Acceso usuarios
Yuguang Lu
Purpose: To explore how artificial intelligence (AI) can transform criminal litigation dispute resolution in real-world judicial practice, focusing on resolving persistent challenges in cross-regional mediation, including legal application discrepancies and mediation model heterogeneity. Design: We integrate data from the national court electronic case file system and the Supreme People's Court’s judicial big data platform to construct a criminal mediation case database covering eastern, central, and western China. Using a domain-adaptive approach to standardize legal elements, we build a dispute prediction model combining deep adversarial networks with temporal attention mechanisms. Multidimensional feature vectors incorporate behavioral trajectories, social relationship networks, and historical caFse similarity analysis. Findings: The model achieves 81.4% accuracy in predicting successful mediations—33.6 percentage points higher than traditional legal analysis. In small-sample scenarios (e.g., border regions), it maintains 79.2% accuracy, demonstrating robust adaptability for resource-constrained courts. Conclusion: This study confirms AI’s capacity to unify regional judicial standards and provides an interpretable framework for data-driven dispute resolution in criminal practice.
