18.12.2018 Filipe Galvão Teles

The Fifth Arbitrator? The Role of Artificial Intelligence to Tribunals in International Arbitration

1. Technological Wave

Around 20 years ago, Richard Susskind predicted that lawyers would regularly use e-mails to communicate with clients[1]. Time proved him right. Thereafter, 10 years ago, the same author predicted that “the foundations of dispute resolution [would] be rocked by a combination of electronic disclosure, e-filing in the courts”[2]. Once again, time proved him right. He now predicts that new technologies are one of the main drivers of a big transformation in the legal market[3]. Time will, most likely, prove him right yet again. Technology is, indeed, making its one way into the legal market.

According to “Moore’s Law”, a number of transistors which can fit in a computer double each 18 months to two years[4]. This means that computers increase their power exponentially, “starting out slowly and virtually unnoticeably, but beyond the knee of the curve it turns explosive and profoundly transformative”[5]; and we may well be in the knee of the curve already[6].

Artificial Intelligence (AI) is one of those new technologies. AI is a sub-field of computer science[7]. It is usually defined as the “science of making computers do things that require intelligence when done by humans”[8]. In fact, AI and other new technologies are already available in different legal areas and being used by law firms and courts around the world.

A good example is predictive coding[9]. Companies such as RelativityModusand Open Text provide services that operate under this technology and are already being used by dispute resolution teams in law firms. Courts in the USA[10], Ireland[11] and, recently, in the UK[12] have already permitted its use.

Likewise, technologies that enable users to search legal content, such as Ross Intelligence (ROSS), are also being used. ROSS and other similar tools allow users to ask legal questions using natural language search, then read through the entire body of law and return an answer with its level of confidence and relevant readings from legislation, case law and secondary sources[13].

This context, as it will be briefly explained, may change the way tribunals work in international arbitration. In particular, these technologies may partially substitute or complement the work currently carried out by arbitral secretaries, which are responsible for assisting arbitral tribunals.

2. Arbitral Secretaries: the Forth Arbitrator?

Almost 20 years ago, Constantine Partasides suggested that arbitral secretaries could be acting as the forth arbitrators in international arbitration disputes[14]. The author intended to discuss, clearly with a provocative tone, the role that arbitral secretaries were enjoying and the extent to which they ought to be doing so. Since then, the discussion has not evolved significantly and the regulations on the subject are scarce and divergent.

It is a well-established practice in international arbitration for arbitrators to seek assistance from arbitral secretaries to carry out not only administrative tasks, such as organizing meetings and hearings, but also more substantive functions, such as researching legal matters, providing summaries of legal submissions and partially drafting orders and awards[15].

However, it is generally recognized that arbitrators are under a duty not to delegate their responsibilities or tasks to third parties[16]. Therefore, the arbitral secretaries’ role has always been a controversial topic. The arguments against the use of arbitral secretaries – at least, to perform certain tasks – are grounded on the fact that the arbitrator’s mandate is intuitu personae and should not be shared with third parties[17]. For some commentators, the involvement of an assistant comprises a risk of improper influence on the arbitral tribunal[18].

3. Artificial Intelligence: the Fifth Arbitrator?

The technologies abovementioned and countless others for document analysis or generation of legal forms, briefs and memos are already developed and being used in the legal market. Therefore, it is my opinion that personal assistants may start giving place to technological assistants. Although secretaries may still be used extensively – not least due to all the benefits to the system arising from the education of younger lawyers on the arbitration practice, when acting as arbitral secretaries[19] –, some of their functions may start being performed by a faster and more accurate computer.

Thus, it would be interesting to shift the discussion on the use of secretaries to the use of AI tools. If an arbitrator uses an AI tool to make a legal research, and then uses the case law, laws, rules, guidelines, legal doctrines or any other information provided by the AI tool in an order or an award, he or she is actually delegating part of his or her role as an arbitrator. Regardless of the eventual review carried out by the arbitrator over the results provided by the computer, part of the arbitrator’s tasks related to the legal research was not performed by him or her.

