The objective of the research carried is to facilitate collaborative activities in knowledge management software. More particularly, we aim to improve collaborative activities by the implementation of indicators. We intend thus to facilitate the progression of one or several implemented collaborative processes and to support the social networks within the activities. For that purpose, we propose to observe the users’ activity by providing collaboration indicators calculated from the digital traces of activities. The first objective of this thesis is to identify the characteristics of the collaboration indicators. In order to make the best possible use of indicators, we propose a classification composed of six main independent characteristics which correspond to the fundamental elements for the elaboration of collaboration indicators. This classification supplies us with indications to build, model and set up an indicator in a collaborative environment such as those studied in the context of the project. Moreover, this thesis presents a framework for the management of indicators based on a user-centred approach. Indeed, by observing the current use of indicators in platforms of knowledge management, we realized that the process of elaboration and acceptance was long and complex. Consequently, when an indicator is established in an environment and when it works properly, it is interesting to reuse it in other contexts. This reusing can be done entirely or partially. We thus propose a type of modeling which takes into account the reuse issues as well as those of the personalization and adaptation of indicators according to the user and\or to the context of the activity. In our approach, the indicators are calculated from the activity traces left by the users. To facilitate the implementation of these indicators, we have developed a tool which assists the users during the elaboration of indicators: GINDIC (Generator of INDICATORS). With this tool, the designers are able to build indicators: their definition, the calculations made on traces to obtain the indicator value, and their displaying. We have chosen to manage indicators by using a rules-based system, where traces are facts, whereas the calculation processes for indicators are rules. This thesis has been written within the framework of a project financed by the French National Research Agency entitled ProCoGeC (Collaborative Software package of Management of Knowledge). A rules-based language was compared with the needs of the ProCoGeC project. It helps us to describe the calculation processes for all indicators stemming from the users’ requirements. An experiment with the tool has been carried out with one of the industrial partners of the project, who is the client of a collaborative platform of knowledge management.
