Arbitration and AI – friends or foes?
This article is part of our Artificial Intelligence Insights Series, written by McCarthy Tétrault’s multidisciplinary Cyber/Data team. This series brings you practical and integrative perspectives on the ways in which AI is transforming industries, and how you can stay ahead of the curve.
View other blog posts in the series here.
Arbitration is a flexible, efficient and cost-effective way for parties to resolve disputes, often with the added benefits of a confidential, tailor-made process that suits the parties and the dispute, and that can lead to a quick and final determination. The inherent flexibility of arbitration should allow arbitration counsel and arbitrators to be among the vanguard of the legal profession in adopting artificial intelligence (“AI”) and other technologies. When used appropriately, AI can limit and/or eliminate time, costs and unconscious biases in the arbitration processes, for example using AI technology to review millions of records more objectively for relevance, privilege and materiality.
On the other side of the debate are concerns about whether the use of AI makes the arbitration process less human. If true, it could impact some of the reasons for which parties choose arbitration in the first place, such as the ability to have arbitrators apply their expertise and industry knowledge to the dispute. Balancing these potential positive and negative implications of AI in arbitration will be key as AI continues to play a more significant role in arbitral dispute resolution.
Choosing an Arbitrator
When selecting an arbitrator — arguably the most important decision in the process — parties and their counsel typically weigh a number of criteria. Parties consider the arbitrator’s experience, research, reputation, and word of mouth testimonials to generate lists of potential arbitrators as a first step to seeking agreement with the opposing party. This can be a time-consuming process impacted by biases and informational deficits of the humans involved.
AI could offer improved objectivity and diversity in arbitrator selection. AI powered tools such as Arbitrator Intelligence and Billy Bot collect information and feedback about arbitrators from around the world and make recommendations to parties in light of their stated preferences, from area of expertise to cost. AI tools used in the arbitrator selection process can also promote diversity and, to the extent that an arbitrators’ past decisions are accessible, can provide the parties with some predictability in how their dispute may be resolved.
AI technologies are instrumental in the e-discovery and information gathering process that arbitration entails. For complex and technical disputes — and many commercial arbitrations can be so described — the parties may be required to sift through hundreds of thousands or even millions of documents. AI can make this process significantly more efficient. Heavy volumes of evidence can be translated, summarized, and assessed by simple classification, coding and clustering models that would otherwise require thousands of hours of human attention.
Legal Research and Drafting
AI tools can translate, transcribe, and summarize evidence. They can assist parties and their counsel — and arbitrators themselves — with legal research and analysis, at significant cost savings. AI may even assist arbitrators in preparing their awards, particularly when an arbitral institution has imposed specific requirements for the structure and formulation of arbitral awards.
AI can also be a “tool of predictive justice” and consequently, a cost-effective way to advise a party to a dispute on the chances of success. With this knowledge, a party can model their approach to a particular dispute with a realistic sense of its possible outcome.
The (potential) Negatives
An area of growing concern within the arbitration community is the use of AI tools as a standalone decision-maker. AI has already been used as a tool for decision-making in several countries: Estonia is currently developing an AI “judge” that can adjudicate small claims disputes of less than €7,000; a court in Beijing is presided over by an AI judge; and American courts have – controversially - used AI tools in criminal proceedings to determine the possibility of parole based on likelihood of recidivism.
The idea of using AI to resolve straightforward disputes is attractive in some circumstances. However, it becomes far less attractive when you take into account the prevalence of complex and novel legal issues, public policy and equity concerns, and AI’s general inability to understand the humanity in legal disputes. With these concerns in mind, trusting the resolution of complex legal disputes to a machine should not be considered a replacement for thoughtful, independent arbitrators now (or maybe ever!).
Complex Nature of Disputes
Arbitration is designed to solve unique disputes between specific parties in a unique and tailor-made way and is not designed to apply widely to other parties or issues. AI systems of all kinds require large data sets to produce reliable outcomes. Insufficient data sets (which result from the confidential nature of arbitration and the lack of publication of awards) means that complex, uncommon, or novel features of a dispute cannot get the consideration they deserve and would receive from a human decision-maker.
Administration of Justice Concerns
While AI systems are incredibly complex and are at the forefront of technological innovation, they, like humans, are far from perfect. These imperfections can result in arbitral decisions that run contrary to public policy.
For one, AI systems frequently fall into a trap called “overfitting” – in which the AI system learns the idiosyncratic features of the data so much that the machine will create a pattern that fits the data rather than allowing the data to speak for itself. 
Additionally, AI systems can fall victim to bias both in how they are made and in the information available to it. If the arbitral decisions that are used to train and build an AI arbitrator reflect a pattern that is biased, the resulting decision could reflect that same bias. Beyond this, biases are often more easily observed in human ways, like by how the arbitrator questions the parties, or by how much attention they pay or don’t pay when one party is speaking. With a machine, the biases are much harder to spot and rectify.
