It is widely admitted that poor coordination among stakeholders leads to inefficient resource use, economic and environmental damage, negative externalities, and social conflict. Diverse stakeholders use resources for different purposes, with differing perceptions of their dynamics, and adopt various strategies to cope with problems. Consequently, the number of social conflicts is increasing and they are frequently reported in the national and international media, for example, the cases of water sharing at rice transplanting among farmers and villages in Bhutan, and conflicts over land use between highlanders and lowlanders in northern Thailand uplands. To manage these problems, new legislative frameworks decentralizing the management of renewable resources are being introduced in many countries. To be efficient, these new systems of governance of local resources require a greater involvement of the concerned stakeholders. Their success depends on the quality of the local coordination among stakeholders, who often lack tools, methods, and trained managers to achieve success. Thus, there is a demand for innovative approaches and tools to improve coordination processes among an increasing number of stakeholders using common resources at the rice agroecosystem level. Our hypothesis is that an understanding and modeling of the diversity of stakeholders' perceptions, associated with participatory simulation sessions, can be used to improve resource management through better coordination of people's actions in any type of rice ecosystem. Companion modeling (ComMod) is a methodology for the collective implementation and use of simulation models, and, more precisely, multi-agent simulation systems (MAS) which are suitable to represent complex heterogeneous and open systems and for integrating knowledge from various sources (Barreteau 2003). The ComMod approach proposes an iterative and evolving modeling process: participatory modeling is used to facilitate stakeholders' int