A multi-agent approach to involve multiple knowledge models and the case base reasoning approach in decision support systems

The building of Decision Support System (DSS) in complex domains like management, medicine, environment...requires to involve various data and knowledge models and to support the reasoning modes taking place in the professional cognitive process during his or her job practice. We propose a Multi-Agent Decision Support System (MADSS) architecture to integrate the different categories of knowledge representation. We describe the functionalities and capacities of cognitive agents implementing these different kinds of knowledge and bringing them to cooperate in elaborating the decision. Agents are able to commit to build sub-decisions and thus, the system provides a whole sequence of decisions fitting the user's usual cognitive scheme. We trust in a supervised architecture because such a MADSS must be goal driven by the user's requirements. However, the agents remain autonomous and keep the opportunity to commit or not to the task exhibited by the supervisor at the moment.