Résumé
In the last decade, the design of collaborative discovery learning environments (CDLE's) has received increasing attention. Such a design perspective brings up to the educational context the possibility of exploiting the so-called Model-Based Reasoning approach, by providing to learners the opportunity of collaborating while building models to represent observations. In this paper, we are concerned with the design of CDLE's in which building models instantiates as a theory formation process. Such a process is provided as a synergetic combination of both inductive and hypothetical-deductive approaches. The “moving engine” allowing a theory to evolve is the notion of contradiction: learning is supposed to occur as a side effect of contradiction detection and overcoming during theory formation by learners guided by a coordinator. Inspired in recent relevant work concerning Computer Supported Collaborative Learning (CSCL) architectures, we propose an architecture to support the process of asynchronous theory formation, allowing a student both to work individually and to contribute to the group discussion. Out of the proposed architecture, we draw-up related questions that would address supporting a coordinator to guide discussions on the basis of group interaction analysis.