Dynamic Composition and Personalization of PDA-based eLearning – Personalized mLearning

With the increasingly demanding schedules of formal and casual learners and the proliferation of terminal devices available for content delivery, learners are demanding ubiquitous access to their learning. The multi-modal nature of the variety of devices employed by these learners will necessitate that learning experiences can be started on one device and transitioned seamlessly to another as the learner changes the context of their learning. This movement of learning experience between devices, which may have dramatically different capabilities, presents a usability challenge to ensure our learners do not feel their learning experience has been adversely affected. Personalized eLearning offerings can be tailored not only to the learner, but also to the device they are using. Generally, the evaluations that have been carried out of personalized eLearning systems have concentrated on systems that are delivered via devices, such as desktop computers, that have generous screen real estate available for rendering the personalized courses. When screen real estate and characteristics such as bandwidth are limited appropriate personalization becomes much more important. This paper examines the usability of dynamically composed eLearning experiences on Personal Digital Assistant (PDA) devices. INTRODUCTION According to [Kukulska-Hulme, 02], usability is typically defined as being the effectiveness, efficiency and satisfaction with which specified users can achieve identified goals in particular environments. In order to get away from the traditional “one size fits all” attitude to eLearning, where every user is presented with the same content in the same order, an approach which takes into account the user’s aims and prior knowledge, encompassing their tasks, pre-empted the development of systems which would be capable of adapting to individual users. These systems are known as adaptive hypermedia systems. [Brusilovsky et al, 02] outlines that a distinctive feature of an adaptive system is an explicit user model that represents user knowledge, goals, interests, and other features that enable the system to distinguish different users. The user model is used to provide an adaptation effect that tailors interaction to different users in the same context. An adaptive system automatically adapts to the user given a user model, whereas an adaptable system requires the user to specify exactly how the system should be different. Adaptive Hypermedia and web systems are essentially collections of connected information items that allow users to navigate from one item to another and search for relevant items. The adaptation effect in this reasonably rigid context is limited to three major adaptation mechanisms Adaptive content selection, Adaptive navigation support and Adaptive presentation. While this paper touches on all three of these mechanisms it is particularly concerned with the presentation of Adaptive Hypermedia. Presentation of content and navigation structures has to be considered in developing an eLearning application that is targeted for use on many types of terminal devices from which the learner may be accessing learning content. Wireless mobile devices are becoming increasingly widespread, with more and more people choosing to access the Internet and various Web services while on the move. This has directly influenced the growth in mLearning [Sharples, 00] where people wish to access eLearning resources from their mobile devices. This opens up a whole new range of design issues for eLearning interface designers .For example, the smaller screen real estate and decreased processing power of PDAs could adversely affect access to eLearning solutions. These, and other, restrictions of such devices make them a challenging environment on which to deliver personalized eLearning experiences. Many of the adaptive hypermedia techniques employed in current personalized eLearning systems will be equally applicable for personalized mLearning. The primary difference between personalized eLearning and personalized mLearning is the medium through which the learning is presented and the presentation of navigational structures and learning content. The mobile nature of mLearning means that the content delivered should not only be tailored towards the learner’s need, but also to their current situation or learning context. In order for this context, and in particular the device the learner is using, to be catered for in mLearning the presentation axis of adaptation should be considered. In many personalized eLearning systems the presentation axis is often seen as less significant when compared to adaptive content selection and adaptive navigation support. In mLearning, however, its significance increases substantially as it is the mechanism through which the personalized learning experience is formulated. This paper describes the architecture and implementation of dynamically composible eLearning courses for PDAs. The first section of this paper will discuss some current Adaptive Hypermedia Systems, and also some systems that have taken into account some of the issues dealing with automatically configuring interfaces for specific devices. The architecture of the Adaptive Personalized eLearning Service (APeLS) [Conlan et al, 2002] that has been developed in the Knowledge and Data Engineering Group at Trinity College, Dublin will then be described briefly, including the specific changes made to a personalized Structured Query Language (SQL course running on APeLS to enable it for Personalized mLearning. The penultimate section will discuss the initial experiment that was conducted to evaluate the feasibility and usability of using APeLS to deliver mLearning using mobile-enabled PDAs. The paper will conclude with a Future Work section. STATE OF THE ART The goal of this section is to provide a background in which this trial and evaluation of eLearning on a mobile device, or mLearning, has been carried out. It briefly reviews Adaptive Hypermedia (AH), including a discussion of a number of Adaptive Hypermedia Systems (AHS). It also describes some of background relating to designing content and interfaces for PDAs. Adaptive Hypermedia Brusilovsky [Brusilovsky, 96] provides the following definition for