Massive Open Online Courses (MOOCs) have spurred considerable interest, with several researchers from the fields of education and ICT working actively on producing high quality products and a highly improved learning experience for students of these courses. Yet, a pedagogy and set of learning outcomes designed for students in one setting (e.g., country, level, subject), are often not appropriate for students in different settings, mainly due to differences in culture, language and terminology and educational backgrounds. Moreover, even for students sharing the above characteristics, differences in their personal learning styles, strengths and weaknesses, mean static MOOCs can never attain the same level of rapport with the student as a teacher.To this end, PRIMES will deliver: (a) a platform allowing for automatic multi-level personalisation of MOOCs, and (b) a first set of courses provided as a proof-of-concept. For every student registered with PRIMES, a detailed educational and social profile will be created, encoding the students’ educational background, academic interests, and possible future aspirations, along with their attainment in several courses and grasp of threshold concepts. Course material in PRIMES will be available in several languages (through subtitling and/or dubbing) and formats (e.g., video, audio-only, braille, etc.). Each course and lecture will be annotated, to allow for selective extraction of self-contained excerpts of sessions/lectures with specific (language, theme, level, difficulty, etc.) characteristics. PRIMES will then employ novel machine learning and recommendation algorithms, taking into account the individual student profiles and the courses/sessions on offer, to produce personalised recommendations to further each student’s progress and attainment; our platform will be able to extract and recommend, for each subject and intended learning outcome, parts of lectures and/or courses, along with matching (formative and/or summative) assessment tasks, matching the current level of attainment of each student, while covering the areas where the student faced difficulties in grasping or applying the corresponding knowledge, methods and techniques.

Although most HE institutions have embraced the potential of e-learning methods and have invested in technology-enhanced learning environments and tools, we do not have a clear picture of students’ online learning habits. Moreover, e-learning so far has not received much attention within quality assessment procedures. The understanding of concrete learning behaviour and uses of electronic courseware and online resources is an important prerequisite to assess the quality of autonomous, lifelong learning. Another challenge are the high dropout rates associated with e-learning, not in the least where MOOCs are concerned. Among numerous other variables, an important factor is the lack of engagement and motivation, when students don’t know how they are progressing and what their peers’ achievements over time are. Students involved in e-learning often have a limited knowledge of their own learning habits and which rate of studying with the online material is required. To succeed in (semi-)autonomous learning, however, a higher level of self-regulation is needed.This project addressed the Erasmus + challenge of raising the quality of education through the use of learning analytics. Learning analytics is a new and promising research field which can be defined as “the measurement, collection, analysis and reporting of data about learners in their context, for purposes of understanding and optimizing learning and the environment in which it occurs” (Siemens et. al., 2011). The recent evolution of web-based learning and the possibility of tracking students’ online behaviour offers promising new ways of measuring actual self-study activities. This project aimed to establish a clear image of how higher education students in different European countries learn online. The goal was to map existing learning patterns in 4 different types of online language learning and teaching and maths courses and to feed back this new knowledge to the most important educational actors themselves, being the students and their lecturers. The learning analytics approach was bottom-up, taking the perspective of the learning process, focusing on the courses used and the students’ learning trails through these courses. Process mining techniques were used for the analysis of the data. Therefore, a complimentary and cross-disciplinary consortium of teams from three universities and a private open source company was set up.First, this consortium implemented tracking of learning data based on the new Experience API standard for interoperability with other learning environments and reporting tools. A piloting phase was organised to check the technical & pedagogical validity of the data tracking & data analysis instruments.In the main data collection phase, the learning behavior of several student groups enrolled in distance learning or university programmes with an important self-study component was tracked during one semester. The data were collected in a central repository (Learning Record store). An important point of concern in the project was the privacy of the students who were monitored. Participants were asked to give their consent to collect and use their data for the aims that were clearly described. The data was kept and analysed anonymously and the EU data protection directive was taken into account.In the data analysis phase, the processes of autonomous learning were mapped and compared to the intended pedagogic objectives of the tools. Patterns of learning behaviour were detected, leading to different user profiles and feedback about used learning resources.Finally, the project developed and implemented data visualisation tools in the form of learning dashboard applications for students and for teachers. Special care was given to the ease of use of the dashboards for non-specialist users. These applications allowed the students to understand how they learn online, to follow up their progress but also to compare their profile to user patterns of their peers (optional). Educators got dynamic and real-time overviews of how their students were progressing, which students were potentially at risk of dropping out or of failing for the course and which parts of the courses caused specific difficulties and/or required more feedback.The project also developed a generic xAPI model for implementing learning tracking in interactive language learning tools, which we hope will be reused in different educational settings, countries, courses. The project outputs were used by or presented to the student and instructor target groups but more generally also to all stakeholders in the field of educational innovation and research on a European level. All technologies, algorithms, reports, guidelines, recommendations were put at their disposal under open licenses via www.project-vital.eu.