Linking and Timing Information Presentation in Multimedia Educational Systems.

Linking and timing what is to be presented to the learner is particularly important in multimedia educational environments, especially when the same piece of material can be presented to the learner in different media, often more than once. This article addresses the linking and timing attributes of the presentation of learning material in multimedia educational systems by way of a control in teaching strategy selection. The control mechanism is illustrated through the model for teaching strategy selection. The model is based on factors that affect human teachers in their teaching strategy selection, which in turn governs the decision of which piece of material to present, and the timeliness of such. The benefits of the model are demonstrated through ARISTOTLE, a multimedia education system in which the model is deployed.

User disorientation has frequently been discussed in research in multimedia. The freedom in user navigation and the division of information into nodes may disturb the construct provided in conventional linear structures (Chapman & Chapman, 2000). This undermines the benefits of the flexibility of information presentation in multimedia educational systems. Reliance on commercial concerns, such as business involvement and marketability, has prompted a generation of high-street multimedia educational packages, which employ attractive interface designs, yet without much expectations on specific learning goals, less to the activities and information essential to the achievement of the specific goals, and none at all to the formative evaluation and revision of the learner's progress against specific learning goals (Tong & Agius, 1999). The main weakness of multimedia systems in educational environments stems from the loosely-structured links between the different pieces of learning materials, and the lack of educa tional validity for how the pieces of learning material should be related, and when a particular piece of material should be presented.

In multimedia systems in general, the way in which information is fundamentally associated in a system is predefined by the author in a hypermedia structure by way of nodes and links. The order and the timeliness of the presentation, that is, leaving one presentation for the next, is executed by the user. The importance of linking and timing presentations has been addressed by many in the field of multimedia and hypermedia, but often not directly related to educational settings (Courtiat & Oliveira, 1996; Fernando, Soares, Rodrigues, & Muchaluat-Saade, 2000; Hardman, Bulterman, & Rossum, 1993a; Rutledge, Ossenbruggen, Hardman, & Bulterman, 1999). In the education arena, the importance of these two issues has been investigated by research in computer-based instructional systems, as well as research in human teaching. To this end, the emphasis is on the combined implication of the two issues, "when" to present "what," and why (Tong, 1999).

This article aims to define the linking and timeliness of information presentation in multimedia educational systems by employing a set of mechanisms which governs teaching strategy selection. It does not propose new research in computational links and their representations in hypermedia. Its objective is to demonstrate how the flow and the timeliness of presentations in multimedia educational settings can be improved with instructional reasoning. The proposition of a control in teaching strategy selection derived from the fact that teaching strategies are conceptual means through which the teacher imparts material to the learner. The links and timeliness of the physical presentation of the learning material can be conceptually structured through the use of different teaching strategies. The rest of the article is organised as follows. The next section reviews related work in the linking and timeliness of information presentation in multimedia systems, as well as the use of teaching strategies in educational systems. The article then uncovers the characteristics, which constitute satisfactory teaching strategy selection in both machine and human teaching. The model for teaching strategy selection is then presented, followed by a demonstration of its operations and functions as it is deployed in ARISTOTLE, a multimedia educational system for teaching zoology. The benefits of the model are then discussed.

RELATED WORK

The linking and timing of multimedia presentations has been addressed in research in hypermedia. An example of such is CWI Multimedia Interchange Format (CMIF). CMIF is a system-independent representation of multimedia presentations (Hardman, Rossum, & Bulterman, 1993b). It represents two main aspects of multimedia presentations: the structure of the presentation, and the logical assignment of multimedia resources. CMIIF provides a multimedia authoring environment to support the creation of multimedia presentations from existing media segments. The structure is also used for deriving timing constraints in the presentation. The authoring process is divided into two stages: creating the structure of the presentation, and the assignment of detailed timing constraints between constituent media segments. The two stages are maintained by two views in the CMIF environment: the hierachy view that shows the structure, and the channel view, which shows the logical resource usage. The CMIF authoring environment enables hyperlinks to be created between segments within a presentation. The main purpose of CMIF is to aid multimedia authoring by exploiting the inherent modularity in the structure of multimedia presentations. It moves away from the timeline based authoring approach, in which multimedia segments and associated location information are placed on a line along which time flows. CMIF does not acknowledge the importance of why and how the segments should be linked in specific ways.

