Learning Technologies in Support of Self-DirectedLearningGerhard Fischer, Eric ScharffAbstract:Self-directed learning is a continuous engagement in acquiring, applying and creatingknowledge and skills in the context of an individual learner’s unique problems. Effectivelysupporting self-directed learning is one of the critical challenges in supporting lifelong learning.Self-directed learning creates new challenging requirements for learning technologies. Domain-oriented design environments address these challenges by allowing learners to engage in their ownproblems, by providing contextualized support, and by exploiting breakdowns as opportunitiesfor learning.Economies of educational knowledgeconstitute an emerging concept in which communitiescontribute toward the creation of information repositories, which can be reused and evolved byall members of the community for the creation of new environments. We argue anddemonstrate that domain-oriented design environments can serve as models for theseeconomies, that a software reuse perspective provides us with insights into the challenges thesedevelopments face, and that the creation and evolution of these economies are best understoodas problems in self-directed learning.Keywords: Self-directed learning; lifelong learning; domain-oriented design environments;economy of educational knowledge; reuse; seeding, evolutionary growth, reseedingDemonstrations: A demonstration of the WebNetsystem described in this article can be found at 1.IntroductionThe previous notion of a divided lifetime—education followed by work—is no longer tenable.Learning can no longer be dichotomized, spatially and temporally, into a place and time toacquire knowledge (school) and a place and time to apply knowledge (the workplace). In theemerging knowledge society (Drucker, 1994), an educated person will be someone who iswilling to consider learning as a lifelong process. More and more knowledge, especiallyspecialized knowledge, is acquired well past the age of formal schooling, and in many situationsthrough educational processes that do not center on traditional schooling (Illich, 1971). Seenin this context, working, learning, and collaboration become intimately intertwined rather thanbeing three different and distinct activities.Lifelong learning has emerged as one of the major challenges for the worldwide knowledgesociety of the future. Lifelong learning has been given considerable international attention bythe European Community, which proclaimed 1996 to be the “European Year of LifelongLearning,” and by UNESCO, who has included “Lifetime Education” as one of the key issuesin planning for the future. The G7 group of industrialized nations has named “LifelongLearning” as a main strategy in the fight against unemployment. Despite this great interest,there are few encompassing efforts to tackle the problem in a coherent way. Lifelong learningcannot be investigated in isolation by looking at only one small part of it, such as K-12education, university education, or worker re-education. Lifelong learning needs to promoteeffective educational opportunities in the many learning settings through which people pass,including home, school, work, and communities.Supporting lifelong learning requires a suite of complementary approaches including intelligenttutoring systems, design environments, performance support systems, on-demand learning,coached simulation systems, intelligent help and advisory systems, and collaborative systems.Exploring these systems has been the shared objective of the East/West Consortium (seeSpohrer, et al, 1998, this issue). This article focuses on one of the most important approachesin support of lifelong learning: supported self-directed learning. Self-directed learning is criticalwhen learning becomes an integral part of life—driven by our desire and need to understandsomething or to get something done, instead of merely solving a problem given in a classroomsetting. A lifelong learning perspective implies that schools and universities need to preparelearners to engage in self-directed learning processes because this is what they will have to do intheir professional and private lives outsides of the classroom.The challenge for environments supporting self-directed learning is to allow learners to work onauthentic problems and tasks of their own choosing, and yet still provide them with learningsupport contextualized to their chosen problem. Although repositories of objects are anJournal of Interactive Media in Education, 98 (4)Page 2 Learning Technologies in Support of Self-Directed LearningFischer & Scharff important resource for designers creating artifacts within a certain domain, economies ofeducational knowledge, not just of objects, must address the self-directed learning needs of thecommunity that the economy is meant to support. These challenges, some of our approachesto them, and their illustrations in the context of educational economies of knowledge form thecore of this article. This article will first discuss in section 2 the need for and the theoretical grounding of self-directed learning. Section 3 describes domain-oriented design environments, their structure,and process models for their creation and evolution. Section 4 uses different systems, whichrepresent pieces of economies of educational knowledge (such as Gamelan, the EducationalObject Economy, and the AgentSheets Behavior Exchange), to illustrate the need to supportself-directed learning to make these environments sustainable efforts that enhance the practiceof knowledge workers. Section 5 discusses the strengths and weaknesses of different economiesof educational knowledge and articulates some specific challenges for future research.2.2.1Self-Directed LearningCurrent Theories of LearningCurrent trends in educational theory make the following fundamental assumptions aboutlearning(Resnick, 19):Learning is a process of knowledge constructionnot of knowledge recording or absorption(Harel & Papert, 1991)—and it requires environments in which learners can be active designersand contributors rather than passive consumers (Fischer, 1998). Research in end-userprogramming and end-user modifiability (Nardi, 1993) contributes toward this goal. Insystems such as AgentSheets and Hypergami (see Repenning et al. and Eisenberg andEisenberg, 1998, this issue), learners construct new knowledge through interaction with thesystem and the creation of new artifacts with these software tools. Actively constructingknowledge engages learners and emphasizes the need for learners to construct knowledge in amanner appropriate for them.Learning is highly tuned to the situationin which it takes place (Lave & Wenger, 1991;Suchman, 1987)—requiring environments that are domain-oriented and support humanproblem-domain interaction (the connection between people and the domain specific problemsthat they face) and not just human-computer interaction. In a typical activity (such as workingor playing), individuals are acting until they encounter a breakdown and they reflect about thebreakdown (Fischer et al., 1993). These breakdowns (originating from missing knowledge,misunderstandings about the consequences of actions, and so on) are key to situated learning.Journal of Interactive Media in Education, 98 (4)Page 3Learning Technologies in Support of Self-Directed LearningFischer & Scharff Schön (Schön, 1983) calls this reflection-in-action, Norman (Norman, 1993) calls it experi-ential and reflective mode. Because self-reflection is difficult, a human coach, a design critic, ora teacher can help the learner to identify the breakdown situation and to provide task-relevantinformation for reflection. In our own work, we have explored the possibility of using computa-tional critics (Fischer et al., 1993; Sumner et al., 1997) to provide some of this support whenhumans are not present. Critics support learners in their own activities by contextualizingexisting knowledge within a certain design situation. The information spaces presented and