Mauro Fragale and Valentina Grilli[1]
1. Introduction
Artificial Intelligence (AI) refers to the development of computer systems that, by utilising machine learning, neural networks, and data analysis, are capable of performing tasks that would normally require human intelligence. AI has advanced to the point where it can be used to autonomously resolve many day-to-day legal issues: for instance, in copyright infringement cases, courts can employ AI rather than experts to more objectively determine if a creative work is substantially identical to a previous one.[2]
The employment of AI has been expanded. AI is, for example, increasingly utilised in international commercial arbitration, an alternative dispute resolution mechanism often preferred in the resolution of international business controversies for its efficiency, flexibility, and the ability to select expert arbitrators.[3]
One of the applications of AI to arbitration consists in the employment of the machine as the decision maker: AI can help – or potentially replace – arbitrators in deciding the dispute.[4] As a case in point, eBay deploys AI-driven ‘robot arbiters’for the automatic resolution of customer complaints.
However, this paper will focus on a different aspect, which is the use of big data and machine learning to select arbitrators that better match the characteristics of the dispute at hand. Indeed, in the authors’ view, such a system ensures that the person appointed to deliver the award on a specific subject matter is the best equipped to do so in terms of expertise, competence, and preparedness. This circumvents several challenges and limitations that may arise when the parties come to appoint arbitrators.
2. Current Usage of Data Analytics in Arbitrator Selection
In international arbitration, the integration of AI has been predicted to encompass a diverse range of functions, including – inter alia – legal research, streamlining procedural processes, and the identification of appropriate professionals such as experts, counsels, and arbitrators.[5] Noteworthy examples of AI-powered tools in this context include LexisNexis, DoNotPay, ExaMatch, and Ross Intelligence, which are employed for legal research, examination of legal documents, and even seeking legal advice.
Particularly significant to this paper is the utilisation of AI to select suitable experts, counsels, and arbitrators, with software like BillyBot and Arbitrator Intelligence. Specifically, BillyBot is a digital junior clerk capable of understanding user queries, retrieving information to generate responses, and finding the right barrister or mediator for a specific legal problem. In parallel, Arbitrator Intelligence is an algorithm that collects and analyses legal case data to unveil patterns in arbitrators’ decision-making, helping parties make informed choices in selecting arbitrators for their dispute.
There is a widely acknowledged understanding that arbitral proceedings safeguard confidentiality and arbitral awards concerning commercial disputes typically are unpublished. This opacity concerning the actual arbitrators’ experiences and their standard of decision-making renders the task of identifying the most fitting arbitrator a formidable challenge.[6] AI can provide a solution to the lack of transparency, and this is the sphere to which Arbitrator Intelligence belongs.
Arbitrator Intelligence has created the Arbitrator Intelligence Questionnaire (AIQ), a structured feedback instrument crafted to systematically gather insights regarding arbitrators’ historical case management and adjudicative approaches.[7]The process is composed of two stages: Phase I focuses on basic objective information about the case and may be completed by anyone with access to the award or file; Phase II contains questions about the conduct of the arbitration and, potentially requiring professional judgement, should be completed by a lawyer involved in the proceedings.
Essentially, the initial part of the process establishes the factual groundwork, while the second one delves deeper into the procedural dynamics and subjective assessments. Once an adequate corpus of data has been amassed via the AIQ, Arbitrator Intelligence periodically publishes AI Reports. The aspirations of this initiative include delivering insightful feedback, avoiding questions that may inadvertently exhibit biases for specific cultural or legal frameworks, upholding impartiality towards arbitrators, and encouraging methodical and comprehensive responses.[8]
To ensure the efficacy of this tool, it is necessary to foster a widespread participation in the completion of AIQs. In response to this challenge, Arbitrator Intelligence has initiated collaborative partnerships with diverse international arbitral institutions, which take on the role of disseminating the AIQ to pertinent parties and legal professionals upon the culmination of each arbitration. In reciprocation, Arbitrator Intelligence provides collaborating entities with unrestricted access to AI Reports.[9]
3. Benefits and Challenges of AI in Arbitration
As the field of international arbitration continues to broaden – together with the growing involvement of participants coming from geographically diverse areas – the proposal of relying on data analysis and AI progressively becomes more relevant.[10] In this context, it is crucial to weigh the benefits and drawbacks of utilising this technology for the selection of decision-makers.
Undoubtedly, employing AI tools for arbitrator selection offers a wide range of advantages. First and foremost, it allows for an impartial and unbiased selection, as relying on data-driven analysis mitigates the influence of human biases and subjective factors that could potentially infringe upon the integrity of the selection process.
In this regard, comprehensive research has shown a prevailing trend wherein the composition of international panels is primarily directed towards men from North America and Europe, underrepresenting women and other minorities.[11] The discernible dominance of specific demographics among international arbitrators calls for a proactive response; thus, a second advantage of AI-powered selection is the potential to increase diversity by encompassing a broad spectrum of arbitrators from various jurisdictions, backgrounds, and cultural contexts.
