Jialei Nie
Trans-regional carbon governance is a complex system with diverse stakeholders, conflicting indicators, fuzzy cognition and multi-source uncertainties. Traditional MCDM methods use manual weighting and single-agent frameworks, which cannot accurately capture stakeholders’ vague preferences or withstand external disturbances. To address these limitations, this paper proposes a distributed robust multi-criteria optimization decision method combining interval binary linguistic information and large language model (LLM). First, interval binary linguistic variables describe stakeholders’ dual uncertain cognition in carbon governance. We design fusion operators to reduce information loss during evaluation aggregation. Secondly, an implicit preference mining module based on LLM is designed to quantitatively extract risk preference, fairness preference and bounded rationality of decision-makers, so as to realize dynamic and intelligent weight optimization instead of subjective manual weighting. Thirdly, a distributed robust multi-objective optimization model is established for heterogeneous regional governance subjects to balance economic development, carbon emission reduction and governance cost under multiple uncertainties. Finally, a practical case of trans-regional carbon collaborative governance is used for empirical verification, and comparative analysis with classic multi-criteria decision methods is conducted. Results verify its high robustness and practicability, offering a novel framework for intelligent optimization of complex public governance systems.