Supporting the active learning process.

Many e-learning materials lack pedagogical principles and theoretical foundations. The great potential of activating learners and thus enriching learning experience is often unused in instructional software and online courses. Pedagogical theories like constructivist and action-orientated approaches should rather underlie the creation of new computer-based instructional material. In this article we identify several e-learning scenarios which implement constructivist and action-oriented ideas and realizations like the cognitive apprenticeship model. Criteria for a system which supports those scenarios are derived. CleverPHL, a capture and replay program, is introduced as a tool which meets these criteria. With this tool, learners have the opportunity to actively take part in learning by observing and working with artifacts, sharing the process of producing rather than only presenting results.

**********

The number of e-learning materials is rapidly growing. Online courses, instructional software, multimedia presentations and interactive simulations are used in continuous education as well as schools and universities all over the world. Nevertheless, many authors object that lots of e-learning materials lack pedagogical principles and instructional concepts. For instance, Harper and colleagues complain about losing the balance between technology and pedagogy in e-learning materials (Harper, O'Donoghue, Oliver, & Lockyer, 2001). They quote authors who fear that institutions are rushing too quickly into new technologies without actually utilizing their potentialities in education. Bork mentions numerous reasons why computer-based instruction has not yet fulfilled its promises, for example the emphasis on hardware, little focus on learning and weak interaction (Bork, 1995).

We are also convinced that in many cases e-learning functionality is only used to implement traditional teaching with better means instead of innovating the way of learning. One of the most promising aspects of e-learning consists in the opportunity to activate learners and shift the responsibility for the learning outcome more to the learner. This can be done, on the one hand, by integrating communication tools in order to foster collaborative learning and discussion about problems and task solving with teachers as well as among peers. On the other hand, offering interactive learning materials like simulations enables the learner to see the effects of her or his actions immediately. Based on the output and feedback given by an interactive system, a student may check and adjust her or his internal models and concepts of the domain. Newly formed models and hypotheses can then be validated by interacting with the system. In opposition to the behaviorist, reactive design of learning material, the proactive design of highly interactive instructional software assigns an active and constructive part to the learner (Schulmeister, 2002). Bork also promotes interactive learning materials, which in his view have "great potential for major improvements in education" (Bork, 1995). He states that interaction is important for at least three reasons: it individualizes learning experiences, it activates learners and has strong motivational effects.

Still, interactive systems like simulations are just a small subset of e-learning materials. In general, it is more difficult to activate learners in e-learning or distance learning scenarios than in traditional classroom settings. This might be one reason for transferring only old and proven instructional concepts into e-learning scenarios, because it is much more difficult to create learning materials which focus on learners' actions rather than on their prerequisite knowledge, which must be derived from the products of their learning outcomes. In distance learning environments, assisting learners in their learning activities is difficult because the human teacher is not at the same place, and learning and teaching actions do not take place at the same time. Thus, observing actions of the learner is not possible and feedback can only be given based on the end-products presented by the learner. Most information about the process of producing these results is lost.

In the following, we argue for computer supported action-oriented e-learning, in which teaching and learning phases alternate seamlessly, and the responsibility for the learning outcome is shared between teacher and learners. Activating students in situated learning contexts is one of the most important aspects for motivation and other associated emotions, and thus the overall learning effectiveness. At first we outline two theories which stress the importance of students' activation. Then we sketch problems which occur when trying to focus on the activities of students in e-learning scenarios. From this standpoint we identify several educational scenarios where students' activities are focused and we describe criteria for a system which supports those scenarios. CleverPHL, our tool satisfying most of the described criteria is presented thereafter, followed by a small example demonstrating its application in an action-oriented e-learning setting focusing on the process of producing artifacts. Technical considerations and ideas for future improvements complete this article.