Importantly, there are also fundamental parts of the arbitration process that AI cannot yet replicate. Arbitrators are tasked with hearing oral evidence, assessing credibility of witnesses, and providing a well-reasoned decision that explains the analysis and conclusions. AI technology does not allow it to assess the human emotions and behaviours required to make findings of credibility. Additionally, at this time, AI arbitrators are unable to provide well-reasoned decisions that explain their thought process and carefully and systematically dispose of each of the parties arguments. It should be noted this concern is not unique to the arbitration process.
Beyond that, arbitrators are typically careful to ensure both sides receive a full and fair opportunity to present their case and feel heard. AI would not be able to provide parties with the same satisfaction of “being heard” and understood in the way a flesh and blood human can. At this time, AI arbitrators also are not capable of understanding the importance of providing the parties with a fair hearing. In the end, parties turn to arbitration because of its flexible nature and the ability it provides to reach an outcome that is fair and just to both parties. Without fairness, we lose a fundamental benefit of turning to arbitration in the first place.
Fundamentally, the legal framework of arbitration and dispute resolution as a whole is designed with human decision-makers in mind. Inherent in this is cognitive and emotional capabilities that AI does not possess. Human arbitrators do not apply the law purely as a matter of logic or patterns, but through an understanding of the details of a dispute and the parties to it. For example, judges and arbitrators frequently make decisions based on public policy, equity and fairness, principles that would be difficult for AI to understand and consider. While a human arbitrator would be able to weigh the policy and fairness implications of their decision, it is not proven that a machine can do the same. While an AI system may be able to predict the answer to a discrete and straightforward legal problem, the resolution of complex arbitration disputes is far from straightforward.
The Way Forward
While AI benefits arbitration at all steps of the dispute resolution process, we are not yet at the point where it will give human arbitrators a run for their money. Humans are still a necessary feature of the arbitration process and will be best positioned to embrace the benefits that AI has to offer. While we recognize and appreciate the benefits of AI in arbitration (and beyond), we also understand the importance of knowing the current limitations of AI and the need for humanity within the arbitration process. Walking that line is the continued challenge – one we look forward to daily.
 Horst Eidenmuller and Faidon Varesis, “What is an Arbitration? Artificial Intelligence and the Vanishing Human Arbitrator,” New York University Journal of Law and Business (2020) Vol. 17. No. 1, at p. 51.
 Young-Yik Rhim and KungBae Park, “The Applicability of Artificial Intelligence in International Law,” Journal of East Asia and International Law (2019) Vol. 12, No. 1, p. 18.
 See https://arbitratorintelligence.com/about/ for more information about their services.
 See https://robotlawyerlisa.com/billy-bot/ for more information about their services.
 Horst Eidenmuller and Faidon Varesis, “What is an Arbitration? Artificial Intelligence and the Vanishing Human Arbitrator,” New York University Journal of Law and Business (2020) Vol. 17. No. 1, at p. 56.
 With the right team, AI can be utilized to streamline the e-discovery process in a cost effective way. McCarthy Tétrault’s own e-discovery division, MT>3, has deep and extensive knowledge of all the ways that AI can and should be used to assist clients of all kinds.
 Horst Eidenmuller and Faidon Varesis, “What is an Arbitration? Artificial Intelligence and the Vanishing Human Arbitrator,” New York University Journal of Law and Business (2020) Vol. 17. No. 1, at p. 59.
 Aditya Singh Chauhan, “Future of AI in Arbitration: The Fine Line Between Fiction and Reality.” Kluwer Arbitration Blog. (2020) https://arbitrationblog.kluwerarbitration.com/2020/09/26/future-of-ai-in-arbitration-the-fine-line-between-fiction-and-reality/.
 Gizem Halis Kasap, “Can Artificial Intelligence (“AI”) Replace Human Arbitrators? Technological Concerns and Legal Implications”, Journal of Dispute Resolution Vol. 2021 p. 209.
Wisconsin v Loomis, 2016 WI 68.
 Gizem Halis Kasap, “Can Artificial Intelligence (“AI”) Replace Human Arbitrators? Technological Concerns and Legal Implications”, Journal of Dispute Resolution Vol. 2021 p. 224.
 Gizem Halis Kasap, “Can Artificial Intelligence (“AI”) Replace Human Arbitrators? Technological Concerns and Legal Implications”, Journal of Dispute Resolution Vol. 2021 p. 225.
 Derick H. Lindquist and Ylli Dautaj, “AI in International Arbitration: Need for the Human Touch,” Journal for Dispute Resolution (Vol. 2021), at p. 53.
 Derick H. Lindquist and Ylli Dautaj, “AI in International Arbitration: Need for the Human Touch,” Journal for Dispute Resolution (Vol. 2021), at p. 40-41.
- AI and its Role in the Anne Frank Cold Case
- AI in the Canadian Financial Services Industry
- An AI a Day Keeps the Doctor Away?: Regulating Artificial Intelligence in Healthcare
- Considerations and Risks of AI in Employment Decisions
- Could AI get you sued? Artificial intelligence and litigation risk
- AI and Patent Law: Can AI Be an “Inventor”?