Building a knowledge-based recommender for inclusive eLearning scenarios Olga C. SANTOS1 and Jesus G. BOTICARIO aDeNu Research Group. Artificial Intelligence Department. UNED Calle Juan del Rosal, 16. Madrid 28040. Spain {ocsantos, jgb}@dia.uned.es Abstract. When building a knowledge-based recommender along the eLearning life cycle, the following issues have to be considered: a) the user interface design of the tools required, b) the process to design/generate the recommendations, c) the process to select the appropriate recommendations, and d) the management of the users’ interactions. We are defining a user-centered evaluation approach that copes with those issues and drives the recommender building process in three consecutive steps: 1) elicitation of pedagogically sound recommendations validated by users with a collaborative review, 2) acquisition and validation of the user features to select the appropriate recommendations for the current context, and 3) analysis of the recommendations provided and evaluation of their impact on the user. Keywords. Recommender systems, Life Cycle of eLearning, Accessibility. Introduction Building learning systems that care about both learners and tutors and have a good understanding of the variety of learning contexts is a complex task. However, the goal should not be to focus on the system itself, but on fitting this system into the full life cycle of eLearning in a way that provides a personalized support for its users. The real drivers of the learning process are the users and their evolving needs. These needs can be of varied nature. Users should not only be supported in their learning needs, but also on the accessibility preferences in accessing the course contents and carrying out the course activities. Our approach is based on the idea of combining design and runtime adaptations. According to this approach, adaptations should be applied along the full life cycle of eLearning making a pervasive use of standards to support users in the process. The idea behind is that adaptation is not an idea that can be plugged in a learning environment, but a process that influences the full life cycle of learning, which consists on four steps where the user (and not the system) is the focus. Moreover, this user-centered approach relies on an appropriate usage of the technology that removes the accessibility barriers that users with disabilities may face if the appropriate support is not provided. The goal is to provide an inclusive and personalized support. 1 Corresponding Author. In this context, our research is focused on a recommendations model [1] that influences the life cycle of eLearning and can be used to build a knowledge-based recommender system to provide adaptive capabilities to existing learning management systems (LMS). To achieve our goal, we have proposed a methodological approach to drive its development and validation [2]. In this paper, we comment on the four issues that we have identified as relevant when building a knowledge-based recommender and how they are mapped against the life cycle of eLearning. 1. Relevant issues and steps of the recommender building process In the educational domain, recommendations should be pedagogically guided instead of by learner’s taste [3]. Thus, a knowledge-based recommender, which uses knowledge about users and the domain, [4] was chosen as the basis for our approach. That knowledge-based approach aims at generating suitable recommendations and reasoning about what elements of the domain meet the user’s requirements. We came up with the following issues when building a knowledge-based recommender for inclusive eLearning scenarios:: • The user interface design of the tools. There is a need for an authoring tool to create/view/update/delete recommendations by tutors and a player to present them to the user when appropriate, from within the LMS user interface. • The process to design/generate the recommendations. A twofold approach is considered. On a first stage, pedagogically-oriented recommendations can be designed by human experts who have experience in online courses designed for all. In this stage, a bank of recommendations can be constructed by individual proposals of the experts and a peer-reviewing process of the proposed recommendations applying user-centered design methods. On a second stage, the recommendations initially proposed by the experts can be tuned and complemented from the usage experience in the course, by applying some artificial intelligent techniques to the interaction data tracked by the LMS modify the values of the elements of the model initially provided by the experts. Moreover, these algorithms can also identify troublesome or promising situations and suggest the tutor to think of appropriate recommendations. • The process to select the appropriate recommendations. This process should take into account the user features, which include individual preferences (such as learning styles or accessibility preferences) and the progress in the course (in terms of competences achieved), as well as the particular context in which the user is (e.g. the learning objectives that are being worked in the course, users on-line, the capabilities of the device used). • The management of the users’ interactions. The users’ interactions can be gathered in different ways. On the one hand, explicit information obtained from surveys, assessments, questionnaires and tests, or ratings given to objects available in the LMS. On the other hand, implicit information tracked by the system on the actions done by the user on the elements of the LMS. With these issues in mind, we are defining a user-centered evaluation approach that drives the construction of the recommender system along the different phases of the eLearning life cycle through the following steps: • • • At design time, elicitation of pedagogically sound recommendations following user-centered design methods (questionnaires, interviews, observations). As a result of this process, a bank of recommendations for eLearning scenarios validated by the users with a collaborative review can be obtained. An authoring tool to manage the design of the recommendations is needed. At runtime, acquisition and validation of the user features to select the appropriate recommendations for the current context. The evaluation of the recommendations followed right on time can be offered to the user optionally and non-intrusively. A player integrated in the LMS is required to present the recommendations to the user. Once the course has ended, validation of the recommendations provided and the impact on the user. This requires the analysis of the interactions of the user at runtime, by i) analyzing the outcomes in evaluations within the course, ii) querying the user satisfaction, iii) analyzing the interaction data. With these data, compute user indicators regarding the collaboration and knowledge level. This can also be used to feed back the designed recommendations with the results from the course experience to modify accordingly the values of the elements of the model or even generate new rules for recommendations. 2. On going works We are currently researching 1) what user-centered design methods are the most appropriate to elicit pedagogically sound recommendations to validate the recommendations model defined and 2) what user-centered evaluation methods are the most appropriate to evaluate the impact of a recommender system in inclusive eLearning scenarios. Acknowledgements The work presented here is framed in the context of the projects carried out by the aDeNu Research group. In particular, the EU4ALL (IST-2006-034478) project funded by the European Commission and A2UN@ (TIN2008-06862-C04-01/TSI) project funded by the Spanish Government. References [1] Santos, O.C. and Boticario, J.G., 2008. Users' experience with a recommender system in an open source standard-based learning management system. In proceedings of the 4th Symposium of the WG HCI&UE of the Austrian Computer Society on Usability & HCI for Education and Work (USAB 2008). [2] Santos, O.C. 2009. Recommendations support in standard-based learning management systems. Proceedings of the 14th International Conference on Artificial Intelligence in Education (AIED 2009). [3] Drachsler, H., Hummel, H. G. K., Koper, R. 2007. Personal recommender systems for learners in lifelong learning: requirements, techniques and model. International Journal of Learning Technology. [4] Burke, R., 2002. Hybrid recommender systems: survey and experiments. User-Modeling and UserAdapted Interaction, 12, pp. 331-370.