Another research issue which addresses the linking and timing in multimedia presentations is the automatic temporal formatter (Buchanan & Zellweger, 1993). The aim is to aid the author in creating temporal layouts of multimedia presentations. A temporal layout indicates when events, from the same or different segments, should occur in a multimedia presentation. The formatter uses a set of mechanisms explicitly representing temporal relationships among the media segments in a presentation. Temporal relationships describe how the segments should be combined in a presentation in terms of time. The temporal relationships are defined by certain parameters. Those which are directly relevant to the linking and timing of presentations are temporal relation types, flexibility and flexibility metrics. Temporal relation type specifies the order of occurrence of the segments by means of describing the start, finish, overlap, parallel and sequential components, and the duration of a segment. Flexibility specifies whether a temporal relationship is mandatory, that is the relationship must be satisfied in the resulting presentation layout, or optional, that is the relationship is desirable but can be ignored. Flexibility metrics provide the basis for choosing between the two aforementioned flexibility options. The metrics measures the degradation in presentation quality due to ignoring an optional temporal relationship. The emphasis of the work is on the algorithmic automation of the temporal formatting, and again not on the rationale of why the presentations are temporally formatted in the ways defined. The motivation is biased towards how a system would and could order a presentation, and towards runtime and compile time support, rather than how the audience of the presentation would benefit from the formatted temporal ordering.

Research related to the linking and timing of multimedia presentations has tended toward investigation of the relationships between objects and events within a particular media segment, and the sychronisation of different segments (Courtiat & Oliveira, 1996; Czumaj, Finch, Gasieniec, Gibbons, Leag, & Rylter, 1999; Dalal, Quible, & Wyatt, 2000; Fernando et al., 2000; Ossenbruggen & Eliens, 1994). There is another related area of research that addressed the expressiveness of links in hypermedia systems (Moreau & Hall, 1998). The expressiveness of a system is defined by its ability to follow various links in user navigation. They compared the flexibility offered by different types of links, for example, generic links and adaptive links. Nevertheless, research concentrating on the meaning of the linking and timing of presentation amongst the pieces of information (e.g., should information C be presented before information A, does the order of presentation matter at all, what are the situations in which the order of presentation does matter) to the audience is comparatively scarce.

The main reason for the lack of emphasis on the meaning of the links is due to the intended richness of the multimedia presentations, in which a wide variety of media types and information can be simultaneously displayed. Creating a multimedia presentation requires the specification of which media segments are played, when and where, visually or audibly, or both. Nonetheless, for meaningful communication to take place, the author (as well as the audience) ought to think in terms of the essence of the information, that is, the message being conveyed, the ordering of the material, and the flow of the presentation.

Meaningful communication in education between the teacher (human or machine) and the learner relies on the conceptual and physical representation of knowledge by the teacher, and the interpretation of the represented knowledge by the learner (Wenger, 1987). The physical representation contributes to the efficient transfer of knowledge, especially when there is a well defined learning goal (Shapiro, 2000). The linking and timeliness of presentation in this respect focuses mainly on the effect of the presentation structure to the learner. In empirical studies of human teaching, the use of different teaching strategies has been found to form a substantial part of the overall teaching interaction (Tong & Angelides, 2000). It is certainly important for an educational system to provide more than one teaching strategy, otherwise it may not be able to adapt promptly to the changing cognitive needs of the learner (Sadler-Smith & Riding, 1999). The key to structuring the linking and timeliness of presentation is by no means to have as many teaching strategies as possible, but how to select among them, that is, to decide on when, why, and which one to use.