The circumstances of this delegation are almost identical to the one that often occurs between arbitrators and secretaries. The differentiating factor is only the nature of the mandated entity, which changed from a human to a computer. Therefore, the debate about the use of arbitral secretaries may (and should) also be applied to the use of AI tools.

Thereby, the tasks that are considered as validly delegable to secretaries should also be considered as validly delegable to computers. If, under the legal framework applicable in international arbitration, secretaries are usually allowed to perform legal researches, to make summaries of legal submissions or evidence produced and to draft non-controversial parts of the tribunals’ decisions, there is no reason to substantiate a different position on the admissibility of the delegation when the mandated entity is a computer.

As far as the use of the AI program is not wrongfully influencing the decision-making process of the tribunal, its use should indeed be admitted under the same terms as those regulating the use of secretaries. In fact, the benefits assigned to the use of secretaries, such as the increased efficiency, are actually fostered with the use of AI tools[20]. Thus, it appears to me that the approach towards arbitral secretaries may, in general, be analogically applicable to computers.

However, it will certainly be more problematic to decide whether the tasks that are not validly delegable to secretaries may, under this new context, be validly delegable to computers. For instance, an AI tool able to critically assess the reliability of a certain piece of evidence or which is actually capable of developing a legal analysis of a case, if used, is performing a function which is generally forbidden for secretaries to carry out. Despite the positive impact on the tribunal’s management of the dispute, this delegation would certainly be unwelcomed by the advocates of a stringent approach on the tribunals’ obligation not to delegate.

If I previously advocated that what secretaries are allowed to do, computers should also be, in this case, though, I tend to reject the reverse argument which would conclude that, what secretaries are prohibited to do, computers should also be. Effectively, if previously the nature of the mandated entity justified the application of the same regime, now, the nature of the mandated entity may justify the application of a different regime.

Indeed, since computers are not natural persons, permanently affected with prejudices and influences, they may be better positioned than secretaries to perform more substantive functions, as the risk of a biased interference in the decision-making process is less intense or does not even exist.[21] Hence, with computers instead of secretaries, one could envisage (and look forward to) a more lenient approach to be adopted as far as the sphere of powers of the tribunals’ assistants is concerned.

In fact, this approach would certainly be more in line with the international arbitration practice, where arbitrators often delegate several administrative and more substantive functions to secretaries and junior lawyers from their law firms. 

Having said that, it seems appropriate to raise the following question: can, in a near future, AI tools assume the role of a fifth arbitrator as arbitral secretaries assumed the role of a forth arbitrator decades ago? Considering the new technologies already available and the number of companies specifically developing software for the legal market, the answer appears to be affirmative. One may even contemplate the possibility of AI assuming a role within the podium, i.e., being one of the elements of a three-member tribunal or even deciding the dispute on its own as sole arbitrator.

Hence, a more generic question follows: how will the international arbitration community react to this technological upsurge? Will it be keen to adopt and incorporate disruptive instruments which will change the status quo and, eventually, conflict with certain rules or regulations currently applicable?

In my opinion, despite the stand one may adopt on the arbitrators’ powers to delegate tasks to third parties, people involved in developing international arbitration should be generally open to that transformation. Only a prepared community will be able to adapt the applicable framework to the new context, avoiding forcing the new context to fit into outdated frameworks.

As Alfred Korzybski famously said, “the map is not the territory”[22]. The model of the reality should not be confused with the reality itself. Therefore, if the reality changes, the model should be adapted accordingly.


Author's Note:

This article was based on my final dissertation submitted in the context of the LL.M. in Comparative and International Dispute Resolution, at Queen Mary University of London. The dissertation, which is yet to be published, discusses the issues that are just briefly mentioned here. 


[1] Richard Susskind, The Future of Law (OUP 1996), Part V.