theinformation provided should be made relevant to the task at hand—something that computa-tional media can achieve, but is impossible for paper and pencil technologies.Learning isknowledge-dependent, meaning people use their existing knowledge to constructnew knowledge, requiring environments that support user-tailored information presentationssuch as differential descriptionsof new information. For example, if someone who knows MSWord wants to learn HTML, the explanations and examples provided should be different thanthose given to a learner who knows Framemaker. The cognitive models of users constructed byIntelligent Tutoring Systems (see Ritter et al., 1998, this issue) provide some support forpresenting knowledge in a form targeted to a specific user. Design critics may be used to tailorinformation so that it is relevant to the current task.Learning needs to account for distributed cognition(Norman, 1993), by which the knowledgeand effort required to solve a problem is distributed among various participants. The distri-bution of knowledge among humans is based on the “symmetry of ignorance” (Rittel, 1984) orasymmetry of knowledge between different stakeholders in problem solving. Teaching inclassrooms is often conceptualized very differently: it is often fitted “into a mold in which asingle, presumably omniscient teacher explicitly tells or shows presumably unknowing learnerssomething they presumably know nothing about” (Bruner, 1996). A critical challenge is areformulation and reconceptualization of this impoverished and misleading conception.Although this model may be more realistic for early school years (Hirsch, 1996), it is obviouslyinadequate for self-directed learning processes as they occur in lifelong learning, whereknowledge is distributed among many stakeholders and “the answer” does not exist or is notknown. Group discussions, conversations around dinner tables, and classrooms have thepotential to be places where knowledge is created and constructed by communities of mutuallearners. Distribution of knowledge is a central concept for collaboratively constructedknowledge repositories such as Gamelan and the Educational Object Economy (see Sections 4and 5).Learning is affected as much by motivational issues(Csikszentmihalyi, 1990) as by cognitiveissues—requiring environments that let people experience and understand why they shouldlearn and contribute something. For example, learning-on-demand(Fischer, 1991) lets usersJournal of Interactive Media in Education, 98 (4)Page 4Learning Technologies in Support of Self-Directed LearningFischer & Scharff access new knowledge in the context of actual problem situations and delivers informationabout which they are unaware in the context of theirproblem situations. Environments mustallow users to take pride in their contributions and be awarded for them.Learning is not limited to any discrete group of individuals. Even though educational systemsdeal with “teachers” and “learners” as separate groups, in reality these labels are not universallyapplicable. As tasks and responsibilities change, all individuals must be continuously learning.For example, educators who are traditionally “teachers” may desire to investigate newtechnologies to use in their classrooms. However, when exposed to new technology andmethodologies for creating educational curricula, teachers become learners in order tounderstand how to use new opportunities effectively.2.2Self-Directed Learning: Beyond the “Gift Wrapping”Approach of New MediaOne of the major misunderstandings in our current debate about enhancing learning with newmedia is the assumption that technological advances will, by virtue of their very existence,improve the quality of learning. New technologies and media must be more than add-ons toexisting practices. New technologies and learning theories must together serve as catalysts forfundamentally rethinking what learning, working, and collaborating can be and should be inthe next century.A major finding in current business reengineering efforts is that the use of informationtechnology had disappointing results compared to the investments made in it (Landauer, 1995).A detailed causal analysis for this shortcoming is difficult to obtain, but it is generally agreedthat a major reason is that information technologies have been used to mechanize old ways ofdoing business, rather than fundamentally rethinking the underlying work processes andpromoting new ways to create artifacts and knowledge.We claim that a similar argument can be made for current uses of technology in education: it isoften used as an add-on to existing practices rather than a catalyst for fundamentally rethinkingwhat education should be about in the next century. For example, the “innovation” of makingtransparencies available on the Web rather than distributing copies of them in a class takesadvantage of the Web as an electronic information medium. This may change the economicsof teaching and learning, but it contributes little to introducing new epistemologies. Oldframeworks, such as instructionism, fixed and “balkanized” curricula, memorization, decontex-tualized rote learning, and so forth, are not changed by technology itself. This is true whetherwe use computer-based training, intelligent tutoring systems, multimedia presentations, or theWeb. Journal of Interactive Media in Education, 98 (4)Page 5Learning Technologies in Support of Self-Directed LearningFischer & Scharff In the “gift-wrapping”approach, technology is merely wrapped around old frameworks foreducation. What is needed instead are richer conceptual frameworks, leading not just to theaddition of technology to existing practices, but to the exploration of fundamentally newpossibilities and limitations of computational media on how we think, create, work, learn, andcollaborate. To move beyond “gift-wrapping” in the context of self-directed learning leads tothe following requirements for computational environments. Such systems must:•be simultaneously user-directedand supportive, i.e., the choice of tasks and goals(including the learning opportunities offered) must be under the control of theuser/learner, and the support provided by the system must be contextualized to theuser’s task;be sufficiently open-endedand complexthat users will encounter breakdowns. Thesystem must provide means for allowing users to understand, extricate themselvesfrom, and learn from these breakdowns;provide means for significant modification, extension, and evolution by users; support arange of expertise, because such systems will be employed over long periodsof time by their users and must be able to accommodate users at progressively differentlevels of expertise; must promote collaborationby supporting people to overcome the symmetry ofignorance and allow stakeholders to learn from each other and create mutualunderstanding.