Although many multimedia educational systems incorporate more than one teaching strategy, a common basis is lacking on which the rules governing the selection can be applied. This is mainly due to the difficulty in abstracting and making accessible the knowledge that the human teacher has (Tong & Angelides, 1999). Attempts have been made in existing systems to support teaching strategy selection. An example of such a system is DOMINIE (Elsom-Cook, 1991). DOMIME is a multimedia educational system that teaches the procedural skills of using computer applications. It embodies eight distinct teaching strategies, with an overall tree-structure plan for selecting among them. The primary decision the system makes when considering which strategy to use is the balance of teaching, that is to teach new material, and assessment, to evaluate the learner's current knowledge state. Within the decision about teaching or assessment is a decision about the general area on which the system will focus. The system has a set of heuristics to give an order of preference to the possible teaching strategies to be used. The objective of DOMINIE is to capture different teaching strategies at a level of detail which allows them to be distinguished. Although the need for a formal basis for strategy selection is recognised, the justification of why the rules work is omitted.

One of the most significant recent contributions in teaching strategy selection is the "dynamic model selection" in Generic Tutoring Environment (GTE) (Marcke, 1998). GTE is used for developing courseware based on a generic instructional task-based approach. The selection takes into account instructional tasks, instructional methods, and instructional objects. Instructional tasks are activities accomplished by the teacher in an educational discourse. Instructional methods are procedures to carry out tasks, which are equivalent to teaching strategies in the context of this article. Instructional objects are means that methods employ. Instructional methods are chosen according to a numerical applicability value for each method. These values are calculated according to a list of conditions attached to the individual methods. While GTE has comprehensively represented the instructional processes from the teacher's viewpoint, the learner's perspective has not been sufficiently addressed (Elen, 1998). GTE's selecti on could be improved if it could couple the learner's results with the selection mechanism. There is no separate information for the selection procedure in GTE. While having the information embedded in the teaching strategies themselves make their context more explicit, it obscures the relationship between the various elements within the selection mechanism, thus lessening the transparency between the different aspects that affect selection (Murray, 1998).

The previous examples indicate that research in the linking and timing of information presentations has focused mainly on the semantics of the media segments rather than on the semantics of the messages put across. The rationale behind the flow of the message itself, the timeliness of the message presented to the audience, and the order of presentation, is not always addressed. These issues cannot be neglected in multimedia educational systems for them to be instructionally valid. Although the importance of teaching strategy selection in the linking and timing of information presentation has been acknowledged, a rigorous reasoning behind why a particular teaching strategy is chosen at a given point in time is lacking.

This article proposes a model for teaching strategy selection to provide instructional control over the linking and timing of information presentation, as well as to provide a basis for formalising teaching strategy selection in multimedia educational systems. The next section reveals the characteristics which constitute satisfactory teaching strategy selection in both machine and human teaching, and then presents the derived model for teaching strategy selection.

STRATEGY SELECTION FOR THE FLOW AND TIMELINESS OF INFORMATION PRESENTATION

Teaching strategies refer to any mode or modes of action or specific activities that facilitate the accomplishment of one or more learning goals. "Teacher" in this section can denote either human or machine. Current research in human and machine teaching suggests that when considering the use of teaching strategies, the following issues should be taken into account: (a) subject matter and cognitive differences, (b) test scores, (c) "trial and error" selection, (d) learner involvement, (e) the teaching environment, and (f) human teacher involvement (Tong & Angelides, 1999). Subject matter is important because it represents what is to be taught. The differentiation of cognitive levels of learners is important for individualised teaching. With the analysis of test scores, the teacher gets to know how the learner is acquiring knowledge taught. Trial and error selection is necessary in cases where the teacher would like to explore, which teaching strategy is the most appropriate to use in a particular situation. Learner involvement is important so that he/she can have some control over the discourse. Understanding of the teaching environment is required because it determines the media through which a teaching strategy is delivered. Human teacher involvement is necessary as the human teacher is an expensive and effective resource, which should only be complimented and not replaced by technology (Anderson, Boyle, Corbett, & Lewis, 1990).