adaptive hypermedia: “By adaptive hypermedia systems we mean all hypertext and hypermedia systems which reflect some features of the user in the user model and apply this model to adapt various visible aspects of the system to the user. In other words, the system should satisfy three criteria: it should be a hypertext or hypermedia system, it should have a user model, and it should be able to adapt the hypermedia using this model.” To achieve this view adaptive hypermedia combines research efforts in the area of user-modeling, adaptive systems and hypermedia. A hypermedia system is a collection of linked nodes (hyperdocuments), which have links through which the student can traverse to other nodes. When applied to eLearning AHS support the premise that different people learn in different ways and at different rates. In AHSs the learning experience can be tailored to their specific capabilities and past history. The techniques frequently employed in achieving this tailored learning are adaptive navigation structuring, adaptive content selection and adaptive presentation. Studies have shown that users of educational AHSs are faster, more goal-oriented and take fewer steps to complete a course. It was also noted that they are less likely to repeat the study of content that they have already covered [Conlan, 00]. A general-purpose Adaptive Hypermedia Engine [De Bra] may be viewed as being comprised of at least the following three elements: • Domain Model – Consists of concepts and the relations that exist between them. Typically the domain model gives a domain expert’s view of the domain • User Model – Consists of relevant information about the user • Adaptation Model – Consists of a set of rules or triggers, which indicate how various user actions update the user model and affect how the pages are adapted accordingly AHSs apply different forms of learner model to adapt the content and links of hypermedia pages to the user. In many AHSs there tends to be separation of the learner model and the content model, but the narrative or pedagogical model is usually embedded in the content or the engine. In these cases, adding new or different pedagogical models to the content model is more difficult and involves a re-authoring of the content model. This results in learning content that is difficult to reuse or an engine that is domain specific. One means of enhancing the educational impact of eLearning courses, while still optimizing the return on investment, is to facilitate the personalization and repurposing of learning objects across multiple related courses [Conlan, 02]. The AHA project [De Bra, 01] was designed to implement a system that extends Web servers with adaptivity functionality in a transparent manner. A module registers users automatically and adapts the content and navigation structure to the information stored in a user model which is maintained by the system. This module is application independent and the source of the content may not necessarily be internal to the system. The AVANTI [Avanti] project was aimed at catering multimedia information to the information needs of different users by adapting the content and the presentation of web pages to each individual user. A domain model, as well as user models which contained information about the characteristics of individual users, groups of users and various usage environments were maintained by the system and used in the adaptation process. Adaptation was invoked at the content level and at the user interface level. This dual adaptation effect allowed the system to optimize the way the content was applied and presented. The AVANTI system was aimed at providing personalized information across a wide range of users, with differing knowledge and abilities. The information it conveyed was hypermedia information about a metropolitan area. The Adaptive Personalized eLearning Service (APeLS) [Conlan et. al., 02] approach taken to implement an adaptive education system is to provide a generic model for integrating the learner model (which describes the pertinent learner characteristics), content model (which describes the pedagogical qualities of the content and the narrative model (which describes a mechanism for combining the content to produce a coherent educational courseware component). To achieve a high level of adaptivity the models require a high level of detail. Any models which are used as the basis for the learner or content models will need to be augmented to support the level of adaptivity hoped to be achieved. The principle metadata which needs to be added to all the models are more focused pedagogical elements. These should be pertinent to how the adaptive engine can combine the models to provide effective and coherent course material to the user. Elements containing information about the users’ preferences should be reflected in all the models. The adaptive engine provides the facilities for reconciling the content, learner and narrative models to produce individualized content. The system is an adaptive metadata driven engine that composes, at runtime, tailored educational experiences across a single content base. The TCD system has a clear separation of content, learner and narrative models, and a generic adaptive engine that employs a multi-tiered AI model to achieve effective adaptation to the learners requirements. The approach is to have very little semantics actually embedded in the adaptive engine itself. The adaptive engine reconciles, at runtime, the personalized course. The dynamic building of the course is controlled by each learner via appropriate pedagogic instruments. This approach enables multiple narrative models to be constructed to fulfill different learning goals, while these goals may be achieved from a common repository of content. User Interface Design Usability is defined as the measure of the quality of a user’s experience as they interact with a system [Usability]. In order to establish basic human factor goals, a study of the user community and their set of tasks is necessary. Five of these which are important to evaluate are [Shneiderman, 98]: • Ease of learning: How quickly can the user learn to use the interface? • Efficiency of use: How fast can a user perform his/her tasks once they know the system? • Retention over time: Can users who have used the system