Using problem solving models to design efficient cooperative knowledge-management systems based on formalization and traceability of argumentation Myriam LEWKOWICZ, Manuel ZACKLAD Laboratoire Tech-CICO Université de Technologie de Troyes 12, rue Marie Curie BP 2060 10010 Troyes Cédex France myriam.lewkowicz@univ-troyes.fr manuel.zacklad@univ-troyes.fr Abstract. We present here a groupware (MEMO-Net) based on a model (DIPA), which uses and simplifies the concepts of Problem-Solving methods. This model comes from a review of Design Rationale formalisms that gave rise to the ABRICo formalism. MEMO-Net enables exchange structuring in order to improve dialog quality and reusability. MEMO-Net prefigures a new KM approach, named “cooperative KM” where we propose virtual work environments for groups that will consist in new “coordination mechanisms”. 1. Introduction In accordance with Zacklad and Grundstein [17], we can classify knowledge management (KM) approaches in three complementary categories: top-down, bottom- up and cooperative. In the first one, models are used with experts to formalize their knowledge (MKSM for example). In the second one, huge corpus are memorized and formalized afterwards (text-mining methods). And in the third one, we consider that organizations’ critical knowledge comes within a collective competence that is not enough or badly formalized. Groupwares generally use two kinds of models to structure interactions and to manage knowledge in organizations: some of these models allow a relationship standardization instead of others that focus on know-how standardization [18]. Among models based on relationship standardization, we can quote the famous Speech-Act theory from Winograd and Flores [15] and its derivative [6]. These models allow a more efficient management of contractual processes in firms. They fit in KM strategies focused on “commitment traceability”. On the other hand, models used for decisions’ traceability in the scope of know-how standardization are closer to Design Rationale (DR) researches. In these studies, models are decision oriented. We can quote QOC [9] or IBIS [1].