••••3.Computational Support for Self-Directed LearningCreating computational environments in support of self-directed learning prohibits designersfrom completely anticipating and determining the use context (as is done in intelligent tutoringsystems), because this context is only partially known at design time. Figure 1 differentiatesbetween two stages in the design and use of an artifact. Often, system developers createenvironments and tools, including help systems, guided tours, forms, etc., where they have triedto anticipate at design time the situational contexts and tasks users will be engaged in. For printmedia, a fixed context has to be decided at design time, whereas for computational media, thebehavior of a system at use time can take advantage of contextual factors (such as backgroundknowledge of a user, the specific goals and objectives of a user or the work context) only knownat use time(Fischer et al., 1993). The fundamental difference is that computational media haveinterpretive power: they can analyze and critique the artifacts created by users—and users actingJournal of Interactive Media in Education, 98 (4)Page 6Learning Technologies in Support of Self-Directed LearningFischer & Scharff as designers will create artifacts of all kinds. The challenge is to create new innovative systemcomponents that allow users to articulate these contextual factors.3.1Contrasting Different Computational Approaches toSelf-Directed LearningIntelligent tutoring systems(Wenger, 1987) represent a teacher- or system-driven approachtoward learning in which the problem or the task is determined at design time. At use time, thesesystems use their intelligence to individualize instruction with user modeling techniques andsupport the learning process by providing feedback to users’ solutions, visualizations, andsimulations. In general, users of a tutoring system are learners who interact with informationand solve problems previously provided by teachers at design time. (Of course, the creators ofthe tutoring system were themselves learners when creating the tutoring system, but theiractivities of self-directed learning are not captured within the system).Interactive learning environments(Papert, 1980) are learner-driven by providing powerfulprogramming environments, often enriched with domain-specific abstractions and microworldsupport (Repenning & Ambach, 1997; Resnick, 1996) that enable learners to tackle complexproblems. But during use time, they provide little support when the learner gets stuck, offeronly limited feedback on the artifacts created, and have restricted access to information spacesbehind the artifact (such as design rationale or a catalogue of related solutions). Thus, users ofthese systems must act as teachers and learners at the same time. Without capturing, selectivelypresenting, or sharing information, users are individually completely responsible forconstructing and reflecting upon information.Domain-oriented design environments model domains but not individual tasks within thedomain. These environments are intermediate between the other two approaches by providinga more distributed approach to interacting with domain problems. They face the challenge to(at least) partially “understand” the activity in which the learner is engaged. A framework forsolving problems within a domain is provided at design time, and learners create artifacts oftheir choosing in a self-directed fashion (and extend the domain framework) at use time.Sources that are used to provide domain-specific assistance include: (1) the focus on the domain,(2) the partial construction of an artifact, (3) the partial specification provided by a learner, and(4) the information spaces visited (Fischer & Nakakoji, 1994).3.2DODEs: Environments for Self-Directed LearningDomain-oriented design environments (DODEs) (Fischer, 1994) address the problem of self-directed learning by providing concrete learning support within a particular domain. UsersJournal of Interactive Media in Education, 98 (4)Page 7 Learning Technologies in Support of Self-Directed LearningFischer & Scharff learn as they design artifacts following their unique interests and needs. Instead of attemptingto incorporate knowledge about specific problems at design time, DODEs provide a frameworkthat allows users to construct many different problems withina domain in which they areinterested at use time. Example domains that we have explored in the past include kitchendesign (Fischer et al., 1993), graphical user interface layout (Fischer et al., 1990a), telephonevoice messaging systems (Sumner, 1995), and Local Area Network design (Fischer et al., 1992).We believe that some of the techniques used in the development of DODEs are importantingredients, comprising economies of educational knowledge as well as being effective tools forsupporting such economies. Figure 1: Design and Use TimeFigure 2 illustrates a prototype for a DODE for Local Area Network (LAN) design 1. Incontrast to general-purpose environments, DODE components are instantiated in theframework of the given domain. Components are expressed in terms of domain concepts, andinformation is presented in the context appropriate for that domain. DODEs provide specificfunctionality for manipulating, exploring, and communicating about domain entities. Thecomponents of a DODE (illustrated by the numbered elements in Figure 2) include thefollowing:•A specification component (4) allows the specification of design constraints and goalsby users during use time. This provides the system with a more specific understandingabout particular tasks at hand and enables the system to offer guidance and suggestionsrelevant to those situations.WebNet is a prototype DODE for Local Area Network (LAN) design. •Domain-specific components (2) contain the individual elements that designers use tocreate design scenarios. These components, which can be shared and modified,provide a language of concrete items that augment communication through a domain.Critiquing mechanisms (not shown) represent the accumulated “wisdom” of a designcommunity, such as criteria and rules about what constitutes “good” design. Critiquing mechanisms monitor design actions, provide explicit and task-relevantfeedback, and identify breakdowns in a design developed by users at use time. Thisfeedback leads to opportunities for self-directed learning.Organizational and artifact memories (1) support the capture of design rationale andargumentation embedded within design artifacts. Embedding rationale within designsallows users to explore the rationale behind specific parts of an artifact and to use theartifact as a tool to ground domain communication.Case libraries (5) allow reuse at a higher level of granularity than individualcomponents. They facilitate a different kind of reuse and design-by-modificationmethodologies by modifying previously constructed artifacts.Simulation mechanisms (3) support users in understanding the behavior of acomponent or a complete artifact.••••The combination of approaches used in the DODE framework go beyond the “gift-wrapping”approach and address the challenges faced in supporting self-directed learning. Specifically:•Orienting use of the system through construction of objects within a domain gives theuser choice over what artifact to construct (and grounds learning within authenticactivity). Construction is user-directed, and critics support situated learning byanalyzing the user’s current construction.Employing design criticsthat can analyze different situations creates an environmentin which users will be informed of problems as they arise. These breakdown situationsprovide opportunities for users to understand and learn from the breakdown withinthe context of their personal design activities.•Journal of Interactive Media in Education, 98 (4)Page 9 Learning Technologies in Support of Self-Directed LearningFischer & Scharff Figure 2: Example of a Domain-Oriented Design Environment for Computer Network Design•Supporting different levelsof modification (including changing organizationalmemories, adding new components, creating new cases for the case library, andaltering the domain simulation) provide support for a wide range of expertise,providing the ability to augment the system when necessary but not requiring users todo so. Applying different sets of critics and using specification components alsosupports different levels of expertise by providing users with different schemas foranalyzing their domain-centric activities.Providing mechanismsfor evolution such as case libraries and organizational memoriesallows users to easily add knowledge to the existing framework. Simulations andcritiquing systems serve as the driving support for evolution. End-user modificationmechanisms are built to facilitate changes using domain concepts, minimizing theneed for a shift from domain activity to changing low-level system implementation.