before instinctively use it when they come again? • Error frequency and severity: How often are errors made by users and how easy is it for them to recover? • Subjective satisfaction: How much does the user like using the system? Half of all software support calls are due to poor usability [Nielsen, 00]. Users’ satisfaction with a system increases when that system is easy to use. Ease of use increases when the system follows a consistent usage pattern and provides a familiar lookand-feel to the user. Users don’t like to have to undergo training in order to be able to use a system. Systems must focus on user satisfaction, allowing the users to be able to relax, enjoy and explore the various features the system offers [Shneiderman, 02]. PDA Interface Design It was reported in [comScore, 02] that 9.9 million Internet users in the U.S. use a personal digital assistant (PDA) or cell phone to access the Internet. [CompIndAl, 02] projected that 48% of Internet users in 2005 will be mobile Internet users. The World Wide Web consists of a huge number of pages for people to browse for work or pleasure. Unfortunately, most of these pages are designed to be displayed by computers with a large and high resolution monitor [Chu, 01]. When viewing pages on a device with a small and low resolution screen, display problems arise, making the experience very unsatisfactory for the user. Therefore, the presentation of information on a small screen device provides many usability challenges. Some which have to be taken into account, according to [Usability] are as follows: • Instructions and other similar text should be used sparingly and only when necessary • Links should be brief and contain only necessary key words • As with web pages, specific options should be presented before general options • PDA users expect to find web-like interfaces on their handhelds • Due to low screen resolution and small screen size, long narrative descriptions are not effective The Dygimes [Coninx et. al., 03] project stems from a desire to create multi-device interfaces. This results from several observations of the growing trend of mobile devices. One of these observations is that the reuse of existing interfaces on new devices is problematic, specifically when the interface is constrained by hardware or software issues. The user also grows familiar with a “brand” of user interface design, which puts a pressure on user interface designers to create a consistent look and feel across several devices. In the framework developed in Dygimes, the design of the interface and the application code is separated, with smooth integration of the functionality and the interface supported. Therefore the resultant interface can be used on many devices, using standard functionality. This allows flexible reusability of existing designs. In the framework, the user interface is less dependent on device-specific properties, and is implemented through the use of XML-based UI descriptions. Task models, interaction models and context-dependent mapping rules are combined with these high-level XML-based UI description in order to generate interfaces which are device and system independent [Dygimes]. The Pebbles project [Pebbles] is exploring how handheld devices, such as Personal PDAs and mobile phones can be used if they are interacting with other electronic equipment, such as other handheld devices, appliances or normal desktop PCs. Appliance are becoming more advanced as more and more functionality is added. With the increase in functions often the interface to the appliance becomes more complex and harder to use. The Pebbles project proposes that separating the interface from the appliance may help the users to interact with the appliance. The idea is that every user would carry a personal universal controller (PUC) [Nichols, et. al, 02], a device which would allow the user to interact with all of the appliances or services in his/her environment. In order to control an appliance, the PUC communicates with the appliance, downloads the details of the functions of the appliances and automatically generates a remote control that is suited to the user and his/her PUC. ARCITECTURE The Adaptive Personalized eLearning Service (APeLS), and more specifically a personalized SQL course based on APeLS, is used as the basis for the personalized PDA-based eLearning System trialed and evaluated in this paper. This section briefly describes APeLS and the specific implementation required to apply the multi-model, metadata driven approach for adaptive PDA-based eLearning. APeLS is an adaptive system that is designed in accordance with the multi-model, metadata driven approach [Conlan et al, 02]. It supports the separation of the models that impact upon the personalization process. While the approach supports a multitude of models the three core models of learner, content and narrative are usually included. The learner model describes information about the learner that is pertinent to the personalization of the material; the content model describe the learning resources and the narrative model describes the concepts and pedagogical strategies that may be employed to teach a specific domain. It is important to note that the models remain discrete and separate. This, and the candidacy abstraction layer [Dagger et al, 03], facilitate their reuse. Fig 1. Architecture of APeLS The ability to describe the narrative of a course independently of the content used to teach it enables the same narrative to be used in different versions of the course. For example, the PDA-based SQL course and the standard desktop version share the same narrative. The candidacy abstraction layer enables multiple versions of content for each concept in a course. Candidate selection is performed just in time as the course is being delivered to the learner. A separate model is employed to facilitate this selection process – the environment model. This model primarily describes the device the learner is accessing the personalized course from. This enables the candidate selection process to choose the most appropriate piece of learning content for a given concept and for the device the learner is using. Knowledge of the learner’s device also enables the most appropriate navigation paradigm to be employed. Figure 1, below show the APeLS architecture. Fig 2: Sample Table of