Collaborative CBR-based agents in the preparation of varied training lessons

International audienceCase‐Based Reasoning (CBR) is widely used as a means of intelligent tutoring and elearning systems. Indeed, course lessons are elaborated by analogy: this kind of system produces sets of exercises with respect to student level and class objective. Nevertheless, CBR systems always result in the same solution to a given problem description, whereas teaching requires that monotony be broken in order to maintain student motivation and attention. This is particularly true for sports where trainers must propose different exercises to practice the same skills for many weeks. We designed a system based on CBR that takes into account any previous lessons offered and designs new ones so as to vary the exercises each time: this system takes into account the solutions previously proposed so as to avoid giving the same lesson twice. In addition, this system is based on collaborative agents, each taking into account the exercises proposed by others so that each activity is proposed only once during a lesson. A sports trainer tested and evaluated the ability of this system as a means to design varied aïkido training lessons and proved that our system is capable of creating classroom activities that are diverse, changing, pertinent and consistent