. Cognitive Science, 17(1),1-6.
Norman, D. A. (1988). The design of everyday things. Reading, MA:Addison-Wesley.
Norman, D. A., & Draper, S. W. (Eds.). (1986). User centered systemdesign: New perspectives on human-computer interaction. Hillsdale, NJ:Lawrence Erlbaum.
Riel, M., & Levin, J. A. (1990). Building electronic communities: Successand failure in computer networking. Instructional Science, 19,145-169.
Rumelhart, D. E., & Norman, D. A. (1978). Accretion, tuning andrestructuring: Three modes of learning. In Cotton, J. W. & Klatzky, R. L.(Eds.), Semantic factors in cognition (pp. 37-53). Hillsdale, NJ:Lawrence Erlbaum.
Ruopp, R., Gal, S., Drayton, B., & Pfister, M. (1993). LabNet: Toward acommunity of practice. Hillsdale, NJ: Lawrence Erlbaum.
Shapiro, A. M. (1994). Complex concept acquisition and the representation ofknowledge: A study in hypermedia-aided instruction. Paper submitted forpublication.
Spiro, R. J, Feltovich, P. J., Coulson, R. L., & Anderson, D. K. (1989).Multiple analogies for complex concepts: Antidotes for analogy-inducedmisconception in advanced knowledge acquisition. In Vosniadou, S., &Ortony, A. (Eds.), Similarity and analogical reasoning (pp. 498-531).Cambridge, MA: Cambridge University Press.
Spiro, R. J., Feltovich, P. J., Jacobson, M. J., & Coulson, R. L. (1992).Cognitive flexibility, constructivism, and hypertext: Random access instructionfor advanced knowledge acquisition in ill-structured domains. In Duffy, T. M.,& Jonassen, D. H. (Eds.), Constructivism and the technology ofinstruction: A conversation (pp. 57-75). Hillsdale, NJ: Lawrence Erlbaum.
Spiro, R. J., Vispoel, W. P., Schmitz, J. G., Samarapungavan, A., &Boerger, A. E. (1987). Knowledge acquisition for application: Cognitiveflexibility and transfer in complex content domains. In Britton, B. K., &Glynn, S. M. (Eds.), Executive control processes in reading (pp. 177-199).Hillsdale, NJ: Lawrence Erlbaum.
Suchman, L. (1987). Plans and situated actions: The problem of human-machinecommunication. Cambridge: Cambridge University Press.
Wittgenstein, L. (1953). Philosophical investigations. New York:Macmillan.
This material is based upon work supported by the United StatesNationalScience Foundation under Grant No. RED-9253423. The United States governmenthas certain rights in this material. Any opinions, findings, and conclusionsor recommendations expressed in this material are those of the authors and donot necessarily reflect the views of the National Science Foundation. Thiswork was supported by a grant of equipment from Apple Computer, Inc. Portionsof this paper have been revised from earlier conference presentation papers onthis research. The authors acknowledge the contributions ofYoungcook Jun andYasuhiro Uno for their programming work on the Message Assistant. We alsothankMatthew Stuve,Pia Bombardier,andEvangeline Secaras for theirassistance on establishing the University of Illinois College of EducationLearning Resource Server.NOTES
原文链接：http://www.ed.uiuc.edu/tta/papers/JL_EdTele/