When considering how to select a teaching strategy for a given point in time, different teaching strategies are needed for the different pieces of learning material within the domain, and for different cognitive levels among learners (Khan & Yip, 1996). This implies that a teaching strategy selection is needed when there is a change in the nature of the learning material, a change in cognitive needs among the different learners who interact with the teacher, and a change in cognitive needs within the same learner as he/she advances along the educational discourse. One question that needs to be addressed is "when" specifically a teaching strategy selection should be triggered. One answer is when a teaching strategy appears to have failed. A learner's test results on the material presented may reflect a degree of teaching strategy success and failure (Hannafm & Land, 1997). Test scores may also indicate whether the learner should revise the same material before moving on to new ones, which, in turn, determine the scope of application of the next teaching strategy, that is, whether it is to be used for teaching/reteaching, or testing what has been taught (Woolf & Hall, 1995). Teaching strategy selection is needed when test scores of a learner are consistently excellent or poor. The choice of the next teaching strategy may depend not only on the history of the teacher-learner interaction with the different strategies. A trial and error selection may be used to explore how various teaching strategies may suit a learner, for example, at the time when the learner's portfolio has not yet built up enough information for the teacher to infer which teaching strategies are compatible with the learner. The learner's own learning preference with specific teaching strategies ought to be considered as well. It implies learner's control, which contributes to intrinsic motivation in learning (Merrill, 1998). However, the learner may not always know what is best for him/herself (Shyu & Brown, 1995). The teacher should therefore be able to provide an environment, in which the learner has the impression of control over the presentation of material, with the amount of control carefully monitored.

The nature of the material affects the teaching environment where the learning takes place. The teaching environment in turn governs the type of communication channel between the system and the learner. Different teaching strategies that are used to teach different material within the same domain require different teaching environments. For example, teaching procedural knowledge may call for the animated display to demonstrate each process to the learner, whereas teaching factual knowledge may only need a textual display. Teaching strategy selection is therefore needed when there is a change in the presentation method for a different piece of learning material. The choice of which teaching strategy to use (which is conceptual) may depend on which teaching environment is required, since these environments facilitate the delivery of the teaching strategies (which is physical). Teaching strategy selection in machine teaching should also allow for human intervention in the process to ensure system efficiency and to explain to the learner about the presentation when necessary to avoid confusion.

The Model for Teaching Strategy Selection

Subsequent to the characteristics of teaching strategy selection uncovered previously, the model for teaching strategy selection consists of the following aspects: (a) learning goal, (b) use of teaching strategy (teaching/ re-teaching or testing/re-testing), (c) the subject domain (content and type), (d) the learner profile (advancement level, specific teaching strategy success rates, learner's preference), (e) the teacher's repository (the teaching strategies, general teaching strategy success rates), (f) teaching strategy delivery environment (text, graphics, audio/visual, animation, combinations of two or more), and (g) external human teacher support.

Learning goals define the objectives of the learning session. Objectives state specific skills, tasks, the exact material to be learned, and give the teacher and the learner a standard by which to judge if they are moving towards achieving the learning goal. The use of a teaching strategy is determined by the scope of application, which is distinguished between teaching/ re-teaching and testing/re-testing. The content of the subject matter indicates the message to be conveyed. The type of subject matter refers to whether the material to be imparted is procedural or declarative knowledge. The advancement level of the learner is distinguished between novice and advanced. Specific teaching strategy success rates indicate the different teaching strategies' prior successes with the individual learner. Learner's preference refers to the learner's choice in the teaching strategy to be used. Teaching strategies are divided into multimedia and non-multimedia based. General teaching strategy success rates give informa tion on the overall success rates of the teaching strategies with all the learners who have interacted with the system. Teaching strategy delivery environment provides the learner with the scenario in which the learning material is introduced. External human teacher support calls for the intervention of the human teacher.

The structure that governs the teaching strategy selection, and in turn the linking and timeliness of the learning material presented is as follows. The model commences selection with the retrieval of all appropriate teaching strategies indicated by the learning goal. This set of teaching strategies is a subset of the full set of teaching strategies available. It then examines the subject domain to identify the teaching strategy delivery environment needed according to the content and type of the learning material needed to satisfy the learning goal. For example, if the learning material in hand needs to be taught with audio and video, then those teaching strategies that support multimedia, that is, non-textual presentation are included in the next stage of the selection.

