Second Chair
1,501 curated sources on AI and the legal profession — scholarship, industry evidence, primary legal materials and practitioner interviews, 1949–2026 — assembled by David B. Wilkins and Anthea Roberts for the AI and the Legal Profession research program of the Center on the Legal Profession at Harvard Law School.
How the record is organised
The pipeline read 1,501 sources into 369 topics, grouped the topics into 46 themes, and the themes into 5 families. Open a family or theme to read it.
Swipe across the diagram to follow the branches, or use the list below.
Browse the hierarchy as a list
Legal-AI market transformation 187 topics
AI evaluation from testing to deployment 54 topics
- Legal AI assessment6
- AI deployment governance10
- Professional and agent benchmarks6
- Clinical AI assistance5
- Due-diligence extraction benchmarks3
- Legal capability rankings5
- Coding copilots2
- Adaptive medical AI devices2
- Professional roles and task allocation4
- AI request taxonomies2
- Hiring-bias audits2
- Medical AI device inventories5
- Expertise development2
Legal judgment amid institutional and technological change 49 topics
- AI and legal professional power4
- Access to justice2
- Delegated legal work7
- Lawyer error9
- Legal career inequality4
- Legal education and reasoning4
- E-discovery review4
- Digital information handling3
- Rule-based legal automation2
- AI training copyright2
- Federal legal aid funding2
- Privilege claims and case citations3
- Suspect citations in appeals3
Legal workforce effects of AI 55 topics
AI-enabled expert-service markets 24 topics
A pipeline, and a way of reasoning over what comes out of it
The corpus underlying this project was structured and analyzed by Insight Bridge and then analyzed through Dragonfly Thinking’s multi-perspective agentic method.
The structure and the analysis
Insight Bridge opens a dense body of evidence to whoever has a question, and makes what is inside it findable in the words you would actually use. Seven stages take the evidence base Anthea Roberts and David B. Wilkins assembled for this project and turn it into a structured relational database that enables cross-cutting analysis. The same pipeline runs over public submissions, oversight reports and research literatures; Second Chair is what it produced when pointed at this one.
The multi-perspective reading
Dragonfly’s methodology is compound vision: put a question through many differently shaped lenses rather than one, and hold what they find in tension before integrating it. It takes its name and its warrant from Tetlock’s finding that single-framework thinkers forecast worst, and that the ones who outperform go looking for multiple perspectives and ways of looking at the evidence. We are applying analytical lenses to the corpus: topic mapping and actor mapping are published, with driver and dynamics mapping and scenario mapping in progress.
The corpus browser covers 1,501 sources. The previously published analyses retain their earlier 600-source snapshot; their findings should be read against that evidence base.