Existing distance learning systems provide collaboration tools (like forum, chat, wiki, etc.) presenting some limits that reduce the effectiveness of such tools from the standpoint of collaborative learning. Accordingly, a conceptual framework is needed that bridges the gap between the interaction forms characterising collaborative learning and existing learning collaboration tools. In this context, Multi-Agent System (MAS) seems to be a suitable paradigm to engineer such systems as it promotes effective collaboration. This paper shows how the Agents and Artefacts (A&A) meta-model for MAS could be exploited to build a conceptual framework for collaborative learning systems