Multi-agent systems companion modeling for integrated watershed management : a northern Thailand experience

Integrated watershed management means a collective management of the land reconciling ecological dynamics and social processes to ensure a sustainable and equitable use of renewable resources. Because different stakeholders have different representations of a local agricultural system, its integrated management can be seen as a collective learning process. New types of models, closely articulated with field work and other participatory tools, can be used to facilitate such collective learning. Particularly, they are useful to elucidate relationships between agent behaviors and interactions, and resource dynamics at different levels of organization. This paper describes and discusses the development and preliminary field testing of such a companion modeling approach, based on the use of multi-agent systems (MAS) associated with role playing games. Companion modeling supports on-farm, interdisciplinary and action-oriented participatory research, to facilitate dialogue, shared learning, negotiation, and collective decisionmaking among multiple stakeholders. The principles of this approach, the characteristics of agentbased models and associated role playing games are explained. Their validation and use with stakeholders to manage renewable natural resources are also presented. The article is illustrated by an on-going case study to improve steep-land management by limiting land degradation in rapidly diversifying and market integrated farming systems of Akha villages in upper northern Thailand. Field work and modeling activities are seen as very complementary, and are closely linked in an iterative way. The collective construction of a common artificial world with stakeholders leads to the emergence of a shared representation of a complex system, and of the concrete problem to be addressed. Later, such a common representation can be used among stakeholders as a coordination and negotiation support tool to identify and to assess scenarios of desirable futures. On the basis of a shared understanding of current systems dynamics, this approach helps to identify acceptable rules for an improved regulation of collective uses of land resources. When a policy of decentralization of natural resource management is implemented, companion modeling can be used to integrate knowledge, to stimulate dialogue and establish adapted coordination mechanisms regarding multiple uses of the land by multiple stakeholders, but also to assess suitable innovations and desirable scenarios of land use changes for the future. (Resume d'auteur)