Constructivist Theories and Action-Oriented Learning

The traditional teaching and learning philosophy focuses on the primacy of instruction. Most important aspects are the systematic planning of lectures, strict differentiation of domains, teacher-centric classes with the teacher as presenter and controller of learning and embedded assessment. Learners stay mostly inactive and perceptive, because self-direction is largely limited by the instructional design of the teacher and the authors of learning material.

In contrast, constructivist learning theories put the learner into control. "Central to the vision of cognitivism is the notion of the organism as 'active'--not just responding to stimuli, as in the behavioristic rubric, but engaging, grappling, and seeking to make sense of things" (Perkins, 1991). Knowledge building is an active construction performed by learners based on the interaction with their environment. Thus, instruction plays a less important role. The teacher rather coaches learners in their learning activities and helps them to solve complex and authentic problems (Gerstenmaier & Mandl, 1995). Moderate constructivist approaches accept the importance of instruction and re-integrate instruction into the constructivist framework. This understanding of a moderate constructivist learning approach is currently being widely accepted by theorists as well as practitioners.

Action-orientation is a learner-centered pedagogical concept which emphasizes the importance of students' activation. Learning takes place while solving complex tasks based on the students' interests (Giani & Schroeder, 2004), often in collaboration and social interaction among peers. A teacher initiates, accompanies and helps to organize the activities of the students. Again, the role of the teacher has changed in contrast to traditional teaching scenarios (for more information on action orientation see Gudjons, 2001).

Both theories are based on the primary assumption that learning in general is an active and constructive process, which ought to be self-directed, situated, and embedded into social interaction. Thus it is emphasized that students must be assigned an active role in order to learn effectively.

One model for the implementation of constructivist theory is the cognitive apprenticeship model as introduced by Collins, Brown and Newman (1989). This model enhances pure situated and case-based learning by instructions and specific assistance within the first steps of active learning. Several phases of learning are distinguished. At first, learners are learning from experts by observing them solving authentic complex problems. Next they are confronted with simpler tasks themselves, for instance by having to finish problem solutions started by the teacher. In this phase learners are asked to recall and replay the behavior they observed before. These actions are closely supported by experts, who now observe how the cognitive apprentices handle the problem, and give feedback on the process of problem solving rather than only on the result of this process. In the following stages the expert fades out of the problem solving support while the complexity of the tasks are increasing. Finally, learners reflect on their problem solving strategies on their own and discuss those with their peers.

Implementations of situated models have shown that internal decision processes of experts cannot immediately be understood and interpreted by observers. There must be steps for the explication and externalization: experts demonstrate and explicitly comment on their problem solving, then they accompany the exercising by letting learners take (smaller) parts of the whole process at first and redraw from the process stepwise. When training in multiple contexts, learners should also articulate their thoughts and comment on their problem solving, and finally reflect upon their own problem solving process and compare it to the one of the expert (see Figure 1).

It is crucial for the implementation of the cognitive apprenticeship model in e-learning scenarios and in similar educational settings where actions and activities are emphasized, that mutual observations of activities (by teachers and students) are possible. A teacher, who has no access to his students' activities, cannot give specific feedback on their problem solving processes. Once there are bigger numbers of students involved, the learning progress becomes only indirectly accessible by looking at the products of achievement. On the other hand, students should also be in the position to observe the teacher's behavior who models an exemplary solution process. Students must also have the chance to observe their peers' activities to discuss alternative solution processes and reflect about their own achievement.

[FIGURE 1 OMITTED]

Action-Orientation in Distance Learning Scenarios

Observation of processes and activities in e-learning scenarios is quite more difficult than in classroom settings. In distance or self-directed learning contexts students must be able to learn at any time and at any place. Therefore, the performance of learning and problem solving activities and their observation take place in an asynchronous way. If a teacher introduces and demonstrates a solution process, this must be captured, stored, and transmitted to the students who can watch and comprehend the process after they have received it. Critical for this method is the representation format of the process. In traditional cases, a teacher would produce slides or images of significant states of the process, thereby diminishing information by this act of discretization. Text descriptions of the process, combinations of text and images, and animations are also often an appropriate way. Still, these means neglect the actions for problem solution. Only if the teacher can record problem solving activities and make it available as a video, the complete information can be made available to learners. On the other hand, if students wanted to present their actions to teachers, tutors or peers, they would be forced to learn additional software to create videos, slides or animations.