•Journal of Interactive Media in Education, 98 (4)Page 10 Learning Technologies in Support of Self-Directed LearningFischer & Scharff •Encouraging collaboration is inherent as users working within the same domain caneasily share their new ideas and changes. Individual components, catalog examples,and rationale about designs are not static entities in DODEs. As users interact with adomain oriented environment, they create and compose new artifacts that themselvesbecome part of the system. Therefore, components in a DODE are specificallydesigned to handle an ever-changing domain space to cope with the constant flux ofproblems in the real world.3.3The Seeding, Evolutionary Growth, and ReseedingProcess ModelMost intelligent systems (including systems in support of learning such as Intelligent TutoringSystems (Wenger, 1987), Expert Systems (Stefik, 1995), and Simulation Environments such asSimCity 2) have traditionally been developed as “closed” systems. The basic assumption wasthat during design time, a domain could be modeled completely by bringing domain experts(designers) and environment developers (knowledge engineers) together and the knowledgeengineers would acquire the relevant knowledge from the domain experts and encode it into thesystem. This approach fails for the following reasons: (1) much knowledge is tacit and onlysurfaces in specific problem situations; and (2) the world changes, and intelligent systems thatmodel this world must change accordingly. Thus, closed systems are inadequate to cope withthe tacit nature of knowledge and the situatedness of real-world problems. DODEs are designed as “open” systems, where opportunities for change are built in as a centralpart of the system. Tools and techniques developed for DODEs present important examplesand highlight specific challenges for supported self-directed learning environments. Byproviding components that incrementally evolve, these environments allow a constant flow ofinput from designers and users, bridging the gap between “design time” and “use time.”Although complex systems must evolve in order to be effective (Simon, 1996), it may not beclear how to conceptualize the changes that will take place over time in these systems. In ourresearch, we have developed the Seeding, Evolutionary Growth, and Reseeding process model(Fischer et al., 1994) to address these problems. This model has been explored in our work withDODEs and postulates three major phases: A seedwill be created through a participatory design process (Henderson & Kyng, 1991)between environment developers and domain designers. A seed is not a fully realized systembut instead is a sufficiently expressive entity that can be used to address some specific real-worldproblems. Mechanisms for evolution must be built into these initial conceptualizations.Postulating a seed as an objective (rather then a complete domain model or a complete2SimCity is a registered trademark of Maxis, Inc. See knowledge base) sets this approach apart from other approaches in intelligent systemsdevelopment and emphasizes evolution (Dawkins, 1987; Popper, 1965) as the central designconcept.Evolutionary growthtakes place as individuals use the seeded environment to undertakespecific projects. During these design efforts, new requirements may surface, new componentsmay come into existence, and additional design knowledge not contained in the seed may bearticulated. During the evolutionary growth phase, the environment developers are not present,making end-user modification(Girgensohn, 1992; Nardi, 1993) a necessity rather than a luxury.End-user programming both supports learning (by leading to the creation of a new computa-tional artifact) and requires learning (in order to be able to create the new artifact). Figure 3illustrates this dual relationship between learning on demand from a system and thesimultaneous need to add knowledge to the system. Figure 3: The Duality between Learning on Demand and End-User Modifiability in SelfDirected LearningReseeding, a deliberate effort of revision and coordination of information and functionality,brings the environment developers back to collaborate with domain designers to organize,formalize, and generalize knowledge added during the evolutionary growth phases. Informationabout how the system has evolved is essential in determining how the system must be reconcep-tualized. By looking at the system evolution, it is possible to postulate which extensions createdfor specific design projects should be incorporated into future versions of the generic designenvironment. Drastic and large-scale changes occur during the reseeding phase.Journal of Interactive Media in Education, 98 (4)Page 12 Learning Technologies in Support of Self-Directed LearningFischer & Scharff The SER model is a useful framework for understanding the processes inherent in thedevelopment of open systems. For example, the development of open-source software systemssuch as the Linux operating system (Raymond, 1998) provides an interesting existence proofthat reliable, useful, and usable complex systems can be built in a decentralized “Bazaar style”by many, rather than in a centralized, “Cathedral style” by a few. The Linux development modeltreats users simultaneously as co-developers and self-directed learners (see Figure 3) and iscurrently being tested in a number of new areas (e.g., in the Netscape Communicator 3 ).Gamelan, the Educational Object Economy, and the AgentSheets Behavior Exchange are threeexamples of open systems that are forming incipient economies of educational knowledge, wewill now analyze using the SER framework .4.Economies of Educational KnowledgeEconomies of educational knowledge (EoEK) address the problem of self-directed learning byleveraging the knowledge of the (world-wide) community to promote knowledge dissemination.One might consider the entire Web to be an EoEK in which a distributed community presentspublicly available information on any subject of interest to the contributors. However, the mereexistence or availability of information does not imply that this information will be useful.Web-based EoEKs tend to focus on supporting certain kinds of knowledge. The idea of usinga specific domain (and supporting a community of practice) that has emerged in successfulEoEKs has been a central concept in the DODE framework for some time. The idea for economies based on the interchange of educational knowledge is not new. Morethan 25 years ago, Illich (Illich, 1971) introduced the concept of “learning webs,” a scheme fortransforming the creation and dissemination of knowledge into a problem in which all peopleplay an important role. Illich envisioned a world in which the mass distribution capabilities ofthe currently extant technology could be used to facilitate access to and sharing of information.Believing that people are capable of being both teachers and learners depending on the circum-stances, Illich envisioned an economy that encouraged people to become active teachers andproducers of educational knowledge as a result of self-directed learning activities.The rapid growth and increasing ubiquity of the Web have made it a popular vehicle forestablishing educational knowledge repositories. Three informative examples of current reposi-tories are Gamelan, the Educational Object Economy, and the AgentSheets Behavior Exchange.All three are Web-based systems that share a common goal: to support the learning needs of acommunity of software developers. Gamelan is a resource for information about Javatechnology, the Educational Object Economy provides computational resources for designingeducational software, and the Behavior Exchange allows AgentSheets simulation developers to3 Mozilla.org is a group within Netscape that is chartered to act as a clearing-house for theNetscape source. Learning Technologies in Support of Self-Directed LearningFischer & Scharff share simulations and simulation components. However, as our analysis will show, thesesystems do not adequately support the interchange of educational knowledgeper se. Repositoriesalone leave it up to learners to frame information in a way appropriate to their problems andself-directed learning activities. In their current state, they provide little support for locating,comprehending, and