AbstractOne strategy to deal with unauthorized groundwater withdrawals by farmers is to impose penalties. Because farmers’ over-exploitation is subject to social, economic, cultural, and agricultural characteristics of the region, feedback assessments of a penalty policy using socio-economic simulation is indispensable. This study presents an agent-based framework to study the interaction among agricultural, environmental, and regulator agents within an agricultural system. The behavior of the agricultural sector was simulated at two levels: agricultural sub-agents, and agricultural group-agents. For agricultural sub-agents, individual benefit was maximized under behavioral and physical constraints simulated by a fuzzy inference system (FIS) and mathematical programing. A coupled linear optimization and nondominated sorting genetic algorithm (NSGA-II) were adopted to identify the best solution, from the agricultural sector’s viewpoint. The proposed framework was implemented in the Najaf Abad hydrological unit, situated in Isfahan province, central Iran, considering fixed and stepwise penalty approaches. Sensitivity analysis of the amount of base penalty in the fixed penalty scenario indicated that, under a $0.12 per cubic meter penalty, the annual over-exploitation declined from the current 79 million cubic meters (MCM) to zero. Furthermore, the results indicated that if the width of the step increased from 1 to 9 MCM/year, the over-exploitation increased by about 18 MCM/year.