AI Courts Could Preserve Human Judgment

Overview

Global court systems are struggling under the weight of mounting backlogs, a crisis that artificial intelligence threatens to exacerbate rather than solve. With millions of cases pending in jurisdictions from India to Brazil, the introduction of low-cost legal tools risks flooding already overwhelmed dockets with an surge in filings.

“When a plausible legal document costs almost nothing to write, courts will see more cases as well as more filings in each case.”

Economists Anand Shah and Joshua Levy provide credible evidence that this shift is already underway in U.S. federal courts. Before the rise of large language models, AI-generated text appeared in roughly 0.1% of complaints. By early 2026, that figure had climbed to 18%, alongside a jump in self-represented litigants from 11% to 17% in 2025.

While increased access to justice is a positive development, unchecked growth in litigation volume risks degrading the quality of judicial outcomes. Judges face a grim choice: reduce the time spent on each case, potentially leading to poorer decisions, or take longer to deliberate, extending waits for all parties. To address this, Justin Curl proposes a novel solution: creating parallel AI adjudication tracks within the public court system.

The Agentic Flooding Crisis

The core problem driving this proposal is what Chris Schmitz of the Hertie School calls “agentic flooding.” This term describes the AI-driven surge in requests directed at public services, a phenomenon that mirrors the classic “tragedy of the commons” when applied to the judiciary.

If every individual can leverage AI to flood the courts with lawsuits, the collective result is a system that no one benefits from. The cost of filing becomes negligible, removing the traditional friction that filtered out frivolous or overly complex claims. Consequently, judges are left with insufficient time and attention to devote to the cases that truly require nuanced human judgment.

Current attempts to solve this by simply urging judges to adopt AI tools themselves are fraught with danger. There is a significant risk that judges or their clerks may lean on algorithmic assistance too heavily, potentially outsourcing moral reasoning or failing to keep pace with the technology’s limitations. Rather than forcing humans to adapt to accelerating demands, the legal system should instead create structural boundaries around where human oversight is mandatory.

A Parallel Track Inspired by Arbitration

Curl’s proposal draws a direct analogy to private arbitration, suggesting that courts establish a distinct pathway for disputes that do not require full judicial intervention. This track would operate similarly to how magistrate judges function today, serving as a formal, publicly funded alternative to both traditional bench trials and private arbitration.

The mechanism for this model already exists in the private sector. The American Arbitration Association (AAA) has launched an AI-based arbitration process specifically for construction industry disputes. In this system, AI agents interact with both parties to clarify claims and gather evidence through iterative questioning.

Once the facts are established, the agents present the case to a specialized AI “clerk” agent, which drafts an award based on domain-specific knowledge. This draft is then reviewed by a human arbitrator, who confirms or edits the decision before issuance. The process is significantly faster and cheaper than traditional methods, and every interaction is logged and auditable.

“This track... would strengthen the role of humans in the judiciary. It would allow courts to be explicit about how human judgment is allocated.”

By adopting a similar model publicly, courts could handle high-volume, document-heavy cases efficiently while preserving human judges for matters involving complex legal precedents, ethical considerations, or disputed facts that algorithms cannot adequately navigate. State supreme courts have the authority to launch pilot programs for such tracks immediately.

Why Private Arbitration Alone Is Insufficient

While private AI arbitration demonstrates technical feasibility, relying solely on existing private mechanisms to relieve public court burdens presents serious systemic risks. First, courts lack the power to compel parties into private arbitration; both sides must consent. Even in jurisdictions like Arizona, where arbitration can be mandated, either party retains the right to demand a new trial afterward, which undermines the efficiency gains.

Second, private arbitration suffers from entrenched incentives that favor corporate defendants over individuals. Arbitration clauses are often buried in terms of service that consumers rarely read, and companies typically select the arbitration provider. This creates a market pressure where arbitrators who rule in favor of businesses are more likely to be selected for future cases—studies suggest they are 40% more likely to be chosen again.

Third, and perhaps most critically, arbitration is sealed off from the development of public law. Because arbitral awards do not set binding precedent, hard legal questions never reach appellate judges. A public parallel track avoids this by ensuring that while routine cases are filtered through AI, significant legal questions can still emerge and be resolved by human judges, maintaining the integrity of the common law system.

Preserving Human Judgment

The ultimate goal of this proposal is not to replace human judges but to protect their capacity for meaningful decision-making. As AI continues to lower the barriers to legal action, the volume of cases will likely continue to rise. Without a structured way to triage these disputes, the quality of justice will deteriorate for everyone involved.

By explicitly designating certain cases for AI adjudication, courts can make transparency a virtue. litigants and the public will know when a decision was generated by an algorithm versus a human judge. This clarity allows for targeted appeals and audits, ensuring accountability.

Ultimately, this approach reframes the role of technology in the judiciary. Instead of viewing AI as a tool for doing more of the same work faster, it should be seen as a mechanism for doing different work better. Saving human attention for the cases where moral nuance, equity, and legal interpretation are paramount is the only sustainable path forward.