Legal writings and conference papers discussing technology in arbitration almost inevitably touch upon the ability of artificial intelligence (AI) to assume the decision-making role. In a recent Young ICCA blog post on machine arbitration and machine arbitrators, Jack Wright Nelson offers an interesting analysis as to why this type of dispute resolution would be acceptable.
On a personal note, I am concerned by these trends, simply because I still hope to earn my living by practising international arbitration in 20 years’ time. I cannot disagree with the conclusion that machine arbitration should be possible even under the current legislation, let alone legislation that will inevitably change over the course of two decades, what with the technological advancements in all spheres of our lives. I also have no reason to doubt the assumption that, very soon, AI could take over some part of legal business (for example, legal research and drafting) if it is able to produce more consistent and reliable results at much cheaper costs. However, I keep asking myself (perhaps selfishly) whether machine arbitration is something users would actually want.
Arguably, AI will be better able to conduct legal research than humans. Likewise, it is quite possible that AI will be better at fact-finding. On the face of it then, I believe that the answer is likely to be yes. Most businesses would be happy to have a dispute resolution mechanism which produces consistent and accurate results at a fraction of the costs involved in today’s arbitrations. There is perhaps less room for scepticism about machines judging humans in the commercial context. For this reason, AI-conducted dispute resolution will likely take some part of the alternative dispute resolution (ADR) marketplace sometime soon. Arguably, it will leave very little room for standard disputes. By using the same AI technology, without formally engaging in a dispute with an opposing party, outcomes might be easier to predict. So, to be precise, it will be an alternative dispute prevention mechanism.
That being said, AI will not wipe out arbitration as we know it now (at least I hope not), because resolving international disputes goes beyond establishing the factual matrix of a case and applying established legal principles.
Firstly, it may be easy to analyse a wealth of well-reasoned common law judgments and come to more or less accurate conclusions. Humans can do it, so AI could do it better. However, in many jurisdictions, court judgments (if at all available) are much shorter and the reasoning is much sketchier, with very little (if any) account of the relevant facts and parties’ arguments. As a result, legal research requires experience in a particular jurisdiction and knowledge of the historical roots of various legal principles and traditions. It becomes more nuanced and, to a large extent, resembles a guessing game where you have to use your intuition and everyday life experience to understand the reasoning behind a particular judgment. International arbitration is there to resolve disputes from all over the world between the parties from different jurisdictions with different legal traditions. There is no reason to assume that AI would match the quality and accessibility of information available for research in every jurisdiction any time soon.
Similar considerations could also go to the fact finding. The analysis of the information, which is well documented and consistent, is, again, not too difficult a task for a human; therefore it could be done easily by AI. But the fact-finding function of the arbitrator goes beyond that because, in many complex disputes, the factual matrix is fragmented. Some facts may not be established with sufficient certainty. Even the standard of proof applicable in non-criminal law cases requires something which goes beyond pure analytical skills. Arbitrators may need to draw on their personal business experiences and, indeed, everyday life. They may need to utilise their understanding of various cultures, business traditions and human psychology to find which interpretation is more likely (on the balance of probabilities) or which one is actually to be believed (on the inner conviction standard); they then need to understand the links between isolated proven facts.
Furthermore, in very many cases the application of the law to the established facts does not necessarily produce the right result for one reason or another. Some crucial facts cannot be established with the required certainty. Sometimes, the law, if applied strictly, may lead to unfair results. Sometimes the law may not even exist to regulate a particular situation and the parties may even ask their arbitrators to resolve the dispute ex aequo et bono (according to the right and good). All of these situations require that human senses of justice and fairness are involved in the decision-making process.
So AI may be suitable for smaller and more straightforward cases. But the real value of international arbitration as the means of peaceful resolution of international disputes between various actors is not in producing standardised and consistent results: it is not a dispute resolution conveyor belt. Its real value is that the people involved in the process (arbitrators, counsel, experts) use their own cultural, business and legal backgrounds, life experience, and senses of fairness and justice to achieve the resolution of the dispute. This is often not only correct as a matter of strict law, but is most important as just and fair under the circumstances of the case. Personally, I do not believe that this is something which can be replicated by AI, no matter how developed it becomes. This gives us a chance to ensure that we are still in our jobs 20 years from now. Or perhaps I’m just trying to comfort myself.