Today problem solving in most domains is carried out by utilizing software tools. If learning and problem-solving activities with software tools are to be presented to others, producing a video or slides with animations and screen shots with textual descriptions may be appropriate, but only if the process is only to be explained. In other cases, when students need to work with the artifact and resume the constructive process, those formats are insufficient, especially if both should be captured, the process of interacting with the tools for producing artifacts and the states and preliminary stages of the resulting products. In addition, when a teacher transmits a process to be observed by the students as a video or combination of text and images, students would first have to re-perform the process in their own programming environment while they watch or read the representation provided by the teacher. It can be assumed that switching between the two sources (process demonstration source and programming environment) may cause a split-attention effect and, therefore, impair the cognitive resources needed to understand the process (Chandler & Sweller, 1991).

A better solution to presenting problem-solving actions is a system which captures actions performed on a domain-specific environment, and stores representations of those actions in a file, which can be copied and transmitted via networks. A capture and replay-tool (CRT) is a program which records user actions on a graphical user interface (GUI) of a given program (target application) and can replay those interactions. User interactions consist of mouse and key events, for example. The actions can be stored in different formats which determine the way they can be utilized. In order to advance the benefit of digital videos or animations, interactions should be represented in a symbolic format. Thus, when replaying recorded interactions, the target application must be invoked and all captured events are dispatched as real actions to the real instance of the application, similar to the macro functionality built into some programs. With such a system, the teacher can simply perform the solution process in his version of the domain-specific environment. The CRT records all steps in the background and stores them for later reuse. Students on the other side use the same system to replay all recorded actions on a real instance of the same environment on their own computer. After replaying the sequence of captured actions, the tool remains in the resulting state of the replay process, so the students can easily continue the task started by the replay process.

A student can also utilize the same system to capture her or his solution process to facilitate remembering in the future, or to send the solution process to a tutor for assessment, feedback or help. The tutor can replay the student's process on the local installation of the tool, correct or resume the process, add hints and comments as audio annotations and send the whole record back to the student who replays it and so on. The same process can be utilized for peer discussion and peer help. In any case, the process of working with an artifact as the main aspect of action-orientation is emphasized and perceived together with the resulting product.

Capture and Replay Tool Functionality in Action-Oriented Learning Scenarios

In the previous sections, it has been argued that special systems and formats for the storage of action sequences are needed to support action oriented e-learning, and that a capture and replay tool with specific characteristics can satisfy those special demands. In the following section we want to describe some examples of educational applications in which a capture and replay tool can be utilized to enhance effectiveness of instruction and feedback (see also Schroeder & Spannagel, 2003a and 2004).

Scenario 1: E-lectures Integrating Problem Solving Demonstrations

Innovation of classical lectures can be accomplished by incorporating demonstrations of processes rather than only presenting facts. Since many steps of problem-solving are carried out by utilization of computer programs, these problem-solving steps should be demonstrated in the classroom by utilizing professional tools. When students recall the lecture they should not only be supported by slides and lecture notes, but rather should be able to recall the utilization of the software tools presented in the lecture. Therefore, the usage of the program must be captured with a CRT and stored for transmission and later replay. This file can be downloaded by the students and executed in their own CRT instance. In this case executing means replaying the events contained in the file. Before the replay process begins, the new instance of the target application is started. After the replay process has been finished or stopped by the user, the target application remains in the current state. This offers the student the opportunity to gain control over the whole interaction process and to continue the usage of the target application at any point in the recorded interaction.