The model then identifies what the teaching strategy is to be used for, that is, to teach or to test. This is indicated by the learner's test scores, which also shows the history of successes and failures of specific teaching strategies with the learner. This is recorded by two success ratings for each teaching strategy, one for teaching and one for testing. If the specific success rates do not indicate a particular preferred strategy, for example, because two or more have the same ratings, the general success rates from the teacher's repository are used to filter the set further. The general success rates gives the overall success ratings of a teaching strategy with all of the learners who have interacted with the system, as opposed to with an individual learner. The system "learns" what is a generally appropriate teaching strategy for certain situations as it builds up its teaching experience through increasing interactions with increasing numbers of learners. If the general success rates do not indicate a particular teaching strategy, then the advancement level of the learner will be referred to. If the learner is a novice, the human teacher will be called to decide on the information to be presented next. If the learner is advanced, he/she will be asked for his/her preference in what is to be presented next. The model for teaching strategy selection increases the learner's control as the learner progresses. The relationships of the aspects, together with the sequence of selection which govern the linking and timing of information presentation in the model for teaching strategies selection are summarised in Figure 1.

Implementation of the Model for Teaching Strategy Selection

The sequence of the selection process performed by the model has been described in the previous subsection. In implementation, this sequence for linking and timing information presentation is supported by an order engine in the model which controls the linking and the timeliness of the presentation as a whole. Embedded in the engine are certain rules that ensure consistent decisions. These rules govern teaching strategy ratings and the selection. There are four types of rules: (a) suitability rules, (b) proficiency rules, (c) blockage rules, and (d) sequence rules. Suitability and proficiency rules both measure a teaching strategy according to the learner's profile. Blockage rules assess a teaching strategy according to its previous successes and failures in similar situations. Sequence rules measure a teaching strategy according to the order of the use of teaching strategies. Suitability rules favour the teaching strategies that have had successes previously with the learner. Proficiency rules select accord ing to the learner's proficiency in the domain, that is, his advancement level. For instance, advanced learners are given more control than novices. Blockage rules lower the priorities of those teaching strategies that have failed previously. Sequence rules favour teaching strategies that are similar to the one used previously. Sequence rules are consistent with the idea that the use of teaching strategies should be coherent. There are two exceptions to the sequence rules. One is when similar teaching strategies have been used and effective learning has not taken place. Then, a completely different teaching strategy may have a higher priority compared to the ones similar to the previous ones used. The other one is when the human teacher decides the sequence rules are to be broken, for example, when a learner has perfect progress with one particular teaching strategy for a considerable length of time. This is to check whether the learner still finds it challenging to learn with that particular teaching strateg y to avoid boredom.

The model for teaching strategy selection has been developed as an "add-on" component to the architecture of any full-scale multimedia educational systems. Such systems should be able to offer different teaching strategies and physical representations of the information to be presented, some history of system-learner interaction, and some contextual structure of how the multimedia documents can be linked. The model for teaching strategy selection has been deployed in ARISTOTLE (Agius, 1997), a multimedia educational system for zoology. ARISTOTLE was developed using Asymetrix Multimedia Toolbook 4.0 under Microsoft Windows. It was originally a prototype to demonstrate the use of "multimedia frames" to represent video and audio semantics, in terms the spatial and temporal attributes of entities in a segment, actions of the entities, and events triggered by the actions of entities. ARISTOTLE initially includes: multimedia-based and non-multimedia based teaching strategies, which were originally selected on the basis of availability of the media segments in the system, a multimedia resource for the physical representations of the subject domain, a system-learner interaction in the form of how much the learner has yet to achieve, by referring to the given set of learning goals, before completing the course of learning, and "multimedia frames" (m-frames) (Agius, 1997) which contain information on how the multimedia documents can be linked. ARISTOTLE uses two types of m-frames: syntactic m-frames (SYMs) and semantic m-frames (SEMs). SYMs contain information about the syntactic co-ordinates within a frame of video clip. SEMs contain information, which represent the semantic content of a sequence of video or audio frames. All of ARISTOTLE SEMs, with the exception of the teaching strategy SEMs, are stored as Borland Paradox databases. Each record of the databases contains two fields: perspective, and instances. Teaching strategy SEMs are procedural in nature, and are stored as Openscript codes in Multimedia Toolbook. Figu re 2 shows a conceptual representation of a learning goals SEM.

