Conditioning Large Language Models on Legal Systems? Detecting Punishable Hate Speech

TL;DR

This study explores how large language models can be conditioned on different levels of legal knowledge to detect punishable hate speech, revealing significant gaps compared to legal experts and highlighting challenges in model understanding and consistency.

Contribution

It investigates methods to condition LLMs on various legal abstraction levels for hate speech detection, providing insights into their capabilities and limitations in legal assessment tasks.

Findings

Models conditioned on abstract legal knowledge often hallucinate and lack deep understanding.

Concrete legal knowledge conditioning helps identify relevant target groups.

Performance gap remains significant between models and legal experts.

Abstract

The assessment of legal problems requires the consideration of a specific legal system and its levels of abstraction, from constitutional law to statutory law to case law. The extent to which Large Language Models (LLMs) internalize such legal systems is unknown. In this paper, we propose and investigate different approaches to condition LLMs at different levels of abstraction in legal systems. This paper examines different approaches to conditioning LLMs at multiple levels of abstraction in legal systems to detect potentially punishable hate speech. We focus on the task of classifying whether a specific social media posts falls under the criminal offense of incitement to hatred as prescribed by the German Criminal Code. The results show that there is still a significant performance gap between models and legal experts in the legal assessment of hate speech, regardless of the level of abstraction with which the models were conditioned. Our analysis revealed, that models conditioned on abstract legal knowledge lacked deep task understanding, often contradicting themselves and hallucinating answers, while models using concrete legal knowledge performed reasonably well in identifying relevant target groups, but struggled with classifying target conducts.