Consensus reaching decisions in groundwater remediation system: A reinforcement learning decision model addressing...

Groundwater remediation projects (GRPs) are critical for reducing groundwater contamination and protecting public health and ecological stability. However, decision-making in these projects is challenging due to long implementation cycles, uncertain environmental conditions, and the involvement of multiple stakeholders with conflicting interests. Unexpected events further intensify these challenges by altering stakeholder perceptions and priorities, leading to fluctuations in consensus. This study proposes a simulation-optimization framework to support adaptive decision-making under such conditions. The simulation component employs a concatenate Friedkin-Johnsen model to represent how stakeholder opinions evolve in response to unexpected events, while the optimization component integrates real options theory and reinforcement learning to determine whether to expand, defer, or abandon the consensus reaching process. Results show that consistent risk perception and opinion fusion strengthen consensus resilience, whereas the option to defer or abandon enhances adaptability under deep uncertainty. The proposed framework provides a systematic approach for managing stakeholder consensus in long-term GRPs and offers broader insights for adaptive environmental decision-making involving multiple stakeholders.