The introduction of the AI Arbitrator has sparked strong reactions across the legal community. Some lawyers and arbitrators are intrigued by its potential, while others question whether AI can play a meaningful role in deciding disputes.
I fall in the first camp. For me, the AAA’s AI Arbitrator holds great potential and is a worthwhile – and necessary – experiment.
The speed at which AI is evolving is unlike anything the legal field has seen before. While still relatively new, AI systems are already capable of doing things that would have seemed implausible at launch, and they continue to improve at an extraordinary rate. Against that backdrop, to say that AI will never be in a position to resolve disputes – which has been a common reaction of those in the arbitration community – strikes me as short-sighted.
Are there any types of arbitral disputes that AI can decide right now? I do not know. There is only one way to find out – by trying. That is exactly what the AAA is doing with the AI Arbitrator.
Importantly, the AAA is releasing this product in a narrow and controlled way. Currently, the AI Arbitrator is limited to two-party, document-only construction disputes. Participation is entirely voluntary, meaning both parties must opt-in to use the platform. As a practical matter, this means the cases will tend to be relatively small and straightforward. Because there are no live witnesses, there is no concern about AI deciding issues of witness credibility.
It is also important to be clear about the role the AI will play in these disputes. The AI Arbitrator does not issue the final decision. A human arbitrator reviews the AI’s analysis and draft decision and is responsible for issuing the final award. Some cynics will say that the human arbitrator will just rubber-stamp the AI decision, but I prefer not to be cynical. I prefer to believe that the human arbitrators will recognize that they are part of a very important experiment and will be extra vigilant in scrutinizing the decision of the AI Arbitrator. That is what I would do in that situation.
With all the other tasks that AI is performing, why should properly trained AI models not be able to "decide" simple legal disputes, especially document only disputes? And, if not now, then almost certainly soon. Court dockets are clogged with many relatively small, relatively straightforward cases (though not necessarily that small to the parties). The need for a human judge to decide each such case is the obvious bottleneck in the system. If, on an opt-in basis, parties can choose to have an AI Judge decide their case immediately, that could be a boon for access-to-justice concerns. But the only way that can happen is if there is proof that AI "judges" get the "right" result. Establishing in the arbitration setting the use case for accepting AI as a sound method of dispute resolution is, in my view, a step down the path of being able to use AI in the courts.
Whether the results ultimately support broader adoption remains to be seen. And even if the results are positive, there will definitely be other issues to address, such as whether courts will enforce an award from an AI arbitrator. But engaging with the technology—rather than dismissing it—is essential for the legal community to better understand where and how it can be used effectively.
At this stage, the most sensible course is a simple one: give AI a chance.