IMS LD reusable elements for adaptive learning designs

This paper presents an approach to designing adaptive learning environments based on IMS LD, which separates its elements (i.e. objectives, prerequisites, method, learning activities, adaptive rules, personalization properties, etc.) in order to use them in different Learning Designs and enforce their reusability and exchangeability. Moreover, it briefly presents an authoring tool under development to define adaptive learning designs compliant with IMS LD.


INTRODUCTION
In Chapter 12 Towle and Halm (2005) explain the modelling of Adaptive Learning using IMS LD (2003) by inserting the adaptive logic within the IMS LD element <method>.They exemplify how three kinds of adaptive strategies can be modelled using IMS LD.These strategies include synchronous vs. asynchronous interactions, rule-example vs. example-rule presentation of the content, and feedback adaptation.
Subsequently, the authors point out that one of the limitations of IMS LD for adaptive learning is its "manifest-centred" schema.That is to say, all the necessary information for interacting with a Unit of Learning (UoL) is inside the manifest of the UoL.For them, the problems of this representation are (p.

225):
(1) The difficulties inherent with rule interactions for multiple characteristics.
(2) Once delivered, manifests cannot be changed to take advantage of new adaptive strategies.
(3) The same strategy is encoded in multiple manifests, causing redundancy in authoring and storage.
(4) The knowledge about learning objects is often embedded in the manifest, and not accessible through metadata for use in new or arbitrary strategies.
The authors argue that a solution to tackle these problems is to move from "manifest-centred" schema, which forces static adaptivity, to a "server-centred" approach.This can be done by removing the adaptive logic from the manifest and using a LD player as a client (or agent) that communicates to the server what the learner has done.The server, then, will send back to the client the ID of the most appropriate next activity to follow.However, problems 3 and 4 are not because of the specification, but the way the specification is used.If repositories are not used, or the creation of learning activities or methods has to be done for each Learning Design, then redundancy, inefficiency and lack of reusability are a fact.
In this paper we present a proposal we are developing to tackle these two issues.We claim that if IMS LD elements (i.e.learning objectives, prerequisites, learning activities, acts, plays, conditions, and so on) are defined as independent elements, they become reusable and exchangeable elements.In this contribution we present this approach, and outline related work.

ADAPTIVE LEARNING DESIGN (ALD)
We are investigating if a learning design with adaptive characteristics, or ALD, can be reusable and exchangeable among different courses, contexts, and applications.
An ALD is a UoL that contains personalized behaviour in order to provide each student with a learning flow adequate to her/him characteristics (Berlanga and García, 2005).In order to permit reutilization, ALDs are semantically structured and designed according to IMS LD.That is to say, ALDs elements are the same as IMS LD elements (with the exception of learning objects that are compliant with IMS LOM, 2001).However, elements are defined and stored as separate components that can be reused an exchanged among different ALDs, learning contexts, lessons, and courses.
The separation between learning activities and their learning resources is a key premise of IMS LD.A learning design can be repopulated with different contents and resources to use it in a new learning context (Richards, 2005), and/or a set of learning activities can be packed in different courses (McAndrew and Weller, 2005).Likewise, there are three kinds of reusability of an ALD: • ALD as a template, where an "empty" ALD is provided in order to fill-in the desired elements (e.g.learning resources, properties, learning activities, conditions, etc.).
• Reusable ALDs, where an ALD is modified in order to suit new settings or contexts.
• Reusable elements of ALDs, where specific components of an ALD are exchanged among other ALDs.

