Within the universe of Alternative Dispute Resolution (ADR), arbitration enjoys a prominent status. If the disputing parties need a binding decision, but do not want to go to court, arbitration is the preferred dispute resolution method. Arbitration is a private and consensual method of dispute resolution, resulting into a binding decision: instead of state courts, it is a private tribunal appointed by agreement of the parties which renders a binding decision—the arbitral award.
Traditionally, the tribunal is composed of human arbitrators who conduct hearings in person. As the development of modern arbitration, and international commercial arbitration in particular, took place in the course of the 20th century, human-powered arbitration was the only technologically feasible possibility. Technological developments, however, especially digitization, artificial intelligence (AI), and blockchain technology, are currently disrupting the traditional format and conduct of arbitrations. Stakeholders in the arbitration market are exploring how new technologies and tools can be deployed to increase the efficiency (lower costs, higher speed) and quality of the arbitration process. Empirical research has shown that the latter factor, in particular, is crucial for parties when choosing arbitration over other dispute resolution processes. Intelligent machines hold the promise of more rational, consistent, and unbiased decisions when compared to human actors.
The COVID-19 pandemic will accelerate the trend towards using smart technologies to increase the efficiency and quality of arbitrations. For example, if physical hearings are not feasible, parties and tribunals require online meeting, desktop sharing, and video conferencing software that enables them to meet via the Internet in real time. Practical necessities and constraints prompt rapid, technology-assisted adaptations to the traditional way of ‘doing arbitrations’ by humans.
In a recent article, we investigate more radical questions, which come up if one thinks about the potential endpoint of the ongoing technological revolution of arbitrations. Does an arbitration require human arbitrators? Can it be conducted entirely by (artificially intelligent) machines? More specifically, can AI-powered systems manage a legitimate and fair arbitration process? Can they render a binding decision that qualifies as an arbitral award? If so, how good, in terms of costs and quality, are machine arbitrators compared to humans?
These questions, and the answers one gives to them, have a huge practical importance: for the economics of arbitrations (costs, speed) and for key legal issues, such as the quality of the delivery of justice, or the existence/challenge of an arbitral award. Even more importantly, these questions raise foundational conceptual issues. They force us to rethink the concept of an arbitration by considering its functions and desired effects, particularly legal effects. The concept and constitutive elements of an arbitration are ‘normatively loaded’. They reflect what a community of scholars, practitioners, and lawmakers (in a particular jurisdiction) believe an arbitration is, given the functions the process fulfils and the (legal) effects this community ascribes to it. These functions and standards are, then, translated into international and domestic legal standards.
In our essay, we argue that, when fully AI-powered arbitrations become technologically feasible, they will be able to functionally perform the same tasks as human actors and should be permitted by the law. There is nothing in the concept of an arbitration that fundamentally requires human control, governance, or even input. We further argue that the existing legal framework for international commercial arbitrations, the ‘New York Convention’ in particular, is capable of adapting to and accommodating fully AI-powered arbitrations. We anticipate significant regulatory competition between jurisdictions to promote technology-assisted or even fully AI-powered arbitrations, and we argue that this competition would be beneficial. In this competition, we expect that common law jurisdictions will probably enjoy an advantage—machine learning applications for legal decision-making can be developed more easily for jurisdictions in which reasoning is bottom-up, with case law playing a pivotal role.
States that do not wish to compete vigorously in the new market for technology-assisted or technology-driven arbitrations will nevertheless want to review their arbitration statutes and domestic legal systems generally to prepare themselves for more radical change, should this be required by market developments. This implies making arbitration laws enabling rather than mandatory unless compelling reasons require otherwise. Furthermore, it supports easing the coding of arbitration-related decisions and regulations and arbitral awards by making them public and easily accessible.
Horst Eidenmüller is Professor of Commercial Law, University of Oxford, and Professorial Fellow of St Hugh’s College, Oxford.
Faidon Varesis is a PhD Candidate, University of Cambridge.