The Rise of Artificially Intelligent Arbitration
Can a computer resolve a legal dispute? At first blush, the idea sounds ludicrous—even dystopian. But AI technology currently being developed seeks to predict how a judge will rule given a set of facts. Why not cut out the middleman, and simply have computer programs that adjudicate?
In the context of commercial dispute resolution outside traditional legal fora, the prospects for adjudication by algorithm, or even AI, are bright. Considering the main goals of alternative dispute resolution (ADR)—reducing costs, promoting adjudicatory expertise, and preserving business relationships between the parties—artificially intelligent arbitration will increasingly come into focus as the technology allowing digital decisionmaking develops. The rise of “smart contracts,” which are written in terms of machine-readable “if-then” statements, will also enable artificially intelligent arbitration.
Due Process and Contracts that Speak for Themselves
In my most recent post on legal automation, I wrote that the true value of hiring an attorney lies in making a real person responsible for your problems. In the context of commercial contracts, however, there is a particular drive to take the human element of interpretation out of the dispute. The goal, in other words, is to let the contract speak for itself. Technology makes this possible in a much more literal way than could ever have been envisioned by the framers of traditional, “four corners” contract law.
Arbitration by AI also sidesteps some of the due process concerns that would prevent, for example, replacing federal district court judges with supercomputers. Parties in digital arbitration would consent to the dispute resolution process in advance and a court could review the decisions of the “arbitrator” under the same principles as current arbitration jurisprudence. Arbitration rules, such as the American Arbitration Association’s Large, Complex Commercial Disputes procedures, can be easily adapted to fit AI adjudication. Parties could limit the issues that go before the digital arbitrator. For example, if the parties wanted to argue about discovery in front of a person, rather than a computer, they could so contract.
The Value of a Human Judgement
What is the value of human judgement in adjudication? It’s easy to imagine questions of law or fact that can be best resolved by a computer, such as the calculation of damages or applicability of certain contractual language to a particular interpretive issue. More difficult judgements involve considerations of justice, such as the weight that should be given to parties’ bargaining power in determining consent or the materiality of information to a particular dispute. Short of hooking witnesses up to a heart monitor during deposition, it’s nearly impossible to rely solely on objective information in determining the credibility of testimony. Over time, however, an artificially-intelligent arbitration program could conduct a number of these adjudications and “learn” how to balance considerations of justice and fairness. AI adjudicators are only as cold as their programming.
Over repeated iterations, AI will gain sufficient experience to enable it to conduct all but the most difficult, fact-intensive adjudications. Consequently, the value of a human adjudicator will diminish. Human arbitrators will assist, rather than direct, the process of submitting evidence and conducting interrogatories. Again, all this is consistent with the movement aimed at reducing the human element in contract law. Moreover, the decision to use automation will be driven by the same consideration that primarily drives arbitration in the first place: cost.
Recommendations for Automated Arbitration
Arbitration rules should provide for judicial review in exceptional circumstances or under a deferential standard such as “clear error.” In order to facilitate judicial review and preserve notions of due process, the program should issue a reasoned judgement in natural text; this would form the basis for a record on appeal. An arbitration program should also be able (similar to a jury) to certify questions to the parties or an overseeing judge. There should be a presumption that certain issues, such as contract formation, are to be handled by a human being. Having a computer determine whether you have consented to its authority crosses the line from efficient to dystopian. In the end, automated arbitration will succeed in contexts where the parties seek to eliminate the ever-fallible human element. In some sense, then, this is but the logical endpoint of the alternative dispute resolution movement.