Repeated discrete choices in geographical agent based models with an application to fisheries. (January 2019)

• Record Type: Journal Article
• Authors: Carrella, Ernesto; Bailey, Richard M.; Madsen, Jens Koed
• Is Part Of: Environmental modelling & software. Volume 111(2019)
• Page Start: 204
• Page End: 230
• Publication Date: 2019-01
• Subjects: Agent-based models -- Adaptation -- Multi-armed bandit -- Fisheries -- Bio-economic modelling; Environmental monitoring -- Computer programs -- Periodicals; Ecology -- Computer simulation -- Periodicals; Digital computer simulation -- Periodicals; Computer software -- Periodicals; Environmental Monitoring -- Periodicals; Computer Simulation -- Periodicals; Environnement -- Surveillance -- Logiciels -- Périodiques; Écologie -- Simulation, Méthodes de -- Périodiques; Simulation par ordinateur -- Périodiques; Logiciels -- Périodiques; Computer software; Digital computer simulation; Ecology -- Computer simulation; Environmental monitoring -- Computer programs; Periodicals; Electronic journals; 363.70015118
• DOI: 10.1016/j.envsoft.2018.08.023
• Languages: English
• ISSNs: 1364-8152

Abstract

Abstract: Most geographical agent-based models simulate agents through custom-made decision-making algorithms. This makes it difficult to assess which results are general and which are contingent on the algorithm's details. We present a set of general algorithms, applicable in any agent-based model for choosing repeatedly from a set of alternatives. We showcase each in the same fishery agent-based model and rank their performance under various scenarios. While complicated algorithms tend to perform better, too much sophistication lowers performance. Further, while some algorithms perform well under all scenarios, others are optimal only in specific circumstances. It is therefore impossible to produce a single, unequivocal performance ranking even for simple general algorithms. We advocate then a heuristic zoo approach where multiple algorithms are implemented in the same model; this allows us to identify its best algorithm and test sensitivity to misspecifications of the decision-making component. Highlights: We describe 13 model-free decision algorithms for geographical agent-based models. Agents in geographical agent-based models often are multi-armed bandits variants. Theoretical worst case regret is not a good indicator of agent performance. Some algorithms perform well in general, some only in special conditions. Imitation and information sharing may reduce overall performance.