THE LEGO METAPHOR AND ITS ELEMENTS
Since we claim that the separation of elements is crucial in order to support their reusability and exchangeability, the definition of an ALD follows the Lego metaphor.Figure 1 represents this approach.Notice that each type of element (e.g.rules, methods, plays, etc.) should be stored in different file folders that could be handled as repositories of IMS LD elements.For readability reasons, not all relationships among elements are presented; see Table 1 for a full list, including the ID of the element, its name, the elements in which it can be included, the elements it can include, and the elements where a learning object can be attached.
For instance, the learning object LO-1 can be attached to learning activity LA-1 (using the <activity-description> element).Then, LA-1 can be included into activity structure AS-1, which could be incorporated in ACT-1, and so on.In the same way, AS-1 could be included in ACT-2.In this manner different components can be reused and exchanged among different applications and tools compliant with IMS LD, and the definition of a new method of instruction does not imply the creation of learning activities, roles, objectives, etc., that have been created before for other ALDs.Reusability of these elements can take place if, for instance, learning activities LA1.doc and LA2.html are removed and authors repopulate them with the learning activities or resources they want (i.e.ALD as a template).Moreover, the components of this ALD can be reused in other ALDs if, for example, the learning activity L1 (Background of Hypermedia) is included in a different ALD as an activity about the evolution of the World Wide Web, or the condition "IF P-Initial-Knowledge < THRESHOLD" is included in other ALD.Finally, this ALD can be modified for other settings if, for instance, a property that contains the final knowledge of the student (e.g."P-Final-Knowledge") is included, and then used in the conditions section to show complementary learning activities if the "P-Final-Knowledge" value is less than the threshold value.

DEFINITION OF ALDS
In order to define ALDs we are developing an authoring tool.Our objective is to support authors in the creation of learning designs without prescribing any instructional approach, variables or conditions for adjusting learning to students' characteristics.
As a result, we are extending the functionality of a tool for creating hypermedia books called HyCo (Hypermedia Composer) (García and García, 2005) with the intention to use it as the ALD authoring tool editor.
HyCo is a multiplatform tool that supports the creation of learning materials.It has sets of galleries that permit authors to manage multimedia resources and bibliographical references, as well as to generate output files in formats such as HTML, PDF, XML or plain text.The current version of HyCo includes a LOM editor compliant with IMS LOM for defining and modifying the metadata of educational resources.HyCo stores these resources in a repository, in such way that, later on, they can be incorporated into elements of ALDs as prerequisites, objectives, components (i.e.learning activities, activity structures), and so on.
The ALD Editor follows the Lego metaphor explained before.Therefore, each element is defined independently from each other.Within the definition of each element, the interface presents a tab structure to group sets of attributes that might be described to annotate the element.Authors can attach to this definition resources created in HyCo (e.g. a chapter or a hypermedia book), or resources referenced by an URL. Figure 3 shows the HyCo-ALD Editor tab to define the description of a learning activity.The editor presents, when possible, default values and combo-boxes.For instance, Figure 4 shows the tab for including learning objectives into learning activities.It provides a selection list that contains possible learning objectives (i.e.those that have been defined before and are in the learning objective repository).Moreover, it is connected to the HyCo-LOM Editor in order to provide authors with a tool for creating metadata.
Similar interfaces are provided for depicting other elements such as learning objectives, prerequisites, roles, learning activities, and so on.Currently, we are working on analyzing and designing the variables needed to define personalization properties and adaptive rules (i.e.IMS LD Level B).Personalization properties contain information about the users that can be included subsequently into adaptive rules, while adaptive rules are prescriptions defined by authors that will be taken into account to adjust the learning design, and that can be included into learning methods.For instance, returning to the example presented before, Figure 2 includes a personalization property named "P-Initial-Knowledge", and an adaptive rule named "IF P-Initial-Knowledge <THRESHOLD".
We are developing two approaches for defining these elements: one for novice users of IMS LD and other for expert users of the specification.In the former case, we are designing a wizard for defining adaptive techniques, and in the latter, we are developing an expression-builder-tool, based on an "ifcondition-then-action" formalism (Berlanga and García, 2004), that permits authors to include learning design elements, personalization properties, and logical and relational operators.In both approaches, authors will be able to save their adaptive rules and properties in repositories and reuse them in other ALDs.

