Research Summary: Ordinary, Reasonable Chatbots: Do AI Models Track Human Legal Judgments?

Research is exploring the capacity of large language models (LLMs) to simulate human legal judgments, particularly concerning the concept of 'reasonableness' in legal contexts. As AI guidance becomes more prevalent, the potential for 'silicon jurors' in courtrooms is being considered, mirroring the existing use of generative AI in social science research ('silicon sampling'). This study investigates how LLM-powered chatbots respond to legal questions regarding the appropriateness of behavior based on reasonableness.

Why it matters

The growing reliance on AI for guidance and decision-making extends to legal domains, raising critical questions about the accuracy and reliability of AI models in simulating human judgment. Understanding AI's capacity for legal reasoning, particularly concerning subjective standards like 'reasonableness,' is vital for developing appropriate governance frameworks and ensuring public trust in AI applications.

Key insights

• Generative AI models are increasingly being considered for roles in legal decision-making.
• The concept of 'silicon jurors' is emerging, suggesting AI's potential in courtrooms.
• 'Silicon sampling' – using generative AI in social science research – is already impacting academia.
• The study specifically examines how LLM-powered chatbots interpret and respond to legal questions concerning 'reasonableness,' a common legal standard for judging behavior.

Source

arXiv — Computers and Society — https://arxiv.org/abs/2609.06769