About JudgeLens
How we build data-grounded Supreme Court simulations — and why it matters.
Methodology
Every response is grounded in real data, not training-set impressions.
Data Source
All transcript data comes from the Oyez Project at Chicago-Kent College of Law — the most comprehensive public archive of U.S. Supreme Court oral argument audio and transcripts. We process timestamped, speaker-attributed transcripts covering two decades of oral arguments.
RAG Pipeline
When you submit an argument, it passes through a six-step pipeline: your text is embedded, semantically matched against thousands of real exchanges, reranked by a cross-encoder for precision, and then used to ground a response generated by Claude. The system prompt for each justice is built from quantitative analysis of their actual speaking patterns — speech markers, question-to-statement ratios, hypothetical frequency, and more.
Grounding Approach
Unlike generic AI role-play, JudgeLens never asks a model to “pretend to be” a justice. Instead, it retrieves the justice’s actual words from cases most relevant to your argument, then instructs the model to respond in that justice’s measured style. The result is a response shaped by data, not imagination.
The Data
Numbers behind the platform — all 9 sitting justices.
Exchanges indexed from real oral arguments
Individual speaking turns analyzed
Oral arguments spanning 1981–2025
Years of transcript data processed
Reflects all 9 sitting justices of the U.S. Supreme Court.
Important Disclaimers
Not Legal Advice
JudgeLens is an educational and research tool. It does not provide legal advice, predict case outcomes, or represent the actual views of any sitting justice. Simulated responses are AI-generated approximations grounded in transcript data — not statements by the justices themselves.
AI Limitations
While grounded in real data, all responses are generated by AI and may contain inaccuracies, hallucinations, or mischaracterizations. The system captures stylistic and rhetorical patterns, not the substantive legal reasoning a justice would actually apply to a novel question. Use outputs as a learning tool, not as authority.
Educational Purpose
This tool is designed for law students, legal scholars, moot court practitioners, and anyone interested in understanding how Supreme Court justices engage during oral argument. It is meant to make the Court more accessible and transparent — not to replace careful legal analysis.
Credits & Attribution
Oyez Project
Transcript data sourced from the Oyez Project at Chicago-Kent College of Law, IIT. Oyez is a free legal resource providing access to Supreme Court audio and transcripts.
Technology
Built with ChromaDB for vector storage, nomic-embed-text for embeddings, cross-encoder reranking for precision, and Claude for response generation. Frontend powered by Next.js.
See It in Action
Present a legal argument and experience data-grounded simulation firsthand.
Open the Simulator