RELATED WORK
Nowadays, IMS LD tools are in their early stages of development and testing.Reasons for this include the relative novelty of the specification and its high level of complexity.Until now, there are not available user-friendly authoring tools for teachers or non-specialists in the development of learning materials; existing authoring tools have not gone beyond research.Nonetheless, the high number of efforts attempting to develop authoring tools is significant [1].
At present, authoring tools for learning designs compliant with IMS LD can be categorized as: • Basic editors, also known as close from specification: authoring tools which interface follows the specification to help users to create a UoL.Therefore, users should have enough knowledge of the specification.Editors that have been developed in this line include those with an interface that uses the tree metaphor for displaying and handling the specification, as CopperAuthor ( 2005), which uses the Coppercore (2005) engine to display a preview panel of the UoL, and the popular Reload Learning Design Editor (2005).Other basic editors include ASK-LDT (Karampiperis and Sampson, 2005), which has a graphical interface, and the aLFanet LD Editor (Van Rosmalen and Boticario, 2005) that, as part of the aLFanet learning management system, guides the designing process using windows for defining each element of the learning design.
• Advanced editors, also known as distant from specification: authoring tools that "hide" the specification to the final user.Editors developed in this line include MOT+ (Paquette, De la Teja, Léonard, Lundgren-Cayrol, and Marino, 2005), which has a graphical interface for creating courses according to the MISA method, and e-Live LD Suite (eLive GmbH, 2005), which is a commercial product under development that provides users with learning design templates for working with the specification.
Table 2 presents an overview of some IMS LD authoring tools that are leading the IMS LD implementations (Griffiths, Blat, Elferink, and Zondergeld, 2005).The  2, the HyCo-ALD editor falls in the basic editors category.Like the aLFanet LD editor, it uses windows to present the specification, but in a stand alone mode.Moreover, HyCo-ALD is part of an authoring tool for creating hypermedia contents; this permits the inclusion of hypermedia learning resources created in HyCo (e.g.chapters, subchapters, etc.) into learning designs, and supports metadata for learning resources conform to the IMS LOM specification.
Nevertheless, the novelty of HyCo-ALD is its approach for reusing IMS LD components and adaptive techniques in different learning designs.Furthermore, learning activities created in HyCo-LD might be exportable components that will work across different learning systems, and vice versa, HyCo could import and take advantage of learning activities compliant with IMS LD.However, HyCo-LD is still under development and much work has to be done to test if reusability of IMS LD components is possible and to what extent.

CONCLUSIONS AND FURTHER WORK
The separation of IMS LD elements in different repositories is an option to avoid the creation and annotation of the same elements (e.g.learning activities, activity sequences, etc.) for different ALDs.
One step further is that those repositories could be distributed in different servers, and UoLs could include URL references anchoring to adaptive conditions or, as Paquette, Marino, De la Teja, Léonard, and Lundgren-Cayrol (2005) suggest, take out Level B and Level C -and limiting IMS LD to Level A -in such way that adaptivity conditions can be stored outside the host system.These could be solutions to avoid static adaptation forced by IMS LD and complement the server-centred schema suggested by Towle and Halm in this chapter.This paper has presented a proposal for defining ALDs using IMS LD.We are finishing the HyCo-ALD editor and depicting a wizard that will guide non-expert users in the creation of adaptive techniques, as well as an expression-builder tool for supporting authors in the definition of adaptive rules.Subsequently, we will test if ALDs reusability is achievable in HyCo, and then examine their possibilities for exchange among systems or applications compliant with IMS LD.

Figure 2 .
Figure 2. Example of an ALD personalization strategy using IMS LD.Only the parts of the ALD relevant to the adaptive strategy are shown

Table 2 .
table shows their type (basic/advanced editor), IMS LD level of compliance, characteristics of the interface, availability status, and authors.IMS LD authoring tools As shown in Table