AI Outperforms Law Professors in Contract Law Reasoning
- In a study, AI-generated answers were preferred over professor-written ones about 75% of the time.
- Google’s Gemini recorded a harmfulness rate of only 3.41%, compared to professors’ rate of 12.06%.
- AI models outperformed human instructors in recall questions and policy discussions.
- Anthropic’s Claude Opus ranked first among AI models, followed by OpenAI’s ChatGPT and Google’s Gemini.
The study led by Stanford University found that large language models (LLMs) align with professional standards in legal reasoning tasks, often surpassing human instructors in preference and performance metrics.
This research highlights the potential of LLMs as educational tools, showing they can provide responses that are not only preferred but also less frequently flagged as harmful than those from human instructors. (Source)