modifying information relevant to the task at hand (see Figure 8). Thus,while community repositories of information are certainly key components, the followinganalysis will highlight additional necessary components for a sustainable EoEK.4.1GamelanOne prime example of a first step toward an EoEK based on community participation is takingform in the software design community. Java developers use the Web to facilitate the learningof Java and to share components created in Java. Due to the contributions of developers aroundthe world, the Java programming community has used community repositories of knowledge toproduce technical advances in a very short period of time. Gamelan4is one of the firstcommunity repositories for Java-related information. The primary users of Gamelan are Javadevelopers looking for information about what other people are doing with Java. Gamelan istherefore a forum to facilitate the self-directed learning of members of the emerging Javacommunity. The software developers who use the content are also the primary contributors,continuously adding new resources to the Gamelan repository. The thousands of developerswho contribute to the Gamelan repository and the estimated thousands who search forinformation in Gamelan every day provide evidence that the Java community has taken a greatdeal of interest in using community repositories to share information.Gamelan was originally designed to be the official clearinghouse for all third-party uses of Java,and the site attempts to support any work that uses Java. Although this encyclopedic coverageattracts developers with many interests, such coverage does not make Gamelan the idealeducational tool. If one considers Gamelan an educational tool used to educate the communityof Java developers, the system does not provide a great deal of support for the learning activitiesthat developers must go through to use resources.4Gamelan was the original repository for Java components. It has grown substantially andnow encompasses resources not only about ava but also other Web developmenttechnologies. The expanded system (of which Gamelan is now a subset) is calledDeveloper.com Figure 4: Browsing Resources Returned by GamelanGamelan resources belong to one or more categories, which are organized hierarchically.Resources are retrieved by browsing through categories or by searching the titles and briefabstracts of resources. Figure 4 shows a keyword search for network visualization andmanagement tools that resulted in a pointer to two Java categories, in this case the “Java ->Network and Communications -> Network” and “General” categories. Retrieved componentsappear in unsorted order. All the descriptions are provided by the contributors of theinformation. The only exceptions are the one-word annotations (marked with fans), which wasadded by the repository administrators. Gamelan does not provide any mechanisms for refiningqueries or organizing returned results.Journal of Interactive Media in Education, 98 (4)Page 15 Learning Technologies in Support of Self-Directed LearningFischer & Scharff 4.2Educational Object EconomyThe Educational Object Economy(EOE) 5(see Spohrer, et al., 1998, this issue) provides amore focused system than Gamelan. Currently realized as a collection of Java objects (mostlycompleted applets) designed specifically for education, the target users of the EOE are teachers(presumably acting as consumers of completed applets) wishing to use new interactivetechnology and instructional designers interested in producing educational software. TheEOE’s primary goal is to provide educators with a collection of useful resources ready to be usedto help students learn. There is an interesting dichotomy apparent in the EOE. Educators wishto create tools that will facilitate the learning of their students, but the teachers are actually thelearners as they search for useful components in the EOE repository.The EOE has two major access mechanisms: a single level (but somewhat hierarchical) classifi-cation scheme based on the Dewey Decimal System, and a keyword or advanced search basedon object descriptions.Figure 5: Search Mechanism for the Educational Object Economy5The Educational Object Economy, EOE Foundation Figure 5 shows the search mechanism for the EOE. The EOE may be queried according toURLs and also by using keywords in the components descriptions. The meta-data6currentlyassociated with EOE objects include the names, subjects, and descriptions in the left columnand the submission dates and source code availability information in the right column.However, the “content-specific” query information must be specified with the predefinedsubject and keyword matches in the description.Figure 6 shows the result of the query for (computer) network tools (Querying “ComputerNetwork” yielded no matches, and the computer example shown seems to be the onlycomputer-network related tool in the repository). Like the Dewey Decimal System, there is aconfusion about whether computer networks belong in the engineering (621) or ComputerScience (004) section. The one relevant EOE component falls only in the latter category.Because of the large number of Computer Science applets, manual browsing is prohibitive.Finding the right category or knowing how to appropriately categorize items is a large problemin both Gamelan and the EOE. This classic vocabulary problem(Furnas et al., 1987) is notunique to computer repositories. Objects matching the search specification are returned intabular form. There is no incremental refinement of queries. As with Gamelan, the repositoryacts more as an index, than as a storage location. The repository stores abstracts and meta-data,but the artifacts themselves are maintained on the personal pages of contributors. Users mustgo to individual web sites for further information about the artifacts.A collection of educational resources is a valuable tool for creating systems that support self-directed learning. However, the act of locating and using educational objects is itself a self-directed learning activity that can benefit from specific support mechanisms. Thus, teachersand creators of educational resources can benefit from supported self-directed learningtechniques as much as learners or “end-users” of educational resources. The EOE supports theexchange of domain-specific components for education, but falls short of supporting the self-directed learning needs of educational designers who need to locate useful components,understand how components work, and modify the retrieved components to fit a specificcontext. 6The IMS Meta-data Information Site Figure 6:Results of a Query for Networks Returned by the EOE4.3Behavior ExchangeThe AgentSheets Behavior Exchange7(Repenning et al., this issue) is an initial prototype of adomain-specific system for sharing computational artifacts. It is a repository that stores agents(entities that can perform computation) created using the AgentSheets system. Like Gamelanand the EOE, the Behavior Exchange consists of a collection of computational artifacts andsome “meta-data” specific to each component. The system’s focus on agents (and sharingcomputation by exchanging agents) allows the system to present information in a mannerappropriate for AgentSheets programmers. The Behavior Exchange was designed to support theneeds of a specific audience who use the AgentSheets system to build simulations. These end-user programmers include college students, primary school students, and professionals in a7The Agentsheets Behavior Exchange variety of domains such as environmental design and biology.Figure 7 shows a sample screen from the Behavior Exchange in the domain of “animal agents.”Components are automatically added to the Behavior Exchange using an uploading mechanismbuilt into the AgentSheets environment. This uploading process combines formal informationfrom AgentSheets with informal information such as categories and descriptions provided byusers. The system then uses these forms of information to provide