To Preserve Human Judgment, Courts Should Create an AI Adjudication Track

Summary

The essay argues that courts need a deliberate way to use AI as caseloads grow, rather than allowing judges and clerks to adopt opaque productivity tools informally. It cites backlogs exceeding 54 million cases in India and 75 million in Brazil, and says AI-generated filings are already increasing in U.S. federal courts: from about 0.1% of complaints before ChatGPT to 18% by early 2026. Self-represented litigants also rose from an 11% historical average to 17% in 2025. The proposed solution is a parallel AI adjudication track inside public courts, modeled partly on private arbitration but retaining human oversight, public accountability, and a route to precedent-setting decisions. In the American Arbitration Association’s construction-related, documents-only process, AI agents clarify each party’s claims and evidence, a domain-customized agent drafts an award, and a human arbitrator confirms or edits it; the process is faster, cheaper, and auditable, according to the essay. The author says private arbitration cannot simply be transferred to courts because binding diversion often requires consent, private providers may have incentives favoring companies, and confidential awards do not contribute reliably to legal development. Cases could be selected first by principle, excluding matters involving life, liberty, fundamental rights, or major value judgments, or empirically, using consent, appeals, and reversal rates to identify suitable categories. Early pilots could randomize referrals or require consent and guarantee human appeals. Later, courts could use the resulting data to route likely-to-appeal or low-consent cases to human judges, audit random samples, and refer unsettled legal questions for human resolution. Risks include bias, parties choosing speed without understanding the tradeoff, excessive deference by human reviewers, and a more mechanical common law. Suggested safeguards include agreed model selection, a council of models, public decisions, independent review, and human certification of important questions. The essay identifies Arizona as a practical pilot site because its existing arbitration system can handle claims below a court-set threshold and has previously supported a county pilot without legislative approval. It concludes that state supreme courts could begin with approved tools, one county, and an outside evaluator, while broader federal adoption would face greater procedural and legislative barriers.