Modeling Legal Reasoning: LM Annotation at the Edge of Human Agreement

TL;DR

This paper evaluates the performance of language models on a complex legal reasoning classification task, highlighting the importance of fine-tuning and human annotation for accurate results in specialized domains.

Contribution

It introduces a novel dataset of Supreme Court opinions annotated by experts and systematically tests various language models, emphasizing the effectiveness of fine-tuning over prompt-based approaches.

Findings

Fine-tuned models outperform prompt-based models on complex legal tasks.

LEGAL-BERT achieves the best performance among tested models.

Generative models perform poorly without task-specific fine-tuning.

Abstract

Generative language models (LMs) are increasingly used for document class-prediction tasks and promise enormous improvements in cost and efficiency. Existing research often examines simple classification tasks, but the capability of LMs to classify on complex or specialized tasks is less well understood. We consider a highly complex task that is challenging even for humans: the classification of legal reasoning according to jurisprudential philosophy. Using a novel dataset of historical United States Supreme Court opinions annotated by a team of domain experts, we systematically test the performance of a variety of LMs. We find that generative models perform poorly when given instructions (i.e. prompts) equal to the instructions presented to human annotators through our codebook. Our strongest results derive from fine-tuning models on the annotated dataset; the best performing model is an in-domain model, LEGAL-BERT. We apply predictions from this fine-tuned model to study historical trends in jurisprudence, an exercise that both aligns with prominent qualitative historical accounts and points to areas of possible refinement in those accounts. Our findings generally sound a note of caution in the use of generative LMs on complex tasks without fine-tuning and point to the continued relevance of human annotation-intensive classification methods.