various ways of presentinginformation in the repository. Multiple categorization schemes (such as projects and categories)are automatically maintained by the system. Agents may be subsequently sorted by name ormodification date. Searches may be refined quickly using keyword mechanisms and formulatingsets of target categories and projects. The information that can be automatically synthesizedfrom the formal agent definition is combined with information contributed by users to providea perspective relevant for simulation builders. Once a designer has found an agent that couldbe useful in a new simulation, this agent can be dragged directly from the Behavior Exchangeinto AgentSheets running on the user’s local computer for immediate use (see Repenning, et al,1998, this issue).Figure 7: Retrieving Components with the Behavior ExchangeJournal of Interactive Media in Education, 98 (4)Page 19 Learning Technologies in Support of Self-Directed LearningFischer & Scharff 4.4Summary: Economies of Educational Knowledge asOpen SystemsOne important feature common to all three of these systems is their support for evolution. Asnew educational knowledge becomes available, members of the community may share newdevelopments with each other. In all three systems, the repository administrators set up aninitial seedthat structures how information is added, presented, and searched by users. The goalis to create useful information repositories in a decentralized fashion (Resnick, 1994). All threesystems allow evolutionby the community who uses the information, although the centralizedauthority plays different roles in the different systems. In Gamelan (and to a lesser extent, theEOE), all new resources are verified and filtered by the repository administrator. In theBehavior Exchange, contributions are unrestricted. In fact, contributors have some control overthe categorization itself because of the ability to add new projects or categories. Evolution inall the repositories is limited to the additionof new content. The ability to refinecontent islimited in all the systems. Because users can add only new resources and all resources stand ontheir own, it is impossible to track the changes of individual components. Finally, over the pastfew years, all three systems have gone through dramatic redesign or reseedingphases in whichcontent is checked and reformulated and revised, entries are related to each other (which mightpossibly have been captured during the evolution phase), categorization schemes change,information presentation goes through major changes, and different searching methods areemployed. In all three cases, reseeding has been performed by the environment developersbased on feedback from the community.Because all three systems are envisioned as tools that evolve at the hands of a community ofusers, all three are prime candidates to study the challenges, strengths, and weaknesses of opensystems. The SER process model that accounts for the evolution of DODEs is a usefulframework for understanding the changes that have taken place in these systems. Table 1summarizes the three systems and how they can be understood using the SER framework.Although most of the repositories were designed for evolution, the seeding and reseeding phaseshave also played important roles in their development. Having an understanding of the processmodel beforehand may be beneficial; for example, Gamelan went through a radical reconcep-tualization when the designers switched from using a single category to using multiplecategories for a resource. This change was in large part due to the inflexibility of the catego-rization scheme - because users could not change categories and because existing categories wereambiguous or insufficient, the existing structure quickly became difficult to navigate. Becauseall changes are eventually handled by Gamelan staff, the evolution that the community couldprovide instead became the responsibility of the site managers. As illustrated by the other twosystems, the more that users become ‘co-developers’, the more the repository begins to resemblean idealized EoEK.Journal of Interactive Media in Education, 98 (4)Page 20 Learning Technologies in Support of Self-Directed LearningFischer & Scharff Table 1: A SER Perspective on Gamelan, the EOE, and the Behavior Exchange5.Challenges for Economies of Educational Knowledge “Hit counts,” the number of times a certain resource has been accessed, seem to be the mostcommon way to evaluate the efficacy of Web-based information repositories. The success ofsystems such as Gamelan are usually given with database usage statistics. Frequentlyencountered database statistics include the number of people who visit a Web site or the numberof resources indexed in a repository. Although these popularity statistics are important, neitherof these metrics provides an insight into how well these systems truly support users engaged inself-directed learning activities. This is not to say that such database resources are worthless—Journal of Interactive Media in Education, 98 (4)Page 21 Learning Technologies in Support of Self-Directed LearningFischer & Scharff a centralized repository where software developers can find resources has proved to be invaluablefor Java developers. Warehouse repositories are simply insufficient tools to support a learningcommunity.Instead, we must analyze the specific needs of users of an EoEK and use technology to helpaddress these needs. The primary goal of the EoEKs described above is to support softwaredevelopers acting in the context of their self directed learning activities. Designers (whether aJava programmer or an instructional designer) who want to take advantage of an existing EoEKwill be driven by their own goals and objectives, which requires, by necessity, support for self-directed learning processes. To address the self directed learning efficacy of these systems, wepresent a model for understanding the self-directed learning needs of software developers andhow the current repositories address (or might address) these needs.5.1Supporting the Location / Comprehension / Modification CycleThe long-term goal of an EoEK is thwarted by an inherent design conflict: to be useful, aneconomy must provide many building blocks, but when many building blocks are available,finding and choosing an appropriate one becomes a difficult problem. Even if an appropriateblock is found, it is rarely usable as is and usually must be modified to suit the new use context(see Roschelle, et al, 1998 for further discussion of this issue). Based on our investigations, aswell as others, we are convinced that to make an EoEK a success, substantially more is requiredthan creating objects and depositing them in a globally accessible information repository. Figure8 illustrates three essential processes as they occur in using an EoEK: location, comprehension,and modification (Fischer et al., 1991).Figure 8:A Conceptual Framework for Software ReuseJournal of Interactive Media in Education, 98 (4)Page 22 Learning Technologies in Support of Self-Directed LearningFischer & Scharff The location, comprehension, and modification cycle serves as a model for understanding theprocesses involved in reusing existing components in new situations. Since systems likeGamelan and the EOE provide repositories of components that are intended to be retrieved andreformulated, it is instructive to determine how these systems would address a user’s unique selfdirected learning needs.We’ll illustrate the self-directed learning challenge in software reuse using a specific softwaredesign task. In this example, an educator (perhaps a Computer Science professor) wishes to findtools to help students learn about the challenges in real-world computer networks. In particular,the teacher tries to find tools that simulate network behavior or visualize the communicationthat takes place between computational components. 