ABSTRACT
Introduction Multi-agent systems (MAS) are increasingly used as a tool to disentangle and explore the complex relationships between land use and land cover change (LUCC), policy interventions and human adaptation. The development and application of these tools has been made possible by the rapid increase in computational power available at modest cost. The strength of agent-based land-use models (MAS/LUCC) lies in their ability to combine spatial modeling techniques, such as cellular automata or GIS, with biophysical and socioeconomic models at a fine resolution. Multi-agent systems are flexible in their representation of human land-use decisions and therefore appeal to scholars from diverse backgrounds such as sociology, geography and economics (Schreinemachers and Berger 2006). The behavior of individual actors can be modeled one-to-one with computational agents, which allows for direct observation and interpretation of simulation results. A large part of their fascination – especially to scholars who are otherwise skeptical of any attempt to quantify and model human behavior – rests on this intuitive and potentially interactive feature. Scholars from CIRAD (French Agricultural Research Center for International Development), for example, combine MAS/LUCC with role-playing games in which a group of resource users, typically farmers using some common-pool resource, specify the decision rules of computational agents and observe how these rules might affect both people’s well-being and their natural resource base (Bousquet et al. 2001; D’Aquino et al. 2003; Becu et al. 2003). In this chapter we reflect on the interactive use of multi-agent models not only for participatory simulation of land-use changes but also for teaching, extension and collaborative learning in general. At Hohenheim University, we have used our Mathematical Programming-based Multi-agent Systems (MP-MAS) software for teaching at the MSc and PhD levels, taught training courses for water resource managers in Chile and parameterized the MP-MAS model for empirical applications in Thailand, Uganda, Chile and Ghana (Berger 2001; Berger et al. 2006; Schreinemachers et al. 2007, 2009). MP-MAS distinguishes itself clearly from most other agent-based land-use models in its use of a constrained
optimization routine, based on mathematical programming (MP), for simulating agent decision-making. Apart from describing the rationale behind this modeling approach, this chapter reports on various case-study applications, and the use of the model for collaborative learning and research.
