Recent Advances, Critical Reflections, and Future Directions in Large-Scale Group Decision-Making: A Comprehensive Survey

This comprehensive survey provides a systematic examination of recent advances, critical reflections, and future directions in large-scale group decision-making (LSGDM), a rapidly evolving field driven by the proliferation of digital platforms and the growing complexity of collective decision scenarios. Building upon a rigorous analysis of 360 peer-reviewed publications from 2014 to 2025, we establish a structured knowledge framework that organizes current research into four core modules: information modeling and representation, dimensionality reduction, behavior management, and consensus model design. Theoretically, our work contributes by synthesizing diverse methodological approaches into a coherent taxonomy, identifying key theoretical gaps in preference modeling, behavioral dynamics, and consensus measurement, while critically examining the disjunction between model sophistication and real-world applicability. From an application perspective, we analyze representative cases across domains, including emergency response, urban planning, and environmental governance, revealing both the transformative potential and practical limitations of current LSGDM methods. Our critical reflections expose fundamental challenges in semantic representation, dynamic clustering assumptions, behavioral intervention effectiveness, and consensus evaluation metrics. Looking forward, we propose a dual-path research framework integrating method-oriented and problem-oriented modeling paradigms, highlighting five transformative technological directions: 1) unified modular information frameworks with adaptive semantic modeling; 2) goal-oriented dynamic clustering techniques; 3) theoretically grounded behavioral models incorporating cognitive and social psychology; 4) AI-enhanced consensus mechanisms leveraging large language models and reinforcement learning; and 5) federated learning architectures for privacy-preserving decentralized decision-making. This survey not only constructs a comprehensive knowledge system for LSGDM research but also provides actionable insights for bridging theoretical innovation with practical implementation challenges, ultimately guiding the development of next-generation group decision support systems that are more scalable, explainable, and human-centered.