8First, the teacher attempts to locatecomponents that could be used to construct a system to explain networks. Then, the teacherneeds to comprehendthe components, determining what each piece does and deciding how thepieces might be used together. The software components would then need to be modified foruse in configurations not anticipated during their design time and to address the specific needsof the teacher. Finally, the teacher may wish to share the newly created educational tool withthe community by adding it back into the repository. The cycle completes when anotherteacher wishes to locate a similar resource in another situation.The first step for a designer would be to locateexisting tools for simulating or visualizingnetworks. A typical “global” repository of software components such as Gamelan provides twomechanisms for locating relevant information: browsing a set of predefined categories orsearching by keywords. The top-level Gamelan categories, such as “Games,” “CommercialJava,” and “Arts and Entertainment,” may be a reasonable way of indexing the whole space ofpossible uses of Java, but they provide little guidance about how to proceed to find a specificcomponent such as a network simulation. A viable alternative would be to provide multiplecategorization schemes that present different views for browsing. For example, a designer maybrowse through genres of courses (looking perhaps for other networking courses), or differentkinds of applications (such as LAN management). Although more selective queries are possible,a simple query search of an encyclopedic database may not provide useful information; akeyword search for “network” and “visualization” in Gamelan returned no matches (particularlysurprising because there are quite a few tools for visualizing networks). Searching for “network”and “simulation” yields a few useful resources and categories (both Java -> Networking andCommunications -> Network and “General”). This example demonstrates that querymechanisms that rely exclusively on keyword matching are extremely susceptible to thevocabulary problem (Furnas et al., 1987) common in the mismatch between a “system” model8 Although we will focus in this section on the teacher’s self-directed learning, the artifactcreated by the teacher could be designed to facilitate the self-directed learning of thenetworking student. Using components to create self-directed learning environments is aninteresting idea but a full discussion is beyond the scope of this example.Journal of Interactive Media in Education, 98 (4)Page 23 Learning Technologies in Support of Self-Directed LearningFischer & Scharff (used by a system to represent information) and “situation” model (the perspective maintainedby the user in a unique situation; see section 5.2 and Kintsch (Kintsch, 1998)). In Gamelan,the user is forced to make this association without support by manually chasing down relevantcategories.Fortunately, better solutions exist. One of our prototype systems for locating code resources(Fischer et al., 1991) combined formal information (extracted from program source texts) andinformal information (in the form of on-line documentation) in order to retrieve as much usefulinformation as possible. The system also used spreading activation to help identify relevantobjects and provided a mechanism for incrementally refining queries. If the system is capableof analyzing the artifact being constructed (as it is in the case in DODEs), a partial represen-tation and a partial specification (see Figure 2) may be used as the basis for a query. The systemcould then return previous objects based on the state of the current design. We have developedsystems that use text analysis mechanisms such as latent semantic analysis (LSA) (Landauer &Dumais, 1997) to find useful information. Ideally, a developer could type a description of thedesired graph visualization system and then incrementally refine the query based on thereturned results.Once the designer finds a set of possibly relevant components, the designer must comprehendthe retrieved resources. Annotations provided by the Gamelan staff such as “well commented,”“sophisticated,” “transglobal,” and “Zowie!” offer little to aid comprehension. In fact, therepository stores little information about a resource other than a link to a Web page maintainedby the resource contributor. It is up to the contributors to present information about theartifacts in question. Although many authors provide a page that demonstrates the functionalityof a resource, usually in the form of a Java applet, these examples provided by the authors donot necessarily address the issues that the designer finds useful. It is unrealistic to assume that,with thousands of Gamelan developers, any documentation or demonstration will be written ina manner appropriate to the entire community. Instead, a developer would probably prefer aranking of relevant resources based on features important for software reuse. The designer mayuse a specification component (see Figure 2) to say that modular code is more important thanfast code, or perhaps that cross platform support cannot be sacrificed. The designer may browsecomments created by other users that are associated with the individual components. Thesediscussion forums would help link developers to the appropriate human contact points.Hopefully, the designer will find a resource that is immediately relevant to the current task, suchas a network simulator or visualization tool. Most likely, the developer will need to modifyexisting components such as a graph subsystem and a network statistic- gathering tool (both ofwhich exist in Gamelan) and compose these to create a new tool. If the resource is a JavaJournal of Interactive Media in Education, 98 (4)Page 24 Learning Technologies in Support of Self-Directed LearningFischer & Scharff component that uses JavaBeans9, the standard Java component model, it would be possible touse a tool specifically designed to compose and modify these components. Most tools forcomposition require a significant amount of Java programming knowledge and low-level Javaprogramming. Components that can be integrated or modified with domain-specificprogramming systems (see Repenninget al., this issue) may make the modification task mucheasier. Although no single tool can be ideal for every programming task, visual and domain-specific programming languages for composing larger applications from existing modules wouldfacilitate rapid system development and prototyping without requiring large low-level codemodifications. A developer may use tools to embed design decisions into the componentsthemselves. Providing the opportunity to track the evolution of individual components wouldallow future developers to see the specifics of how a system evolved and why those decisions weremade.Unfortunately, traditional repositories such as Gamelan or the EOE do not really support thenotion of modifying or refining an artifact. If the designer has augmented an existing networksimulator, there would be no way to submit the new “deluxe simulator” and associate it with thecomponents that were extended. There are a large number of components in Gamelan builtupon each other, but the relations among these components is virtually impossible to determine.Systems such as the EOE and the Behavior Exchange have recognized the need to associate“meta-data” with every component. Right now these “meta-data” do not contain informationabout the relations among components, but this kind of information would be an importantkind of data about components. The infrastructure for the economy can then track the usageof the new component. Associating artifacts with “meta-data” about the evolution (such as peerreviews, comments, or revision histories) would use the intricate relations among componentsand associated comments to provide richer information to locate and comprehend components.5.2DODEs as Models for Economies of EducationalKnowledgeIn this article, we have identified some basic challenges for self-directed learners interacting withEoEKs (acting primarily as designers of some