A similar scenario is the creation of an online tutorial for certain software. With the aid of a CRT's replay process, the learner can watch the steps of interest being performed on his own target application and continue by himself afterwards. Those tutorial sequences can be enriched with audio sequences. Compared to textual annotations, audio information avoids a split-attention effect which occurs when switching the attention between two visual information sources (Chandler & Sweller, 1991; Mayer & Moreno, 2001).

Scenario 2: Interactive Exercise Sheets

Exercises should not always start from scratch. For instance, in programming tasks some modules might be developed up to a certain state and then can be completed or changed by the learner. This can also be implemented by the replay process of a CRT. Therefore, an exercise sheet may contain the task description including an interaction record to start the target application (programming environment) and bring it into the desired state. The student then may add his solution steps to the given record and send the whole record to an online judge who evaluates the transmittal. Thus not only the product (solution to the programming task) but also the process of finding the solution can be evaluated.

The same can be done for the purpose of online assessments. Sometimes the solution process contains valuable steps which are not apparent in the result. Thus submitting the recorded solution process in addition to the result may help the evaluator to come up with a fair evaluation.

Scenario 3: Remote Tutoring

Closely associated with interactive exercise sheets is the remote support by a tutor. If a student cannot completely solve a given task, he may record the steps up to the point of his problem and send the record to a tutor. The tutor can then replay the student's actions on the local instance of the target application. When he has discovered the problem, he may edit the student's record at the position leading to the problem or continue with the next steps leading to a solution of the problem. After that he sends the improved record back to the student who now can replay the tutor's corrections or advice. Again, the focus is being laid on the process rather than on the product only.

Scenario 4: Interactive FAQ Lists, How Tos, Peer Help and Experience Databases

This scenario is also closely associated with the ones before. When interactions of problem solving steps from experts or learners are captured, these recordings can be used as an experience base to be searched by other learners. One important aspect in this scenario is that not only correct solutions to tasks can be recorded but also typical errors can be stored and replayed. This provides for the unique opportunity to learn from typical errors.

Scenario 5: Peer Help

Again this scenario is closely related to remote tutoring. Instead of sending interaction records to tutors, they can be posted in newsgroups and discussed among peers. Again, the advantage of records over end products is that the solution process can be made available instead of the end product only.

In all these scenarios one important aspect of making problem solving permanent in order to be replayed is, that it helps to externalize the learners' or the expert's actions. This facilitates the reflection process by making steps (records) explicit and comparable to other's actions. This aspect is especially important in e-learning context, since new ways of learning need new competencies which must be acquired in the learning setting on a meta-learning level (Farinetti & Schroeder, 2002).

Criteria for a Capture and Replay System Supporting Process-Orientation

From the scenarios described above, criteria for a system which supports performing and observing activities in e-learning contexts can be derived (Schroeder & Spannagel, 2003b). The criteria are related to replaying, editing and analyzing captured processes. The description of every criterion is divided into two parts: first, the general aspects of the criterion are described; second, the importance of the criterion for e-learning is explained in an exemplary scenario.

* Capture and replay functionality: The system supports capturing and replaying user behavior with software. During the replay process, actions are dispatched to a real instance of the software, so that after the replay process has terminated or has been interrupted by the learner, the software remains in its last state. This offers the user the opportunity to continue working with the program from its current state on.

The teacher can simply use the program to capture a sequence of actions with any of the problem solving tools to be demonstrated to the students. Students can simply replay the process on their own installation of the same software. They can resume using the software after the replay process has reached the last recorded activity. They can also interrupt the process at any point and take over control to finish the started task. This is one of the most important aspects in action oriented learning in which self-learning has a huge demand for intrinsic motivation. Students can also capture their actions on software and transmit it to a tutor or to peers for discussion or feedback. This supports remote tutoring, collaborative learning and peer help.

