Who Writes the Judgment—and Who Cares? Public Assessments of the Advantages and Challenges in LLM-Assisted Judgment Writing

Yixiao Wen, Sinong Lu

Keywords: LLM-assisted judgment writing, Trust in AI, Procedural fairness, Human–AI collaboration

Paper Type: Short papers / work-in-progress

TL;DR: We study how the public views AI-written judgments and why it matters whether a human judge or an LLM writes the judicial judgment.

Abstract: Courts have begun experimenting with Large Language Model (LLM)-assisted systems capable of generating judicial opinions, raising important questions about how the public perceives such AI involvement. We conducted a small, rapid qualitative study with seven individuals who had prior litigation experience and had used LLMs to generate legal content. Participants evaluated LLM-generated judgments through three dimensions—fairness, factual accuracy, professional form. They identified advantages such as more impartial reasoning and more formal legal language, alongside concerns about biased training data, hallucinations, and difficulties in meeting legal reasoning requirements. These findings highlight how perceptions are shaped by users’ mental models, prior AI experiences, and legal expertise, offering early insights for designing trustworthy LLM support in judicial decision-making.

Submission Number: 59