A Classification and Grading Framework for Social Conflict Resolution Based on Knowledge Extraction from Large Language Models

Abstract

Due to cultural and institutional differences, social conflicts in East countries are mainly addressed through mediation mechanisms embedded in community governance, while Western societies tend to rely more on formal litigation processes. Originating in the 1960s in Fengqiao Town, Zhuji City, Zhejiang Province, China, the “Fengqiao Experience”, a celebrated practice of conflict resolution, has evolved into an innovative community-based model for resolving social conflicts. A key research question of both theoretical significance and practical relevance is how to distill insights from practical conflict cases, systematically generalize the “Fengqiao Experience,” and construct a standardized paradigm for social conflict resolution. To achieve this goal, this study develops a classification and grading framework for conflict mediation by leveraging large language models (LLMs) to extract structured knowledge from a corpus of practical conflict cases. Subsequently, a knowledge graph is constructed through entity recognition, relation extraction, and semantic integration, followed by a structured analysis to optimize the framework. Within this framework, the classification system categorizes social conflicts into three interrelated types: contradiction, argument, and dispute, encompassing 12 subcategories, which reflect the escalating intensity of social conflicts from latent tensions to open confrontations. The grading system, anchored in eight core dimensions (subject, object, behavior, consequence, impact, source, disposal, and method), integrates 33 specific indicators to enable a multi-faceted assessment of conflict severity, mediation complexity, and resolution effectiveness. The robustness, logical consistency, and practical applicability of the framework have been validated through application and real cases, demonstrating its effectiveness in addressing complex social conflicts. This study contributes an intelligent method that transforms experiential knowledge from practical cases into a standardized paradigm for social conflict resolution using LLMs and knowledge graphs. The findings lay a crucial foundation for upgrading the “Fengqiao Experience” into an intelligent social governance mechanism, offering a scalable solution for modern community-based conflict resolution.