* Different replay modes: The user has the chance to choose between different levels of replay details. Instead of replaying a process as a whole it can be reduced to the relevant states. For example, the system can skip all mouse motions and replays just mouse clicks and the keyboard inputs with a fixed time between these actions. To attract the user's attention, the location of the next action can additionally be highlighted. The user has also the opportunity to select parts of the recorded behavior to be replayed. Also available are other features of replay control known from the VCR such as fast forwarding, rewinding, pausing, or showing a detail screen.

Students can select a replay mode which fits their individual learning style. They can select a level of replay detail regarding to their prior knowledge and competencies. They can also choose the replay speed based on their individual learning tempo. They can select only specific parts of the process to be replayed they are interested in.

* Editing feature: The user can copy, insert and delete parts of the process. He can add comments and change properties of single actions.

The tutor can correct a record which has been sent by the student for help or feedback. He can add comments to give hints where problems may be. Teachers can edit captured action sequences to optimize presentations and interactive assignments. Teachers can give incorrect processes to students who have to correct them.

* Multimedia support: During the capture process, a user can add audio comments describing what he does and what he intends to do. Videos can also be captured, e.g. for usability evaluations. Also, audio or video annotation can be added afterwards, in order to deliver additional information such as generalization from a specific situation. All media can be replayed together in a synchronous manner.

The teacher can describe his problem solving strategies and his intentions while he performs the process. The students can listen to the teacher's articulations and externalizations of his internal processes and ideas.

* Symbolic and hierarchical representation: The captured actions are stored in a symbolic format. The files are human-readable and can be edited with any text editor. All actions are organized in a tree structure which symbolizes the levels of detail. This tree acts as an index to meaningful parts of the captured process. This index is built automatically and can be edited and improved by the user who has captured the process. Users which want to replay just parts of the process can use this tree representation to select the behavior they want to see.

Teachers who have captured a process can structure this process afterwards to make it more concise. Students can then select the part of the process they want to see in the tree representation of the captured action sequence.

* Support for analyses: Since all actions are stored in a symbolic format, the captured process can also be analyzed automatically. The generated tree representation can be used to detect behavior on different levels of detail (Hilbert & Redmiles, 2000) and to categorize sequences of actions. Therefore, it may serve as a method for automatic qualitative analyses. The system also offers a method to extract quantitative data from the captured process as data set and to make it available for statistical analysis systems.

Experimenters may detect patterns in the behavior of students using instructional software. They can extract a data set with a few clicks and load it into a statistical system. Thus the system can be used for learning research.

* Extensibility: The functionality of the system can be extended by users to fit special requirements. New action types, replay modules, detection algorithms for special behavior and editors for actions can be added to the system.

Experimenters can add new action types to capture special semantics of a specific target application. They can also add detection algorithms for particular behavior to find it automatically in captured action sequences.

* No need to adapt target applications: The target applications, whereon processes should be captured and replayed, don't have to be adjusted. They can be used together with the capture and replay system without modifications.

Teachers may find software in the Internet which they want to utilize in their classes. They can download this software and capture tutorials, help sequences or interactive task descriptions without modifications of the target application. Thus neither sources of foreign software nor programming knowledge are required.

* High usability: The system may be used in an intuitive way and its model is easy to understand. Symbols and metaphors known from the VCR help users to navigate through processes.

Younger pupils can use the system as well as students in university to capture and replay processes.

CleverPHL Implementing Action-Oriented Learning

CleverPHL is a capture and replay tool which allows to record and play back action sequences on Java applications and applets. It has been developed in cooperation of the University of Education in Ludwigsburg and the Technical Universities of Darmstadt and Aachen. Special effort has been paid to implement most of the criteria mentioned above. In principle, you can take any interactive software written in Java and use it together with CleverPHL (No need to adapt target applications). Processes can easily be captured and replayed by using the buttons taken from the VCR metaphor (Capture and replay functionality, high usability, see Figure 2).

