Assessing preservation boundaries for historic districts: An integrated framework of explainable AI and game theory

Abstract

The sustainable preservation of historical heritage requires balancing cultural, environmental and socioeconomic factors. This study proposes an innovative framework integrating Explainable AI (XAI) and game theory to address the complexity of assessing preservation boundaries in historic districts. Focusing on the Chongqing Two-River Confluence Historic Urban Area, we applied a Multi-Layer Perceptron (MLP) neural network and SHapley Additive exPlanations (SHAP). This approach facilitates a three-step process: factor analysis, multi-party game equilibrium, and preservation boundary assessment. The findings reveal that (1) historical heritage elements dominate conservation core zone decisions, underscoring the importance of prioritizing cultural value; (2) natural-geographical features, particularly river and mountain proximities, are crucial for ensuring environmental harmony in boundary design; (3) socioeconomic factors, though less influential, contribute to balancing preservation with urban development; (4) boundary assessment should integrate natural-geographical elements that maintain the spatial integrity of historical environments, moving beyond rigid artificial boundaries such as roads or administrative divisions; and (5) irregular and fragmented preservation boundaries correlate with increased likelihood of becoming zones of frequent gaming and adjustment. This study demonstrates the adaptability and interpretability of XAI in sustainable urban planning, providing a scientifically validated framework for heritage conservation.