The entities of interest to the systems are animals. Animals are divided into vertebrates and invertebrates, and further sub-divided into classes such as mammals, reptiles, and arthropods. A learner can choose at the start which class of animals to learn first. The learning goal and the content of the material to be presented are then decided, and the teaching strategy selection triggered. Figure 3 shows the learner choosing what he would like to learn. Figure 4 illustrates the use of teaching strategy aspect in the model manifesting: the next teaching strategy is to be used for teaching. Figure 5 illustrates a multimedia-based teaching strategy is selected, triggering the associated information to be presented.

This section has presented the model for teaching strategy selection as a means for structuring the linking and timing of information presentation in multimedia educational systems, by providing a basis for deciding "when" to present "what," and why. The next section reveals the implications of the model on information presentation, through the discussion of the benefits of the model for teaching strategy selection.

IMPLICATIONS ON INFORMATION PRESENTATION IN MULTIMEDIA EDUCATIONAL SYSTEMS

The defining characteristics of multimedia systems is the way in which these systems offer the user the opportunity to freely consult the available information through a network of inter-related nodes and links. The purpose of the control mechanism in the model for teaching strategy selection which supports the linking and the timing of information presentation is to have a means of generating dynamically, at run-time, appropriate linking possibilities between nodes of learning material. The objective of such support is to facilitate efficient and effective learning. This is the core differentiation to be distinguished between multimedia systems in general, and multimedia educational systems. The latter have to be responsive to the needs of the learner, the type of learning material involved, and the learning goals to be accomplished. In this respect, a form of control is needed for efficient learning, and for the maintenance of the delicate balance between desirable flexibility for the learner, and the nece ssary instructional direction by the teacher, that is, the system.

The insertion of the model for teaching strategy selection improves the ARISTOTLE's educational validity in several respects. While the multimedia resource remains unchanged, the history of system-learner interactions, as well as the mechanisms of information presentation have been modified to accommodate the complete model for teaching strategy selection. The improvements can be divided into two main types. The first type involves the implementation of the aspects of the model for teaching strategy selection that have not been included at all, or not included at a sufficient level in the initial version of ARISTOTLE. They were left out, or not dealt with in enough depth originally, primarily because of the lack of a formal basis for teaching strategy selection, and in turn an educationally valid definition for the linking and timing of information presentation. ARISTOTLE's original subject domain did not differentiate the different types and content of the material. It also included a history of system-lear ner interactions, but did not consider the learner's advancement level and his/her preference. It included different teaching strategies, but did not address the teaching strategy general success rates. With the implementation of the model for teaching strategy selection, all these aspects are now included in order to drive the model.

The second type of improvements involved the implementation of the mechanisms that manipulated information from all the aspects for the model for teaching strategy selection, which is directly relevant to the linking and timing of information presentation. The modifications made to ARISTOTLE facilitate: (a) the calculation and recording of teaching strategy success rates, (b) the elicitation of the learner's preference, (c) the differentiation between teaching/re-teaching and testing/re-testing, (d) the call for human teacher intervention, and (e) the integration of information from all these aspects into the model.

ARISTOTLE started off as a large depository of information, with the capability of multimedia representations. It provided source material and an organisational backbone for an educational environment. The linking and timing of information was largely ad hoc. The model for teaching strategy selection provides specialised functions to assist efficient learning to take place by enforcing formal instructional requirements. Not only does the model for teaching strategy selection offer a set of mechanisms to define the linking and timing of information presentation, it also defines how satisfactory teaching strategy selection should take place in computer-based instructional systems as a whole.

In terms of systems architecture, the benefits of the model for teaching strategy selection essentially lie in its provision of a hybrid intelligent information presentation framework. Such frameworks consist of hypermedia, that is, textual or non-textual, with expert system support (Angelides, 1995). Generically in this hybrid framework, the hypermedia component provides a means of presenting related information units in a conceptual non-linear way. The expert system component is embedded in the system to provide an alternative approach to presenting information contained in a unit, as well as additional information from related units.