Thirdly, AI can identify arbitrators possessing the specific expertise that is most suitable for the subject matter of the dispute. This improves the quality of arbitrator selection by guaranteeing the presence of individuals who possess pertinent knowledge and extensive experience within the relevant domain. Finally, the process of appointing arbitrators guided by AI exhibits transparency, given that determinations are based on measurable criteria and historical data. This transparency fosters confidence among stakeholders by providing a clear rationale for the chosen arbitrators.[12]
Nevertheless, there are legitimate concerns regarding bias and fairness that warrant careful attention in utilising AI tools for arbitrator selection. For instance, by selecting a specific individual based on the predicted decision in a dispute, the AI could potentially strengthen established patterns:[13] although AI is designed to be objective, it can involuntarily perpetuate or even amplify biases present in the data on which it is trained.
With regard to the case of Arbitrator Intelligence, for example, the AIQ incorporates several features intended to guarantee the accuracy and impartiality of the arbitrator assessments. The primary mechanism ensuring equity is rooted in the fact that a vast majority of inquiries within the AIQ are directed towards gathering objective information. Conversely, those questions which deviate from purely objective parameters are crafted to solicit ‘professional judgment’ specifically concerning an arbitrator’s identified conduct and attributes.
Yet another potential concern that may emerge pertains to the inherent opaqueness in the operations of AI systems. This phenomenon is commonly referred to as the ‘Blackbox issue’, which means that the inner workings of certain AI algorithms are not readily comprehensible, rendering their decision-making processes obscure to human understanding. To mitigate the lack of transparency, it is necessary to develop explainable AI models and to cultivate stakeholders’ trust in these tools via regular testing of their operations.
Lastly, concerns can emerge vis-à-vis the obligation to thoroughly examine the impartiality and independence of the appointed arbitrators, i.e. verifying the absence of conflict of interest.[14] Indeed – both in accordance with Article 5 of the LCIA Arbitration Rules 2020 and under the General Standard (1) of the IBA Guidelines on Conflicts of Interest in International Arbitration 2014 – each arbitrator must maintain impartiality and independence from the involved parties from the moment they accept their appointment, continuing to uphold these qualities until the rendering of the final award or the conclusion of the proceedings.
The remedy to this challenge lies in the fact that, while AI technology can intervene in the preliminary identification of potential arbitrators, the ultimate decision to accept the case remains in the hands of the arbitrators themselves. They are entrusted with the responsibility of evaluating their own circumstances and disclosing any potential conflict of interest, ensuring that they can fulfil their role without any bias or partiality.
4. Conclusion
Arbitrators play the most prominent role in ensuring that the delivered award is a fair and effective resolution of the dispute. In fact, in the valuable words of esteemed lawyers Blackaby, Partasides, and Redfern: ‘It is, above all, the quality of the arbitral tribunal that makes or breaks the arbitration.’[15] The challenges highlighted by this paper may indicate that, at the current state, AI is still not ready to carry out the task of selecting the perfect arbitrators for the case at hand, due to its blind reliance on training data, dubious transparency, and incapacity of recognising conflicts of interest.
Nevertheless, AI is a powerful technology that is evolving at a fast rate, has the potential of solving many of the issues related to party selection of arbitrators, and will definitely change arbitration as we know today. As stated by Professor Rogers, founder of Arbitrator Intelligence: ‘Arbitrator Intelligence will liberate arbitrator selection from the 19th Century’s telephone and introduce it to the 21st Century’s data-driven analytic solutions’.[16]
—
[1] Final year Law Students at Bocconi University, Milan.
[2] Georgios I Zekos, Advanced Artificial Intelligence and Robo-Justice (Springer 2022) 245.
[3] Gerard AW Vreeswijk and Arno R Lodder, ‘GearBi: Towards an online arbitration environment based on the design principles simplicity, awareness, orientation, and timeliness’ (2005) 13 Artificial Intelligence and Law 297.
[4] Maxi Scherer, ‘Artificial Intelligence and Legal Decision-Making: The Wide Open?’ (2019) 36(5) J Intl Arbitration 539, 541.
[5] Zekos (n 2) 340.
[6] Faye Fangfei Wang, Online Arbitration (Informa Law from Routledge 2018) 90-1.
[7] Catherine A Rogers, ‘Arbitrator Intelligence: From Intuition to Data in Arbitrator Appointments’ (2018) 11(1) NYSBA New York Dispute Resolution Lawyer 41.
[8] ibid 42.
[9] ibid.
[10] Gary B Born, International Commercial Arbitration (3rd edn, Kluwer Law International 2021) last updated September 2022, [12.03].
[11] Deborah Rothman, ‘Gender Diversity in Arbitrator Selection’ (2012) 18 Dispute Resolution Magazine 22, 23.
[12] Ali Yesilirmak, ‘Transparency and Stakeholders’ Role in the Selection of the Arbitral Tribunal’, in Stavros Brekoulakis, Romesh Weeramantry, Lilit Nagapetyan (eds), Achieving the Arbitration Dream: Liber Amicorum for Professor Julian D.M. Lew KC (Kluwer Law International 2023) 292.
[13] Francisco Uribarri Soares, ‘New Technologies and Arbitration’ (2018) 7 Indian J of Arbitration Law 84.
[14] Mel Andrew Schwing, ‘Don’t rage against the machine: why AI may be the cure for the ‘moral hazard’ of party appointments’ (2020) 36 Arbitration Intl 491, 494.
[15] Nigel Blackaby, Constantine Partasides, Alan Redfern, Redfern and Hunter on International Arbitration (7th edn, OUP 2023) [4.41].
[16] Rogers (n 7) 41.