You can select between several replay modes: the process can be replayed in real time or with a fixed time between chosen actions. In the latter mode, the location of the next event can be highlighted to attract the observer's attention (Different replay modes). The captured behavior is displayed in a tree representation whose leaves are the actions in highest detail. Nodes on higher levels represent interpretations of behavior generated automatically by the set of detection algorithms registered in CleverPHL (Symbolic and hierarchical representation). Attributes of all tree elements can be viewed and edited. Nodes can be inserted into the tree in order to further structure the record and define new higher-order semantic action groups (Editing feature, see Figure 3). Audio annotations can be added to an existing record to comment the process (Multimedia support). The structure tree can be used to detect and categorize action sequences in qualitative analyses. Data sets containing attribute values of behavior elements can be extracted to perform quantitative analyses (Support for analyses). Further detection algorithms, tree nodes, replay modules and editors can easily be added to fit the functionality of CleverPHL to specific purposes (Extensibility).

[FIGURE 2 OMITTED]

Example Scenario

To demonstrate some of the benefits of utilizing CleverPHL in an action-orientated assignment scenario, the following example describes first steps in a modeling and programming course for beginners introducing the programming environment BlueJ downloaded from (BlueJ, 2003).

[FIGURE 3 OMITTED]

In this assignment the students explore BlueJ and consolidate their concepts of objects, attributes and methods. Given a set of classes, BlueJ allows to create instances of the classes and to call methods on these instances directly. To create the interactive assignment, the teacher captures an action sequence demonstraing how to open a BlueJ project, to create an object and to call methods on that object. For example, he opens the example project shapes included in the BlueJ distribution, creates an instance of the class Circle and calls two methods on the instance which make the circle visible and change the diameter of the circle. He adds audio annotations where he describes his intentions. He stores the captured sequence as XML file and transmits it to the students or places it on a server for download.

[FIGURE 4 OMITTED]

The students replay the teacher's action sequence on their own instance of BlueJ (see Figure 4). They can select a replay mode of their choice. After the replay process reaches the last captured action, BlueJ remains in its last state (the project shapes has been opened, an instance of the class Circle has been created and the two methods have been invoked). Now the students can explore the features of the programming environment and the effects of calling methods on the objects created in the teacher's sequence. For example, their task might be to change the diameter of the circle created by the teacher, change its color, and move it 100 pixels to the left. For the assessment of the solution of the task they capture the necessary actions in BlueJ which are appended to the teacher's sequence. Sending back the whole action record (the teacher's and their own actions) to the tutor the formal evaluation is not only based on the end-product of task-solving but can also take the process of constructing this solution into account. Records can also be sent to peers for further discussion or feedback.

Technical Aspects

CleverPHL is implemented as an instantiation of the Java framework Jacareto, and allows programmers to add their own classes with task or target application specific functionality like own record elements, structure elements, and editors for those elements or replay modules. For example, the latter ones are called replayers. Each replayer is responsible for one or more types of record elements. They can be registered with a priority by an instance which manages the replay process. When replaying, this instance asks all registered replayers whether or not they handle the next structure element in order of their priority. The first positive response determines the replayer which is allowed to replay the actual structure element. Of course, own replayers can also be registered by the replay instance to include new replay functionality (see Figure 5).

CleverPHL links into the Java event queue, thus it can record and replay any software written in Java with a Java graphical user interface (GUI); this includes Java applications as well as Java applets. There is no necessity to adjust the target applications to make them work with CleverPHL. CleverPHL detects the structure of the visual components of the observed application. Besides this facilitation, the target applications may be extended to make application specific events available to CleverPHL which cannot be derived from the standard GUI events, but this is not the standard use of our framework in an educational setting.

It has been argued that the format of the stored process has to be symbolic and not image-based. Only symbolic formats allow for performing automatic analyses. Therefore, in CleverPHL every action is stored as a XML element with attributes describing the action. For example, a mouse action is saved as a XML element with the attributes x-coordinate, y-coordinate, the number of mouse clicks and the name of the component, on which the action has been performed.

