【文章內(nèi)容簡(jiǎn)介】 dependent. The same content must be reedited for different means (printed, webdelivered, and so forth). Furthermore, similar contents are sometimes developed from scratch. Finally, administrative and monitoring tasks are done with very, little support. For instance, current environments lack support for assessing users’ behaviors based on their reasoning paths.In order to tackle these problems the “Innova Group” at UNED has developed a dynamic web framework called ALF [2] (Active Learning Framework – educational version). Based on our two years’ experience with more than fourteen thousand elearning users (Boticario et al. , 2001), ALF provides a new set of tools for supporting elearning scenarios. Instead of nonsupported collaboration, several Computer Support Collaborative Learning (CSCL) frameworks have been integrated on the platform. ALF users are involved in a working space of Virtual Educational Communities, which surpass the organizational constraints of classical web courses. Nonassisted guidance has bee scheduled scenarios of learning activities. Wellknown static indicators for learners’ performance assessment (like visited pages and links to accessed materials) have bee tracking facilities where reasoning paths can be easily accessed from users’ interactions. Finally, in order to improve reusability and maintenance, course contents are based on XML specifications.Moreover, we are involved in a more ambitious research scenario where the learner is dynamically guided based on his/her experience and where other users’ related experiences could affect learning experiences as a whole. Thus the WebDL [3] (Web Distance Learning) research group has being working on an adaptive framework where the user modeling system is based on a bination of web mining techniques and ensembles of classifiers for machine learning tasks. This research is included as an essential ponent of a European funded project, ALFANET [4] (Active Learning for Adaptive Internet, IST200133288).In this paper, first a brief review will be done of the problems that have been detected in online training. Then in order to tackle these and other related issues, our approach, the objectives, educational model, a collaborative scenario, course development and experiences after two years in operation with ALF will be presented. The paper will end with a series of conclusions and some future works.2. PROBLEMS DETECTED IN ONLINE TRAININGNumerous studies have been done that reveal the difficulties detected in the first wave of virtual learning environments. In fact, the general view is that there are few socalled best practices (Bonk, 2002). What is particularly surprising is the high dropout rate on online courses, mainly due to not very explicit instructions and no incentive.In depthstudy of online training has detected issues of interest in different areas: cognitive, pedagogical and usability (Chen, 1997). It is thus essential to promote scenarios that guarantee the natural learning model (Schank amp。 Cleary, 1995), where the student learns by doing. Educational frameworks have also been established where active learning, collaboration an the design of munication services for each of the difficulties detected are vital aspects (Boticario amp。 Gaudioso, 2000). Moreover, a direct correlation has been established between the key aspects in the usability field and their effect on learning (Mehlenbacher, 2000).We should also mention the studies that consider the issues related to the management, including the changes in mentality that the different protagonists in the process should experience: lecturers, students and the institution. Without assumed new possibilities and restructured work deployment of elearning cannot be successfully ensured (Dobson, 1998). Related to this issue, problems associated with material development management and faith in standards for solving these difficulties have also been highlighted (Bonk, 2002).On the other hand, often the contents provided by an educational web site are fully static. The navigati