The model for teaching strategy selection provides the expert system support by defining rules for selection based on the teacher's expert knowledge. The advantage of such rules is to separate what information might be useful from what might not, and therefore narrowing the search domain based on the user's needs at a given point in time. In the model for teaching strategy selection, the expert system component is used to integrate procedural knowledge into hypermedia for the problem-solving undertaken by the teacher, such as a diagnosis of the learner's state, and the successes and failures of the various teaching strategies. The expert system capability is associated with ARISTOTLE's m-frames, enabling it to work in synergy with, and communicating important information back and forth to ARISTOTLE's multimedia resource.

The hypermedia structure for learning has been found to be cognitively difficult (Zellweger, Regli, Mackinlay, & Chang, 2000), especially for learners who do not have any background knowledge of the subject domain. The model for teaching strategy selection helps the learner view relevant information. Guidance can be in the form of automatically presenting the next piece of material, or suggestions according to some degree of priority for what the learner should be presented next, for example, by way of learner preference and external human teacher support, backtracking with non-previously visited presentation options, for example, re-teaching or retesting the same material with a different teaching strategy when the initial one appears to have failed. In which case, the model for teaching strategy selection has provided a multimedia educational environment which combines the flexibility of non-linear presentation of information with structure, control, and inferencing capabilities.

Figure 6 summarises the benefits of the model for teaching strategy selection as a means of linking and timing information presentation, by illustrating how best ARISTOTLE has fulfilled the requirements of flexible multimedia systems, as well as the instructional validity of educational systems. The criteria used for evaluating multimedia systems requirements are based on those proposed for the structuring and authoring of hypermedia documents (Fernando et al., 2000; Hardman et al., 1993a; Hardman et al., 1993b). The criteria used for educational systems are based on the evaluation set out for desirable behavioural properties expected from a teacher in any educational environment.

CONCLUDING DISCUSSION

This article has illustrated how the linking and timing of information presentation in multimedia educational systems can be improved, by using a teaching strategy selection control. The article has also demonstrated the viability of such control through the implementation of the model for teaching strategy selection in ARISTOTLE, a multimedia education systems which initially presented information to learners in an unstructured way. The model, derived from the characteristics of human and machines teaching, not only offers a means of linking and timing presentations, but as a basis for formalising teaching strategy selection in instructional systems.

Future work includes: (a) increasing learner's control to further exploit the flexibility of information presentation, (b) fine tuning the granularity of the rating indices for weighing teaching strategy successes, as well as (c) the development of complimentary models supported by expert knowledge to provide a balance in information presentation between flexibility and educational effectiveness of multimedia educational systems.

References

Agius, H.W. (1997). A full-scale semantic content-based model for interactive multimedia information systems. Unpublished doctoral dissertation, London School of Economcis, London.

Anderson, J.R., Boyle, C.F., Corbett, A.T., & Lewis, M.W. (1990). Cognitive modelling and intelligent tutoring. In W.J. Clancey & E. Soloway (Eds.), Artificial intelligence and learning environment. Amsterdam: Elsevier.

Angelides, M.C. (1995). Developing hybrid intelligent tutoring and hypertext systems. New Review of Hypermedia and Multimedia, 1, 67-106.

Buchanan, M.C., & Zellweger, P.T. (1993). Automatic temporal layout mechanisms. Paper presented at the ACM Multimedia 1993, Anaheim, CA.

Chapman, N., & Chapman, J. (2000). Digital multimedia. Chichester: John Wiley and Sons.

Courtiat, J.P., & Oliveira, R.C.D. (1996). Proving temporal consistency in a new multimedia synchronization model. Paper presented at the ACM Multimedia Conference 1996, Boston, MA.

Czumaj, A., Finch, I., Gasieniec, L., Gibbons, A., Leag, P., & Rylter, H. (1999). Efficient web searching using temporal factors. Lecture Notes in Computer Science, 1663, 294-3 05.

Dalal, N.P., Quible, A., & Wyatt, K. (2000). Cognitive design of home pages: an experimental study of comprehension on the World Wide Web. Information Processing and Management, 36(4), 607-621.

Elen, J. (1998). Automating ID: The impact of theoretical knowledge bases and references systems. instructional Systems, 26(3-4), 281-297.

Elsom-Cook, M.T. (1991). Dialogue and teaching styles. In P. Goodyear (Ed.), Teaching knowledge and intelligent tutoring (pp. 123-148). Norwood, NJ: Ablex.