[FIGURE 5 OMITTED]

When a XML file has been loaded into CleverPHL, it can be visualized in several ways. At first, it can be presented as linear list containing the sequence of atomic actions which have occurred during the capture process (see Figure 6a). This view is very complex and detailed. It is difficult to discover behavior of higher order. In order to make the record more concise, it can be structured with a set of detection algorithms. After such a record type has been chosen, the linear list is changed into a tree view which shows the structured record (see Figure 6b). This mechanism can be used to analyze the captured process automatically and to find and categorize action sequences.

The structure can be refined by the user by grouping tree elements together and by placing them into a new tree node (see Figure 6c). This is useful when a user wants to create a demonstration of behavior and wants to display the raw structure of the process. For such singular applications, it is not reasonable to write detection algorithms which find those raw process elements automatically. In this case, the structuring is better performed manually.

[FIGURE 6 OMITTED]

After a student has received such a manually structured process, he can simply select a part he wants to see (perhaps he wants to skip the introduction) in the tree view and replay just this part of the process.

The process of automatically structuring the record is based on methods known from compiler design (see Aho, Sethi, & Ullman, 1985). Formally, the record is structured regarding to a formal grammar called record grammar. Building the structure tree is similar to building the syntax tree in classical programming language settings. Following a top-down approach, there is a detection algorithm contained in every structure element class. Structuring the record means parsing it. Thereby a record tokenizer manages the delivery of the record elements to the structure building process (see Figure 7). New parser modules for specific software can be added to CleverPHL.

CONCLUSIONS AND FUTURE WORK

There are further ideas for more functionality making CleverPHL a valuable tool for teaching and empirical study of educational scenarios. One of them is the implementation of a remote record function: When capturing events, it should be possible to send the resulting record elements directly over the net to a CleverPHL instance on another computer. On this second computer, the arrived record elements will be replayed immediately. With this functionality an online tutor can observe what the student is doing in real-time and help with intermittent feedback.

[FIGURE 7 OMITTED]

The criteria mentioned above are not yet fully satisfied by CleverPHL. There are some ideas for further improvement which are listed below.

* Different replay modes: The user should be able to define different replay modes for different parts of the whole process.

* Multimedia support: CleverPHL just retrieves the behavior on the screen like mouse motions. Events in front of the screen won't be captured, mimics and gestures are lost. It would be helpful to make a video of the actions in front of the screen and link these scenes to the corresponding interaction record. During the replay process, the video could be replayed in a second window simultaneously. With the selective replay mode described above, one could just watch the interesting parts of the video or listen to audio comments on selected behavior elements. The next stage of extension would allow for general multimedia annotation of interaction records. Each record can be annotated with arbitrary multimedia data, which could be the video and audio streams mentioned above, but also hypertext, graphics, diagrams, and even simulations. With this feature, tutorials, presentations and interactive exercise sheets could be integrated completely into CleverPHL without the need of additional viewers and with the advantage of synchronous replay of all media. These enhancements are being developed in ongoing and future diploma theses at Aachen University.

* Extensibility: CleverPHL can be extended by writing classes and implementing interfaces, thus by programming. We plan to develop metatools which allow for adding new features without programming.

* No need to adapt target applications: Although many applications written in Java can be used together with CleverPHL, there are some problems which restrict the practicability of CleverPHL. For example, when the captured process contains an action sequence which selects a file in a file dialog per mouse click, the replay of this process will fail when the observer's file system is in a different state(e.g., if it contains more or less files in the current directory). For many of these pragmatic problems suitable heuristics must be found and implemented.

CleverPHL is part of the capture and replay framework Jacareto which is freeware and open source, and can be downloaded from the website (Spannagel, 2004).

References

Aho, A., Sethi, R., & Ullman, J. (1985). Compilers: Principles, techniques and tools. Reading, MA: Addison-Wesley.