artifacts). They have to cope with the followingchallenges in such situations; (1) they do not know about the existence of components; (2) theydo not know how to access components; (3) they do not know when to use components; (4)they do not understand the results that components produce for them; and (5) they oftencannot combine, adapt, and modify components according to their specific needs withoutfurther help or support.When designers attempt to locate components for use in a new application, they approach the9JavaBeans is the component architecture for Java. Both JavaBeans and Java areregistered trademarks of Sun Microsystems, Inc. search task with their own individual understanding of the world and their own vocabulary.The fundamental challenge of any EoEK is to bridge the gap between a user’s application goalsand vocabulary (situation models) and the repository’s implementation or categorizationschemes (system models). DODEs address this gap by providing information spaces andcomputational tools targeted at specific domains, of interest to certain communities of practices.Thus, the system model of a DODE should already reflect the situation model of the user. Ourprevious analysis of three different repositories illustrated how more-focused repositories such asthe Behavior Exchange acted more like EoEK’s than general purpose repositories such asGamelan.As previously discussed, the lifecycle of DODE’s and other open systems follows the SERmodel. The SER model is motivated by how large software systems, such as Emacs, Unix, andLinux, have evolved over time. In such systems, users develop new techniques and extend thefunctionality of the system to solve problems that were not anticipated by the system’s authors(following the observation that any artifact should be useful in the expected way, but a trulygreat artifact lends itself to uses the original designers never expected). New releases of thesystem incorporate ideas and code produced by users.Unlike these large software systems, DODEs must address an additional challenge to make theSER model feasible: whereas the people in the above-mentioned environments are computa-tionally sophisticated, DODEs need to be extended by domain designers who are neitherinterested in nor trained in the (low-level) details of computational environments (Nardi,1993). Domain designers are more interested in their current task than in maintaining aknowledge base. At the same time, important knowledge is produced during daily designactivities that should be captured. Rather than expect designers to spend extra time and effortto maintain the knowledge base as they design, DODEs provide tools to help designers recordinformation quickly and without regard for how the information should be integrated into theenvironment. Knowledge-base integration is periodically performed during the reseedingphases by environment developers and domain designers as a collaborative activity.One of our basic claims and assumptions is that most future EoEKs should and will have userswho are more like users of DODEs than users of Linux and Gamelan. Because they are neitherknowledgeable nor interested in specific programming issues, their self-directed learningactivities and their own contributions will be more at the domain level than at the programminglevel (e.g., the target users of the EOE include teachers wishing to use and contribute newinteractive technology and educational software).Table 2 is an attempt to summarize and compare Gamelan and the EOE along thesedimensions, and to derive recommendations for self-directed learning based on our work withDODEs.Journal of Interactive Media in Education, 98 (4)Page 26 Learning Technologies in Support of Self-Directed LearningFischer & Scharff Table 2: Summary and Comparison of Systems from a Self-Directed Learning ContextJournal of Interactive Media in Education, 98 (4)Page 27 Learning Technologies in Support of Self-Directed LearningFischer & Scharff 6. ConclusionIndustrial-age models of education and work are inadequate for preparing students to competein a knowledge-based workplace and to be fully empowered citizens in an ‘information society’.A major objective of our lifelong learning approach is to reduce the gap between school andworkplace learning. When learning becomes a part of life, support for self-directed learning isa necessity.New media and new technologies alone will not provide answers to the challenges that self-directed learning presents. When old processes are realized with new technologies, the possiblebenefits afforded by the new capabilities will be largely unrealized. Instead, we must rethink ourbasic assumptions and see how technology can be applied to best solve the fundamentalproblems that people encounter in actual learning situations.Facilitating economies of educational knowledge is a promising direction that supports theneeds of self-directed learners. Based on requirements for self-directed learning, we have arguedthat DODEs, support for the seeding, evolutionary growth and reseeding model, and supportfor the location, comprehension and modification cycle are necessary to create effectiveeconomies of educational knowledge. These economies must allow their designers to use themin self-directed ways, and their content must provide the components required for the creationof self-directed learning environments.In our research we have explored fundamentally new possibilities and limitations of computa-tional media as they complement existing media. The ongoing exploration of these issues willcontinue to raise important questions such as: How can sustainable environments be created forcommunities of practice? How can large complex information spaces be evolved over longperiods of time? How can self-directed learning be facilitated and supported in its importantrole for making learning a part of life?AcknowledgmentsThe authors would like to thank the members of the Center for LifeLong Learning and Design(L3D) at the University of Colorado, who have made major contributions to the conceptualframeworks and systems described in this paper. The authors received important feedback fromthe JIME editors and reviewers to improve earlier versions of this paper. The research wassupported by (1) the National Science Foundation, Grants REC-9631396 and IRI-9711951;(2) the McDonnell Foundation; (3) NYNEX Science and Technology Center, White Plains; (4)Software Research Associates, Tokyo, Japan; (5) PFU, Tokyo, Japan; and (6) Daimler-BenzResearch, Ulm, Germany.Journal of Interactive Media in Education, 98 (4)Page 28 Learning Technologies in Support of Self-Directed LearningFischer & Scharff ReferencesBruner, J. (1996). The Culture of Education. Cambridge, MA: Harvard University Press.Csikszentmihalyi, M. (1990). Flow: The Psychology of Optimal Experience. New York:HarperCollins Publishers.Dawkins, R. (1987). The Blind Watchmaker.New York - London: W.W. Norton andCompany.Drucker, P. F. (1994). The Age of Social Transformation. The Atlantic Monthly (November),53-80.Eisenberg, M. and Eisenberg, A.N. (1998) Shop Class for the Next Millenium: Education throughComputer-Enriched Handicrafts. Journal of Interactive Media in Education, 98 (8). Psychological Review, 104(2), 211-240.Lave, J., and Wenger, E. (1991). Situated Learning: Legitimate Peripheral Participation. Cambridge, UK: Cambridge University Press.GNardi, B. A. (1993). A Small Matter of Programming. Cambridge, MA: The MIT Press.Norman, D. A. (1993). Things That Make Us Smart. Reading, MA: Addison-WesleyPublishing Company.Papert, S. (1980). Mindstorms: Children, Computers and Powerful Ideas. New York: BasicBooks.Popper, K. R. (1965). Conjectures and Refutations. New York, Hagerstown, San Francisco,London: Harper and Row.Raymond, E. S. (1998). The Cathedral and the Bazaar. PAT Algebra Tutor.Journal of Interactive Media in Education, 98 (9).
因篇幅问题不能全部显示，请点此查看更多更全内容
Copyright © 2019- 517ttc.cn 版权所有 赣ICP备2024042791号-8
违法及侵权请联系：TEL:199 18 7713 E-MAIL:2724546146@qq.com
本站由北京市万商天勤律师事务所王兴未律师提供法律服务