Fernando, L., Soares, G., Rodrigues, R.F., & Muchaluat-Saade, D.C. (2000). Modeling, authoring and formatting hypermieda documents in the HyperProp system. Multimedia Systems, 8(2), 118-134.

Hannafin, H.M., & Land, S.M. (1997). The foundations and assumptions of technology-enhanced student-centred learning environments. Instructional Science, 25(3), 167-202.

Hardman, L., Bulterman, D.C.A., & Rossum, G. van. (1993a). Links in hypermedia: The requirements for context. Paper presented at the ACM Conference on Hypertext 1993, Seattle, WA.

Hardman, L., Rossum, G. v., & Bulterman, D.C.A. (1993b). Structured multimedia authoring. Paper presented at the ACM Multimedia, Anaheim, CA.

Khan, T., & Yip, Y.J. (1996). Pedagogical principles on case-based CAL. Journal of Computer Assisted Learning, 12, 172-192.

Marcke, K. van. (1998). GTE: An epistemological approach to instructional modelling. Instructional Science, 26(3-4), 147-191.

Merrill, M.D. (1998). ID Expert: A second generation instructional development system. Instructional Sciences, 26(3-4), 243-262.

Moreau, L., & Hall, W. (1998). On the expressiveness of links in hypertext systems. The Computer Journal, 41(7), 459-473.

Murray, T. (1998). Authoring instructional expertise in knowledge based tutors. Instructional Science, 26(3-4), 243-262.

Ossenbruggen, J. van., & Eliens, A. (1994). Music in time-based hypermedia. Paper presented at the ACM European Conference on Hypermedia Technology 1994, Edinburgh.

Rutledge, L., Ossenbruggen, J. van., Hardman, L., & Bulterman, D.C.A. (1999). Anticipating SMIL 2.0: The developing cooperative infrastructure for multimedia on the web. Computer Networks, 31(11-16), 1421-1430.

Sadler-Smith, E., & Riding, R. (1999). Cognitive styles and instructional preferences. Instructional Science, 27(5), 355-371.

Shapiro, A.M. (2000). The effect of interactive overviews on the development of conceptual structure in novices learning from hypermedia. Journal of Educational Multimedia and Hypermedia, 9(1), 57-82.

Shyu, H.S., & Brown, S.W. (1995). Learner control: The effects on learning a procedural task during computer-based videodisc instruction. International Journal of Instructional Media, 22(3), 217-231.

Tong, A.K.Y. (1999). Developing a multimedia-based intelligent tutoring system. Journal of Computing and Information Technology, 7(2), 165-174.

Tong, A.K.Y., & Agius, H.W. (1999). ARISTOTLE: A multimedia-based intelligent tutoring system for zoology. Journal of Intelligent Systems, 9(12), 107-133.

Tong, A.K.Y., & Angelides, M.C. (1999). Formalising tutoring strategy selection in multimedia tutoring systems. Information and Software Technology. 41(2), 67-90.

Tong, A.K.Y., & Angelides, M.C. (2000). An empirical model for tutoring strategey selection in multimedia tutoring systems. Decision Support System.

Wenger, E. (1987). Artificial intelligence and tutoring systems. Los Altos, CA: Morgan Kaufmann.

Woolf, B.P., & Hall, W. (1995). Multimedia pedagogues--interactive systems for teaching learning. IEEE Computing, May, 74-80.

Zellweger, P., Regli, S.H., Mackinlay, J., & Chang, B.-W. (2000). The impact of fluid doucments on reading and browsing: An observational study. Paper presented at the CHI 2000, The Hague, The Netherlands.

Figure 2 A conceptual representation of a Learning Goals SEM
ENTITY OF INTEREST:           cheetah
GOAL 1:                       SPECIALISATION OF 1 (1,2,3)
GOAL 2:                       HAS PART 2 (M1,M2,M3)
GOAL 3:                       NOISE (M1,M2,M3)
GOAL 4:                       ABLE TO 1 (M1,M2,M3)
GOAL 5:                       DISTRIBUTION 1 (1,2,3)
Links to associated Teaching  Links to associated
Strategy SEMs                 Subject Domain SEMs