Decision and Learning Model Selection for Complex Adaptive Systems
Computer modeling is gaining popularity in study of systems whose underlying processes are hard to measure directly, or controlled experimentation is impossible. Since many real-world phenomena, for instance psychological or ecological, are often complicated, and the models trying to capture their essence relatively complex, selecting the best model from the candidates is a challenge. In this presentation I address model selection in the context of complex adaptive systems, and particularly among classes of models that are used to gain understanding in the most significant single factor behind the global climate change, namely human land-use. In order to understand the impact of the land-use change, not only its consequences but also the underlying mechanisms and socio-economical, political, psychological, and historical factors driving the change need to be explained. I focus on agent-based models of human learning and decision making in the domain of land-use, and propose a criterion to select between these models. The candidate models constitute a set of relatively straightforward reinforcement-based strategies familiar from psychology and economics.