Using artificial agents to understand laboratory experiments of common-pool resources with real agents

Most natural resource systems used by multiple individuals can be classified as common-pool resources. Conventional economic theory predicts that when agents have free access to a common-pool resource they will consume ecosystem services to the point where private costs equal the benefits, whereas externalities are imposed on the rest of the community. This can lead to the well-known tragedy of the commons (Hardin, 1968). Many laboratory experiments have been performed to study this phenomenon. Even in the simplest case of these experiments, without communication between the participants, anomalies where found that are not in line with conventional economic theory, which is non-cooperative game theory (Ostrom et al., 1994). Conventional theory predicts that players in a non-cooperative game will follow a Nash equilibrium. In none of the reported experiments on commonpool resources was such a Nash equilibrium observed. Furthermore, the total consumption of the common-pool resource fluctuates in time (Ostrom et al., 1994). Many studies have focused on this phenomenon. Dudley (1993) analysed individual data from Indiana University on resource-use and classified most of the participants according to the strategy they used, such as noncooperative and cooperative behavior. Deadman (1997) developed a simulation model based on artificial intelligence, which reproduced patterns similar as observed in the laboratory experiments. Casari and Plott (2000) report laboratory experiments that are consistent with their proposed analytical model that includes heterogeneity of social orientation among the agents.