Blue J (2004) [On-line]. Available June 6, 2004: http://www.bluej.org

Bork, A. (1995). Why has the computer failed in schools and universities? Journal of Science Education and Technology, 4(2), 97-102.

Chandler, P., & Sweller, J. (1991). cognitive load theory and the format of instruction. Cognition and Instruction, 8(4), 293-332.

Collins, A., Brown, J. S., & Newman, S. E. (1989). Cognitive apprenticeship: Teaching the crafts of reading, writing, and mathematics. In L. B. Resnick (Ed.), Knowing, learning, and instruction (pp. 453-494). Hillsdale, NJ: Lawrence Erlbaum Associates.

Farinetti, L. & Schroeder, U. (2002). Self-directed distance learning systems--A European-level cooperation initiative. Proceedings of Int. Conference on Open and Distance Learning, European Distance Learning Network, EDEN 2002.

Giani, E., & Schroeder, U. (2004). Situated learning with the learning environment jBat. Proceedings of the International Conference on Web-Based Education WBE04, Innsbruck.

Gerstenmaier, J., & Mandl, H. (1995). Wissenserwerb unter konstruktivistischer perspektive. Zeitschrift fur Padagogik, 41, 867-888.

Gudjons, H. (2001). Handlungsorientiert lehren und lernen: Schuleraktivierung--Selbsttatigkeit--Projektarbeit. Bad Heilbrunn / Obb.: Klinkhardt.

Harper, B., O'Donoghue, J., Oliver, R., & Lockyer, L. (2001). New designs for web-based learning environments. In C. Montgomerie, & J. Viteli (Eds.), Proceedings of ED-MEDIA 2001, World Conference on Educational Multimedia, Hypermedia & Telecommunications (pp. 674-675). Tampere, Finland: Association for the Advancement of Computing in Education.

Mayer, R., & Moreno, R. (2001). A split-attention effect in multimedia learning. Evidence for dual processing systems in working memory. Journal of Educational Psychology, 90(2), 312-320.

Perkins, D. N. (1991). Technology meets constructivism: Do they make a marriage? In T. Duffy, & D. Jonassen, Constructivism and the technology of instruction: A conversation (pp. 45-56). New Jersey: Lawrence Erlbaum Associates, Publishers.

Schroeder, U. (2002). Meta-learning functionality in e-learning systems. International Conference on Advances in Infrastructure for Electronic Business, Education, Science, and Medicine on the Internet 2002, L'Aquila, Italy.

Schroeder, U., & Spannagel, C. (2004): The case for action-oriented e-learning. Proceedings of ED-MEDIA 2004, World Conference on Educational Multimedia, Hypermedia & Telecommunications, to appear.

Schroeder, U., & Spannagel, C. (2003a). CleverPHL: Focusing on the learning process by supplementing instructional software with digital recording and evaluation functionality. World Conference on E-Learning in Corp., Govt., Health., & Higher Ed. 2003(1) (pp. 2413-2420).

Schroeder, U., & Spannagel, C. (2003b). Implementierung von e-learning--Szenarien nach der theorie der kognitiven lehre. In A. Bode, J. Desel, S. Rathmayer, & M. Wessner (Eds.), DeLFI 2003, Lecture Notes in Informatics (pp. 195-204). Gesellschaft der Informatik, Vol. P-37. Bonn: Kollen Druck + Verlag.

Schulmeister, R. (2002). Taxonomie der interaktivitat von multimedia--Ein beitrag zur aktuellen metadaten-diskussion. Informationstechnik und Technische Informatik, 44(2002), 193-199.

Spannagel, C. (2004). Jacareto Website: Download, Documentation and Further Information [On-line]. Available June 8, 2004: http://jacareto.sourceforge.net/

ULRIK SCHROEDER

Aachen University, Germany

[e-mail]

CHRISTIAN SPANNAGEL

Ludwigsburg University of Education, Germany

[e-mail]