There are plenty of dystopian science fiction stories in which computers or robots control society. The stories have a chilling effect: Despite their intelligence and efficiency, machines can be bizarre and cruel.
As litigators begin to incorporate artificial intelligence into their work, some courts are doing the same. A pilot program in Estonia allows litigants in small claims court to submit their disputes to a computer program. Some Brazilian judges regularly use artificial intelligence when drafting decisions. And a virtual judge presides over court proceedings in some parts of China, although human judges make the final decisions.
Before American litigators immediately dismiss this trend, or fully embrace it, I offer two benefits and two challenges that this technology may offer to the judiciary.
Artificially Intelligent Courts Can Produce Consistent, Predictable Results
The first possible benefit to a world with robot judges is predictability. People rely on a consistent legal system to know how to comply with laws. They turn to lawyers to give them guidance about whether certain actions are legal and what to expect in the course of their business. But litigators frequently advise clients that it is impossible to completely avoid risk since litigation is unpredictable. A potential claim could be dismissed as frivolous or a source of substantial financial exposure. Those possibilities can leave people uncertain about their rights and duties and require people to take extra precautions for a wide range of outcomes. Worse, many of the variances in the current system produce results that are worse for its most vulnerable.
One reason the justice system is uncertain is because judges are human beings who themselves can be unpredictable and do not act as one uniform monolith. Two judges in the same court may view the same agreement or the same law very differently. And even the same judge may view the identical disputes differently based on many irrelevant factors, such as the judge’s personal opinions of the parties, subjective bias, or even the judge’s particular mood that day.
Artificial intelligence allows for more uniformity. The same algorithm can be applied to all litigants, replacing the game of chance that currently exists when cases are randomly assigned to a particular judge or arise on a particular day. And an algorithm can decide cases free from many of the subjective biases, attitudes, and influences that affect people.
This uniformity may produce a system that treats litigants more fairly. And the consistency could make it easier for people to know what the law is and how to comply with it.
Technology Can Expand Access to Justice
Another possible benefit to AI judges is that it may reduce costs for people who can barely afford to participate. A major problem in many legal systems is that people without tremendous financial resources do not have access to legal representation. Societies have taken steps to mitigate this problem: There is a right to counsel in criminal matters, public service organizations provide free or low-cost services, and some lawyers perform pro bono work. Even so, many people are still unable to obtain counsel because they do not have the funds to pay. This leads to different legal systems for the rich and everyone else.
Computerized judges could possibly help level this playing field. One way they could do so is by speeding up the process by which cases are adjudicated. In the current system, it may take several months or more for a court to hear and decide a motion to dismiss. And it may take two years or more for a case to proceed from discovery through trial and appeal. Federal law requires district courts to address this issue by publishing a “six-month list” to encourage judges to decide motions in less than half a year. But efforts like that hardly earn the justice system a reputation for swiftness.
These delays not only add to legal expenses, but they put pressure on less wealthy clients who may prefer to settle claims sooner rather than wait for a distant future victory. Because computers can receive and process information quicker than people, and because they can generate decisions faster, too, clients may not need to spend as much time and money to get justice. And clients who cannot afford to wait years for a decision may have access to a legal system that can meet their needs.
Artificial Intelligence May Pose Technical Problems or Potential for Abuse
As for problems, computers mess up. Artificial intelligence is not perfect, and it likely never will be. And so, delegating the power of the state to an algorithm will likely lead to mistakes. Litigators are familiar with the tendency for generative AI software to “hallucinate” or cite nonexistent law; courts now regularly reprimand lawyers who cite fake law because they relied on a computer program without manually verifying the results. And this problem could extend to courts; already, Brazilian authorities have investigated a judge who used AI to draft an opinion that contained fake law. And while it is dangerous for a lawyer to make erroneous citations, it is far worse for the court to make enforceable decisions that directly affect litigants and society based on incorrect authority.
Additionally, sophisticated litigants may attempt to develop a technical understanding of how the software courts use works. They could then craft their arguments to take advantage of peculiarities of the computer system rather than the merits of the dispute. And the authors who write and maintain the software may know about, or even create, “backdoor” secret codes that could give some litigants advantages. For example, if a litigant knew that a computerized judge favored a certain combination of words or specific length of a document, she may try to “game” the system accordingly to the detriment of an adversary who did not try to leverage these technical issues.
To some extent, however, these problems are not new. Human judges make mistakes, too. And some lawyers know judges well and enjoy an advantage based on their relationships and knowledge of judicial preferences. But these problems feel scarier when they arise from something inscrutable, like a computer, rather than something familiar, like human nature.
Computers May Lack Human Values and Popular Trust
Even if artificial intelligence could take existing laws and precedents and perfectly apply them to new cases, it may still not do the work people expect from human judges.
First, human judges apply human values to cases. In some instances, courts grant equitable relief based on principles of fairness even when no statute or clear legal duty applies. And judges frequently make decisions about reasonableness based on a shared understanding of the human experience. While a computer program could attempt to emulate this kind of justice, the changing nature of human values and novel situations that will present themselves may make it difficult to know when to incorporate human values and how to do so.
And second, the justice system functions only because people respect it. Judges wear robes and courtrooms are designed in impressive styles in part to convince people to obey judicial orders. But people may not have the same respect for software, which may threaten the stability of the judicial system.
Generative artificial intelligence will have some role in the judiciary of the future. While we cannot stop it, being aware of its potential benefits and challenges may help us plan accordingly.
Resources
- Amin Ebrahimi Afrouzi, “John Robots, Thurgood Martian, and the Syntax Monster: A New Argument Against AI Judges,” 37 Can. J.L. & Juris. 369 (Aug. 2024).
- Benjamin Minhao Chen, Alexander Stremitzer, and Kevin Tobia, “Having Your Day in Robot Court,” 36 Harv. J.L. & Tech. 127 (Fall 2022).
- David Horton, “Forced Robot Arbitration,” 109 Cornell L. Rev. 679 (2024).
- Christopher Michael Malikschmitt, “The Real Future of AI in Law: AI Judges,” L. Tech. Today (Oct. 18, 2023).
- Gary E. Marchant, “AI in Robes: Courts, Judges, and Artificial Intelligence,” 50 Ohio N.U. L. Rev. 473 (2024).
- Richard Re, “Artificial Authorship and Judicial Opinions,” 92 Geo. Wash. L. Rev. 1558 (Dec. 2024).
- Justin Snyder, “RoboCourt: How Artificial Intelligence Can Help Pro Se Litigants and Create a ‘Fairer’ Judiciary,” 10 Ind. J.L. & Soc. Equality 200 (Feb. 7, 2022). Tara Vasdani, “Estonia Set to Introduce ‘AI Judge’ in Small Claims Court to Clear Court Backlog,” Law360 Canada (Apr. 10, 2019).
- Eugene Volokh, “Chief Justice Robots,” 68 Duke L.J. 1135 (2019).