[2] Richard Susskind, “Hold to your seats, change is getting faster”, (24 January 2016) The Times, referenced in Thomas Schultz, Information Technology and Arbitration: A Practitioner’s Guide (Kluwer Law International 2006), Foreword by Gabrielle Kaufmann-Kohler.

[3] Richard Susskind, Tomorrow’s Lawyers: An Introduction to Your Future, (2nd edn, OUP 2017), ch. 1 3.

[4] Moore’s Law is a technological trend identified by Intel cofounder Gordon E. Moore in 1965.

[5] Ray Kurzweil, The Singularity Is Near, (Penguin 2005), p. 10.

[6] John O. McGinnis and Steven Wasick, “Law’s Algorithm”, (2013) Northwestern Univ. Sch. of Law Pub. Law & Legal Theory, Series no. 12-22, p. 3048.

[7] Kris Hammond, “What is Artificial Intelligence?”, Computer World, available at https://www.computerworld.com/article/2906336/emerging-technology/what-is-artificial-intelligence.html (last accessed 14 October 2018).

[8] Jack Copeland, “What is Artificial Intelligence?”, available at http://www.alanturing.net/turing_archive/pages/reference%20articles/what%20is%20ai.html (last accessed 14 October 2018).

[9] Predictive coding is an algorithm that can be taught how to interrogate a set of electronic data to identify certain types of documents accurately.

[10] See Moore v Publicis Groupe, 11 Civ 1279 (ALC) (AJP) (24 February 2012).

[11] See Irish Bank Resolution Corporation Ltd v Quinn [2015] IEHC 175 (3 March 2015).

[12] See Pyrrho Investments Ltd v MWB Property Ltd [2016] EWHC 256 (16 February 2016).

[13] Karen Turner, “Meet ‘Ross’, the newly hired legal robot”, Washington Post, available at https://www.washingtonpost.com/news/innovations/wp/2016/05/16/meet-ross-the-newly-hired-legal-robot/ (last accessed 14 October 2018).

[14] Constantine Partasides, “The Fourth Arbitrator? The Role of Secretaries to Tribunals in International Arbitration”, (2002) 18 Arb. Int’l.

[15] Michael Polkinghorne and Charles B. Rosenberg, “The Role of the Tribunal Secretary in International Arbitration: A Call for a Uniform Standard”, (October 2014) Dispute Resolution International, vol. 8, no. 2, pp. 110 et seq.

[16] See, for example, the AAA Code of Ethics, Canon V(c), the IBA Rules of Ethics, Article 2(3), and the English case Threlfall v. Fanshawe (1850) 19 LJQB 344 (QB).

[17] Eisemann, “Déontologie de l’arbitrage commercial international”, (1969) Rev. arb., pp. 217-229, Constantine Partasides et. al, “Arbitral Secretaries” in Albert Jan van den Berg (ed), International Arbitration: The Coming of a New Age? (ICCA Congress Series, Kluwer Law International, 2013), p. 328.

[18] Ibid.

[19] Pierre Lalive, “Inquietantes derives de l’arbitrage CCI (sur un récent “Oukase” du Secrétariat de la Cour d’Arbitrage CCI)”, (December 1995) ASA Bulletin 4, pp. 634-640.

[20]Young ICCA, “Young ICCA Guide on Arbitral Secretaries”, available at https://www.arbitration-icca.org/publications/Young_ICCA_Guide_on_Arbitral_Secretaries.html (last accessed 14 October 2018).

[21] A different question that may give rise to controversy is the one related to the way the AI programme has been trained. See, for instance, the issue related to the AI software created by Amazon to automatically sort through CV – which was biased against women – or the polemic around the computer program Compas used by an American court for risk assessment – which was biased against black people.

[22] Alfred Korzybski, “A Non-Aristotelian System and its Necessity for Rigour in Mathematics and Physics” (1931), American Association for the Advancement of Science, Science and Sanity, p. 58.

[Text originally published in Lisbon Arbitration]