Activity-Based Learner-Models for Learner Monitoring and Recommendations in Moodle Beatriz Florian1, 3, Christian Glahn2, Hendrik Drachsler2, Marcus Specht2, Ram��n Fabregat1 1 Institute of Informatics and Applications (IIiA), University of Girona, Girona, Spain bflorian@eia.udg.edu, ramon.fabregat@udg.edu 2 CELSTEC, Open University of the Netherlands, Heerlen, The Netherlands {christian.glahn, hendrik.drachsler, marcus.specht}@ou.nl 3 EISC, Universidad del Valle, Cali, Colombia beatriz.florian@correounivalle.edu.co Abstract. In technology-enhanced learning, activity-based learner models can provide evidence for competence assessment. Such models are the foundation for learning and teaching support, such as: adaptation, assessment, and competence analytics, recommendations, and so on. This paper analyses how to construct activity-based learner models based on existing data in the Moodle learning management system. Based on the activity theory model and the actuator-indicator model, aggregators of learner activities for different activity types were implemented in Moodle. This requires the consideration of the social roles in a course, in order to enable adaptive views for learners and instructors on the stored activity information. The implementation showed that Moodle stores information about course activities that requires filtering before it can get used for higher level processing. The social planes in Moodle reveal a higher complexity than it has been previously described by theories of classroom orchestration, such as actors who are no longer present in a course. Keywords: Activity-based learner models, Moodle, Learners tracking, Learning analytics, Competence assessment, Indicators, TEL recommender systems. 1 Introduction In Virtual Learning Environments (VLE), rich user models based on activity traces are required for different types of personalized learning support such as: analytics of competences��� development, activity-based smart indicators and recommendations. The general process could be described as the record of data interactions and outcomes of activities, the semantic interpretation of collected data and their analysis to produce appropriate support responses. The Learning Management System (LMS) Moodle records a broad range of individual learners' traces in real time. For adaptive systems these interaction footprints had been used to produce new learning paths [1]. However, this data is not easily accessible to Moodle users and the analytics based on this data are hidden to

2 Beatriz Florian1, 3, Christian Glahn2, Hendrik Drachsler2, Marcus Specht2, Ram��n Fabregat1 students and poorly provided to teachers. This hinders practitioners to apply real time educational data in their practice. Although the increasing amount of learning analytics (LA) related papers published nowadays [2], research contributions about how automatically collected activity traces could be effectively used for supporting learning process are rare. The research presented in this technical-design paper is the foundation for future work on applying the concepts of learning analytics for competence assessment and recommendations. This contribution focuses on the concept of social planes and analyses social perspectives for accessing Moodle���s tracking data. The paper analyses the reuse of Moodle���s tracking data for learner and group modelling. Moodle activity log is used to build rich learner models based on learner activities within a social context. Therefore, this paper addresses social facets of Moodle���s log data and their implications for learning support. We have implemented an architecture that based on the Activity Theory model [4] and the actuator-indicator model [3] to have a flexible and extendable interface to Moodle���s tracking data for different roles in learning processes. The resulting framework transforms the collected data into learning analytic information for the social planes ���self���, ���peers��� and ���class���. Furthermore, the implementation considered different social perspectives on the data. At this point these perspectives are coupled to the course roles ���student��� and ���teacher���. The concept of social perspectives on tracking data is useful to integrate aspects of privacy and data-protection while modelling learning analytics functions. The social contexts of the initial implementation were grounded on social planes that have been identified by prior research on instructional design and on collaborative learning. First tests revealed that the given conceptualisations of social planes in education did not fully describe the tracking data of the Moodle logs. This contribution has the following structure. Section 2 outlines the state-of-the-art about technology-supported competence-assessment. Section 3 presents the research objective. Section 4 analyses an implementation of activity-based learner models and learning analytics in Moodle. The implications of the prototypical implementation for technology-supported assessment are discussed in section 5. Section 6 presents the findings. Perspectives towards recommendations support are analysed in section 7. Finally, this paper concludes with an outlook on future research in section 8. 2 Background Most VLEs already provide functions that can be used for supporting activity-centred learning, but the related information is commonly unavailable in a structured form. Semantically structured learner models are required in order to provide technological support for more activity-centred assessment types. An activity-based learner model creates a semantic structure of dynamically generated learner properties that reflect observed actions of a learner. Activity-based learner models are a prerequisite for activity-centred assessment and process support for competence development. Contemporary competence models such as PALO [5] and EQF [6] describe proficiency levels of competences according to types of activities that learners are