Agent-Based Modeling with Classifier System: A New Modeling Tool to Investigate the Dynamics of Social/Ecological Systems with Particular Reference to the Maine Lobster Fishery

Over the last several decades scientific predictions of imminent over-fishing of the lobster fishery have been very inconsistent with the huge increases in landings and population numbers. To acquire a better understanding of the spatial/temporal dynamics of the fishery and explain this discrepancy, a new modeling approach, an agent-based model with John Holland\u27s learning classifier system, is introduced to model the lobster fishery. Instead of modeling the system from prior knowledge of the major factors affecting the population of lobsters and the population of fishers, an agent-based modeling approach models the system from bottom-up, i.e., it simulates the interactive decisions (where to place traps and whether to cut others\u27 traps) made by large numbers of individual fishers. Learning classifier systems allow \u27fishers\u27 to learn from their experiences and adapt their fishing strategies to the changing ecological and social environment. The aggregate patterns of the fishery \u27emerge\u27 as the product of thousands of individual decisions and give us a better understanding of the competitive dynamics that lay the foundation for the growth of self-governance in the fishery. The spatial, temporal and behavioral patterns produced by the model nicely match the patterns found in observations of 988,000 trap placements made by 44 fishers in the Gulf of Maine