Artificial Intelligence Will Soon Replace Many Human Arbitrators

Forbes

11-06-2025

Artificial intelligence is already affecting the way that we resolves disputes.
Arbitration is a process whereby the parties to a dispute agree that it will be privately decided outside of the normal court process. Instead of the formal and heavily-structured litigation procedures of our courts, which can be very expensive and take literally years to reach a conclusion, arbitration offers a much more informal and less-structured process where a result can be reached within just a few months for a fraction of the cost. Instead of a judge, the parties to an arbitration hire what amounts to a private judge, known as the arbitrator of course, to oversee the arbitration process, consider the law and relevant evidence, and come to a conclusion.
The downside to arbitration is that it amounts to a giant corner-cutting of the normal litigation process. The parties to an arbitration may or may not have access to discovery, which at any rate will be abbreviated and typically less than through the courts, and generally there is no effective appeal of the arbitrator's opinion. While the results reached by our courts are certainly not perfect ― there is a reason why we have full-time courts for appeals ― those results are likely by several magnitude to be more accurate than those reached by arbitration.
Basically, the parties to an arbitration are trading the potential quality of the result to get the dispute over with as cheaply and quickly as possible. The entire philosophy of parties who agree to arbitration may be summed up as "get the dispute over with and get back to normal business as soon as possible." Particularly for large and deep-pocketed enterprises that are frequently involved in litigation, and for whom the results of litigation are viewed simply as an aggregate rounding error as opposed to a life-changing event, arbitration is a far more efficient method of resolving their myriad disputes than spending oodles in courtroom litigation.
This now brings us to the arbitrators. The goal of these private judges is, again, to review the law and the evidence and reach a conclusion as quickly and efficiently as possible. In line with the purposes of arbitration, the goal is to resolve the dispute one way or another so that the parties can move on. The arbitrator's award is usually final since the courts may only review an arbitration award for some very serious error, sometimes known as manifest error, such as where the arbitration process in a particular case was corrupt or if the arbitrator exceeds his or her authority. Yes, some arbitration agreements provide for an appellate process, but these are quite rare.
Most arbitrators are highly conscientious professionals who take their job very seriously. They carefully review the law and the evidence and attempt to reach the result that they think is correct under the circumstances. These arbitrators offer the parties a quality of decision that is at least equal to the courts and in some situations (particularly with specialized subject-matter) might even be better. Unfortunately, the better arbitrators naturally tend to end up with the more significant cases. Downstream, the picture isn't so pretty.
The sad truth is that I've seen a lot of bad arbitration decisions of the years. Cases where it was obvious that the arbitrator simply "mailed it in" without spending much time trying to figure out the law or analyze the evidence. This is a result of mediocre lawyers becoming mediocre judges (more often than not through an election) and then leaving the bench early so that they can make the big bucks as a mediocre arbitrator. They don't have to worry about the quality of their decisions either, because there is no court of appeals looking over their shoulders. This isn't rough justice ― it's really not justice at all. Honestly, the parties should instead just flip a coin and live with the result. These mail-it-in arbitrators tend to congregate towards the lower-end cases. You might think this means that they are doing the least amount of damage here, but these smaller arbitrations often involve small businesses and individuals who don't have the financial strength to simply write off a bad decision as a rounding error in the same way that large enterprises do.
Luckily, a new solution is likely to appear in the way of using artificial intelligence as a passable substitute for arbitrators in these smaller cases. Lawyers are already using AI to analyze evidence, summarize witness testimony, and conduct legal research. It would be a relatively small additional step for the parties to an arbitration to put together their briefs and evidence, upload it to the server, and then ask the AI program to render a decision. Two minutes later that decision will appear.
The immediate objection will be that, "AI makes too many mistakes". While it is true that AI can make mistakes, it is also true the human arbitrators make mistakes as well. While the mistakes of most human arbitrators are honest, human arbitrators still suffer from implicit bias, they misread court opinions or err in their analysis of the evidence. Then, as discussed above, some human arbitrators at the lower levels are some combination of intellectually lazy and incompetent which make their arbitration awards little more than a form of random justice ― the parties might as well just flip a coin ― and certainly even a glitchy AI algorithm will be better than that.
To take just one example of implicit bias, there has long been a concern that certain arbitrators working for certain arbitration firms will tend to find in favor of the firms who send them the most business. Or, in other words, some arbitrators are "voting their wallets" instead of strictly in favor of the merits of a particular case. AI holds the potential to eliminate this form of implicit bias, as well as other implicit biases arising from the race, sex, religion, or socio-economic status of a party. Unlike a human arbitrator who cannot "unring a bell" and may be swayed by evidence ultimately deemed to be inadmissible, an AI program can completely discount evidence that it deems to be inadmissible and without subliminally taking it into account.
Admittedly, while the move from human arbitrators to AI takes place, it will likely be necessary for some human review to take place to make sure that the AI analysis hasn't gone totally off the rails. But since AI is inherently a learning algorithm, over time this constant human review may be dispensed with and the courts can review the arbitration awards as a perquisite to enrolling the award as a legal judgment just as they do now. But here it is to be remembered that if a human makes a mistake, probably only that particular human learns from it. If AI makes a mistake, the entire AI service learns from it. So, as we have seen from news reports of lawyer who have inadvisedly filed briefs AI-written legal briefs, AI will often just make up legal authorities out of thin air. But the AI algorithms can be modified going forward to provide for greater self-checking and accuracy, to eventually a degree of accuracy that even the most meticulous arbitrator would be hard-pressed to replicate.
We have not yet discussed another advantage of AI which is the cost-savings that it potentially offers to litigants. Other than their own attorney fees, the single largest cost in any arbitration is of course the arbitrator's fee. With arbitrators charging many hundreds of dollars per hour (and with some superstar arbitrators now having an hourly rate in excess of $1,000), the fees ultimately charged by an arbitrator in a particular case can be very expensive. By contrast, the token charge for AI to make the same decision might be only $50 or something. Indeed, AI arbitration may prove to be so cheap that the parties could submit their arbitration briefs with attached evidence to several different AI services to check the results for accuracy of the decisions and still probably not exceed the cost of a single hour of a human arbitrator's time.
It should be expected that the large legal database firms such as Westlaw and Lexis will develop these AI arbitration services sooner rather than later. They already have up-to-date legal authorities pre-loaded for AI consumption and they are both offering AI legal research services to attorneys already. Presumably, somebody at these firms will one day wake up and see that they are missing out on a potentially significant share of the lucrative arbitration marketplace and soon thereafter add arbitration consoles to their existing service packages.
As with all things, AI arbitration will have its limits. For some time, the larger arbitration cases will likely require human arbitrators to make the difficult decisions regarding the deep nuances of the law and evidence which make the difference in such cases. But for the smaller arbitration cases where the parties are just eager for some quick and cheap outcome whatever it is, AI arbitration holds substantial promise.
AI arbitration is coming, it is only a matter of when.