Multi-Stakeholder Non-Dominated Sorting Genetic Algorithm-II: Beyond Single Decision Maker Optimization

Abstract

Evolutionary multi-objective optimization (EMO) is widely used to solve problems with competing objectives, yet most existing approaches assume a single decision maker and offer limited support for the diverse and often incomplete preferences found in multi-stakeholder settings. We introduce MS-NSGA-II, an extension of NSGA-II that incorporates heterogeneous stakeholder aspirations without requiring negotiation or repeated interaction. Stakeholders specify only the objectives and aspiration levels relevant to them, and a normalized dissatisfaction measure guides selection while preserving the underlying multi-objective formulation and maintaining diversity in the objective space. Experiments on benchmark problems with up to ten objectives and a real-world area design case study show that MS-NSGA-II significantly reduces stakeholder dissatisfaction and uncovers Pareto-superior regions, particularly in higher-dimensional settings.