Neuro-Symbolic AI for Korean Criminal Law: Sentencing Prediction and Document Drafting

2026-07-22

AI Summary

This study addresses the administrative and sentencing burden faced by prosecutors in South Korea’s summary criminal proceedings due to the high volume of minor traffic offenses. To alleviate this, the authors propose a neuro-symbolic hybrid system that integrates large language models for semantic understanding with formalized encoding of the 2026 Korean Sentencing Guidelines for Traffic Offenses and SMT solver–driven symbolic reasoning, maintaining strict separation between neural and symbolic components. A human-in-the-loop review mechanism ensures legal oversight and mitigates hallucinations inherent in large language models. The approach enables fully automated generation of indictments and deterministic sentencing predictions without stochastic errors, thereby significantly enhancing judicial administrative efficiency while preserving legal rigor and interpretability.

Abstract

The Korean criminal justice system utilizes summary proceedings (guyaksik) to expedite high-volume minor infractions, such as simple driving under the influence (DUI), unlicensed driving, and minor traffic casualties. Although this mechanism improves judicial throughput, processing these cases creates a substantial administrative burden for prosecutors, driving the need for automated systems that can precisely translate unstructured legal text into deterministic statutory outcomes. While recent Large Language Models (LLMs) excel at semantic extraction, their probabilistic nature inherently limits their reliability in Legal Judgment Prediction tasks. Specifically, when confronted with the arithmetic constraints of legal statutes, LLMs can produce hallucinations. Given that legal accountability permits virtually no tolerance for stochastic errors, purely neural architectures remain limited in their direct judicial applications. To address these limitations, we propose a Neuro-Symbolic framework that bridges unstructured legal facts with formal verification. Our architecture restricts the LLM exclusively to semantic extraction, while offloading statutory fine calculations to a Satisfiability Modulo Theories solver. This division of labor reduces hallucination risks during computation. Furthermore, we incorporate a Human-in-the-Loop verification scheme to preserve professional legal oversight. We formalize the 2026 Sentencing Guidelines for Traffic Offenses within this pipeline, demonstrating a deterministic approach to supporting summary indictments.

Problem

• Legal Judgment Prediction
• Neuro-Symbolic AI
• Sentencing Prediction
• Hallucination
• Summary Proceedings

Innovation

• Neuro-Symbolic AI
• Legal Judgment Prediction
• Satisfiability Modulo Theories
• Hallucination Mitigation
• Human-in-the-Loop