Orchestrating Learning using Adaptive Educational Designs in IMS Learning Design Marion R. Gruber, Christian Glahn, Marcus Specht & Rob Koper CELSTEC, Open University in The Netherlands, Valkenburgerweg 177, 6419AT Heerlen, The Netherlands {marion.gruber, christian.glahn, marcus.specht, rob.koper}@ou.nl Abstract: IMS Learning Design (IMS LD) is an open specification to support interoperability of advanced educational designs for a wide range of technology-enhanced learning solutions and other units of learning. This paper analyses approaches to model personalised learning experiences with and without explicit adaptive features in IMS LD. The paper has two main parts. The first part analyses the relation between orchestrating learning and IMS LD���s semantic features. The second part compares modelling strategies for educational designs for personalised learning in non-collaborative learning units using IMS LD Level A and IMS LD Level B features. The analysis is based on two worked-out IMS LD units. The paper concludes with a comparison of the two modelling approaches and addresses gaps when integrating adaptation concepts at the levels of the specification. Keywords: adaptation, IMS Learning Design, educational design, personalization, orchestrating learning Introduction IMS Learning Design (IMS LD) is an open specification to support interoperability of advanced educational designs for a wide range of technology-enhanced learning solutions and other units of learning (UoL). The specification has been released in 2003 [1] and has been subject to a broad scientific discussion [2, 3]. Recent developments in the context of the TENCompetence project [4] have led to an improved set of tools and services for modelling and deploying educational designs in IMS LD [5] These developments extend the perspective on the capabilities of IMS LD and allow a better analysis of related educational design approaches. IMS LD provides a semantic framework to formally express educational designs and to model learning processes with innovative technologies. Through scaffolding key parameters of learning IMS LD provides a framework for orchestrating learning processes and personalising competence development. The emphasis on modelling learning processes makes IMS LD a potentially useful framework for tackling the research challenges related to orchestrating learning that have been identified by the STELLAR project [6]. However, only a few research publications have addressed

2 Marion R. Gruber, Christian Glahn, Marcus Specht & Rob Koper personalisation and adaptation with IMS LD in the past years [7, 8, 9]. Related research focused entirely on semantic structures of IMS LD for conditional modelling. This focus is to narrow for reflecting the full potential of IMS LD's semantic framework for orchestrating learning. This paper analyses the application of the conceptual structure of IMS LD for modelling personalised learning experiences. This analysis is based on the recent discussion on personalisation and adaptation in technology-enhanced learning using IMS LD, which is reflected in the background section. It is followed by a section that identifies the research question for this paper and identifies two underlying analytical problems: the relation of IMS LD to orchestrating learning and practical approaches and limitations of IMS LD for modelling orchestration scripts. The section orchestrating learning analyses the dimensions of orchestrating learning based on selected literature. It is followed by the analysis of what semantic features IMS LD refer to the different dimensions. Based on these insights two variations of a UoL analyse the modelling strategies at the different complexity levels of IMS LD. Finally, this paper compares the different approaches and discusses gaps that were identified with respect to the orchestration dimensions. Background Personalisation is increasingly important in technology-enhanced learning. However, personalised learning is not unambiguous. Two general viewpoints on personalisation can be identified. The first viewpoint defines personalised learning as individualised and tailored educational experiences [10]. The personal dimension in this viewpoint is directed towards facilitated educational processes that are unique to a learner. The second viewpoint emphasises the personal relevancy and involvement of individuals in learning processes [11]. From this perspective, personalised learning refers to those processes that support learners to take responsibility and control over their learning and enable them to reflect on the learning on a meta-cognitive level. The two perspectives on personalisation are not mutually exclusive: learner-controlled learning processes may lead to unique learning experiences and automatically adapted educational environments may support deeper learning experiences that allow learners to feel more responsible for their learning. However, learner control can be provided in mass education and fully tailored educational processes can be provided without leaving any control to the learner. Dron [12] argues that personalised learning does not require that learners have full control over their learning, but it requires that some control is left to the learners. Based on this premise, educational designs for personalised learning refer to those educational designs that enable learners to control and regulate their learning based on predefined tasks [13]. Educational designs are adaptive if their task arrangements can reflect the learners' control decisions. Such designs are orchestration scripts for the educational practice. Orchestrating refers to educational practices that organise and arrange learning activities and learning environments to guide learners through learning processes. In this sense the orchestration refers to control and regulation mechanisms for framing parameters of learning processes [14].

Orchestrating Learning using Adaptive Educational Designs in IMS Learning Design 3 From an adaptive educational hypermedia perspective Brusilovsky [15] distinguishes between adaptive presentation and adaptive navigation support to categorise the different approaches for personalising and adapting information. Additionally, a third category has to be included: adaptive sequencing. Although adaptive sequencing is sometimes categorised under adaptive navigation support, adaptive sequencing focuses on the arrangement of processes whereas Brusilovsky���s definitions of adaptive navigation support focuses on navigational features at information level. At this level orchestration scripts define the rules for using and combining the different approaches. IMS LD provides a generic and flexible language to model educational designs in a machine-readable format. As an open specification it allows to prepare orchestration scripts for technology-enhanced learning. One of the requirements for IMS LD was that the language can describe personalization aspects within an educational design, ���so that the content and activities within a unit of learning can be adapted based on the preferences, portfolio, pre-knowledge, educational needs and situational circumstances of the participants in educational processes. In addition, it must allow the designer, when desired, to pass control over the adaptation process to the learner, a staff member and/or the computer.��� [16, p. 14] IMS LD has three complexity levels [17]. 1. IMS LD Level A provides the core elements for creating educational designs. Conditional rules for controlling the learning process are limited to basic aspects of sequencing educational processes. 2. IMS LD Level B adds complex conditional elements that allow learner modelling, fine grained process control, and offers personalisation and adaptation features. 3. IMS LD Level C extends the process model by a notification feature that allows to model event-based processes. Previous research on modelling adaptive educational designs with IMS LD has analysed approaches for modelling units of learning with basic adaptive and personalisation features. These approaches have been primarily guided by user- modelling and adaptation concepts from an adaptive hypermedia research perspective that is discussed by Aroyo et al. [10]. Paramythis and Loidl-Reisinger [18] reflect personalisation at the level of properties and conditions. The authors analyse these features for authoring fully dynamic adaptive processes. The authors conclude that IMS LD is most appropriate to transfer snapshots of adaptation logic between systems. Berlanga and Garcia [7] focus on content adaptation using meta-data fragments to define templates for learning resources. The arrangement of the resources is entirely based on the sequencing features of IMS LD. Instead of utilizing higher order control features of IMS LD for adaptation, the authors propose to store rules for adaptation separated from the educational design. Specht and Burgos [9] analyse what types of adaptation can be modelled using the language features of IMS LD. Their analysis includes partial user- modelling using properties and calculations as well as the application of conditions for adapting aspects of the educational design to the needs of the learner. The authors identify that IMS LD provides strong support for modelling adaptive sequencing, adaptive content presentation, and adaptive navigation support. Van Rosmalen et al. [19] analyse the role of IMS LD in the lifecycle of adaptive e- learning courses. The authors propose that the educational design carries only general