Social, economic and environmental phenomena as well as the presence of groups of stakeholders with diverging and contradictory interests highlight the complexity of urban projects. Traditional instruments of urban management are seldom adapted to these challenges, as they are often founded on sector- and expert-based approaches. Furthermore geographical information is paradoxically often available in a very large quantity, but not sufficiently relevant and synthetic for an understanding of urban complexity. In this context the goal of this thesis is to develop a method for the elaboration of relevant indicators systems from different stakeholders' cognitive representations of urban complexity; and thus to reinforce territorial diagnosis in the context of participative decision-making processes. However these differences of representations are often sources for conflicts. Each stakeholder uses his own representation in order to defend his point of view in a debate. The notice of these differences, from the "problem setting" phase of the decision-making process should contribute to a better structured debate and to a better management of the conflicts around a territorial project. This goal can be reached through consideration of all opinions and through an establishment of a discussion-space between different groups of representations. This allows the development of a coherent diagnosis framework which includes different stakeholders' cognition and perception of the territorial context. Our reflection is concretized by the developing of a prototype of interface that help a stakeholder to construct his own representation or urban phenomena, by organizing them in a dialectic (discussions) and didactic (complexity-training) manner. The phenomena-systems created are in the form of a causal model. The interpretation of this model permits to highlight critical phenomena through considering the preferred causal axes in order to output the stakeholder's goals. On this basis the stakeholder can debate and discuss with other stakeholders who also have constructed their own representations. From these individual representations, the question is how these representations can be integrated to form a collective representation of the territory. This process can be made in a deliberative manner. Before that, it is proposed to structure the debate by regrouping individuals models using analytical tools (classification, aggregation). Converging (collective reasons) and diverging elements of the urban complexity understanding are emphasized in the groups of models. The idea is thus to create a discussion-space for debates in order to find solutions. On the basis of different representation-groups, indicators-systems can be proposed. Indicators are selected in order to translate on one hand common representation of the main phenomena identified, on the other hand of causal relationships that result in strong differences. These indicators can contribute to a partial solution of conflicts by adding another, more objective point of view in the territorial context. Indicators are developed using spatio-thematical operators of geographical information systems. A concept of interface is proposed so as to facilitate the communication of systems of geographical indicators. In order to emphasis our approach, we have illustrated and applied the methodology of systemic urban-phenomena-modeling. We have conducted individual interviews with stakeholders who have participated in a diagnosis process in a neighborhood of Geneva and with stakeholders who are implied in suburban problematic in Québec. These discussions have shown the interest in our approach to help these stakeholders to formalize their representation of urban dynamics. In the same way, regrouping methods have proved their capability to reveal differentiated models of representation between groups of stakeholders. The method described presents an interest in participative decision-making and creates an advantageous framework for the interaction between these stakeholders around the representation of urban reality. The indicators-system contributes to this process and connects the representation of the territory and different stakeholders' cognitive representations according to a dynamic learning process. From this more solid and objectified basis the development of a diagnosis can succeed and the decision-making process can continue towards more operational phases.