A nationwide study has found that generative artificial intelligence significantly improved case resolution in Pakistan’s trial courts. Researchers reported that AI helped judges resolve an additional 1,848 cases annually. Consequently, overall case disposal increased by more than 6 percent without reducing quality.
Study Shows AI Improves Judicial Productivity
The research paper, titled “Courts of Tomorrow: Evidence from a Nationwide Rollout of Generative AI,” examined the use of JudgeGPT. Researchers developed the AI assistant using OpenAI’s GPT-4 family of models and customized it for Pakistan’s judicial system.
Moreover, the Federal Judicial Academy extensively tested the tool before deployment. Authorities later introduced JudgeGPT to 1,559 judges serving across 118 trial courts nationwide.
The study found that targeted AI training produced the strongest results. Judges receiving both AI access and specialized training resolved 6.3 percent more cases than average.
Researchers also found that trained judges adopted the tool more frequently. Furthermore, they continued using it over time and expressed greater confidence in its productivity benefits.
AI Supports Judges Without Replacing Them
The trial divided judges into three groups to compare different training approaches. One group received JudgeGPT with targeted training, while another received only general technology training alongside AI access. Meanwhile, the control group received general training without the AI assistant.
Researchers found that judges mainly used the tool for legal research, summarisation, and improving written text. Additionally, targeted training encouraged judges to use AI for focused support tasks instead of complex legal reasoning.
The study found no meaningful decline in writing quality after AI adoption. Instead, researchers observed slight improvements in overall quality assessments. They also found little evidence of changes in gender or religious bias within judicial language.
Researchers concluded that AI should assist judges rather than replace judicial decision-making. They added that properly trained AI systems can help reduce case backlogs while strengthening the justice system’s overall efficiency.
