KluwerArbitration

Document information Chapter 8: When the Tribunal Is an Algorithm:

Complexities of Enforcing Orders Determined by a
Publication
60 Years of the New York
Software under the New York Convention
Convention: Key Issues and Huang Ying (Irene) Ng ; Valeria Benedetti del Rio
Future Challenges
SUMMARY
Bibliographic Reference Integrating aspects of traditional arbitration with computers or online applications has
been a growing trend with the continued development of technology in the legal industry.
Huang Ying (Irene) Ng and Several processes of traditional arbitration are slowly being replaced by their digital
Valeria Benedetti del Rio, counterparts: filing, submitting and exchanging of arbitral documents can be done
'Chapter 8: When the online, while some smaller arbitration bodies have taken a step further by holding
Tribunal Is an Algorithm: arbitrations exclusively online. Despite the digitization of their processes, one thing
Complexities of Enforcing remains consistent – the arbitrators are human. However, in recent years, much
Orders Determined by a development has been made in training algorithms to decide an ongoing case, much like
Software under the New a robot arbitrator. This chapter intends to challenge age-old definitions and
York Convention', in Katia understandings that are critical in determining whether algorithmic arbitrators or
Fach Gomez and Ana M. algorithmic arbitration awards can be validly recognized under the New York Convention.
Lopez-Rodriguez (eds), 60 Considering the wider application of Artificial Intelligence (AI) or autonomous algorithms
Years of the New York in our daily life, and the numerous debates in discussing whether AI software or software
Convention: Key Issues and with human-decision capabilities should be deemed as a separate legal entity in itself,
Future Challenges, this chapter intends to discuss the public policy rationales in recognizing algorithmic
(© Kluwer Law arbitrators and its potential implication in the arbitration field if they indeed become
International; Kluwer Law recognized as proper arbitrators.
International 2019), pp. 121 P 121
- 134 P 122

§8.01 INTRODUCTION
Integrating aspects of traditional arbitration with computers or online applications has
been a growing trend with the continued development of technology in the legal industry.
(1) Several processes of traditional arbitration are being replaced by their digital
counterparts: filing, submitting and exchanging of arbitral documents can be done
online, (2) while some smaller arbitration bodies (or companies) have taken a step
further by holding arbitrations exclusively online. (3) Despite the digitization of these
processes, one thing remains consistent that most arbitration decisions are still made by
humans. (4) However, in recent years, there has been much discussion on the use of
machine learning in judicial decision-making by the likes of a robot arbitrator, (5) or in
this chapter, referred to as an ‘algorithmic arbitrator’. When an algorithmic arbitrator is
used, the understanding between parties then is that they should accept the decision
generated by the algorithm and such decision is final and binding to parties using the
online dispute resolution system.
This chapter focuses on the enforceability of such decisions made by algorithmic
arbitrators under the New York Convention – are such decisions, even if they are
recognized as valid orders of the tribunals, enforceable under the New York Convention?
In enforcing an arbitration award made by algorithms under the New York Convention,
there are five key provisions in the New York Convention that this chapter intends to
discuss – namely issues arising from enforcement pursuant to Article I.(2), Article IV,
Article V.1(a), Article V.1(d) and Article V.2(b).

§8.02 THE ALGORITHMIC ARBITRATOR AND THE NEW YORK CONVENTION

There are presently numerous developments in the area of using an autonomous software
to render decisions as if such software were a judge. In the US, a software recently made
P 122 the news due to a judge’s controversial decision to rely on a report generated by an AI
P 123 software that would assess the accused’s likelihood of reconviction to send a man to
prison, (6) thereby leading to discussions on consequences arising from AI judges that
mete out sentences. Across the Atlantic, researchers at UCL, the University of Sheffield
and the University of Pennsylvania have developed a machine learning algorithm that

1
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 can predict judicial decisions of the European Court of Human Rights to up to 79%
accuracy. (7)
As further inroads are made in the area of AI judicial making, there are similar
developments occurring outside the judiciary itself, although in different ways. The use of
online dispute resolution to resolve small value claims by a software is rather
commonplace in some large companies where there is a high volume of small value
disputes. (8) Even though most permanent arbitral bodies have yet to deploy an AI
arbitrator or algorithmic arbitrator, issues on the use of an AI arbitrator have already
been raised. (9) The New York Convention, which celebrates its 60th Anniversary this 2018,
however, could not have contemplated the rise of algorithmic arbitrators when it was first
drafted and adopted – the following sections are thus devoted to exploring how arbitral
awards made by algorithmic arbitrators will interact with the provisions of the New York
Convention.

§8.03 ARTICLE I.(2)

In Article I.(2) of the New York Convention, the definition of an ‘arbitral award’ includes
‘not only awards made by arbitrators appointed for each case but also those made by
permanent arbitral bodies to which the parties have submitted’. While the definition of
‘arbitrator’ is self-understood to be a human arbitrator, there is no explicit or implicit
restriction against algorithmic arbitrators in the wording of the New York Convention. The
ambit of who (or perhaps in this case ‘what’) can qualify as an arbitrator has generally
been broad – almost anyone can be an arbitrator. (10)
In this sense, if the definition of ‘arbitrator’ were to also encompass algorithmic
arbitrators, an arbitral award issued by such an algorithmic arbitrator would still be
theoretically permitted for recognition and enforcement under Article 1.(2). When
reviewing the New York Convention as a whole, the use of the term ‘arbitrator’ only
appears another time at Article V.1(b), whereby an arbitral award can be refused for
recognition and enforcement if the ‘party against whom the award is invoked was not
given proper notice of the appointment of the arbitrator’. (11) In this case, the use of an
P 123 algorithmic arbitrator does not affect the interpretation of this clause – as long as the
P 124
party so using an algorithmic arbitrator gives proper notice on the appointment of the
algorithmic arbitrator to the other party, this ground of refusal under Article V.1(b) is
moot.
Giving legal recognition and commercial acceptance to the appointment of an
algorithmic arbitration does not come without fundamental problems. An algorithmic
arbitrator (at least with current technology) will most likely only be able to provide a
decision based on specific algorithms that will most likely remain undisclosed to parties
as third-party software developers that created such an algorithmic arbitrator will be
unlikely to reveal the algorithm powering it, (12) and such an algorithm may be devoid
from actual intellectual reasoning or, even worse, show bias. (13) This can be unfair to
parties that believe that they have a unique point or issue that might otherwise be
ignored by the algorithmic arbitrator. If the use of an algorithmic arbitrator becomes
prevalent, it is suggested that permanent arbitral bodies take the lead by providing a
framework and procedural rules governing arbitral awards issued by algorithmic
arbitrators.

§8.04 ARTICLE IV
An aspect to consider before choosing an algorithmic arbitrator is whether the product of
its decision-making activity is enforceable according to Article IV of the New York
Convention. Indeed, the Convention imposes on the party applying for the recognition of
a foreign award to supply ‘(a) the duly authenticated original award or a duly certified
copy thereof; (b) the original agreement referred to in article II or a duly certified copy
thereof.’ (14)
The content of Article IV has been interpreted as the indication of the minimum formal
requirements that are asked to the party applying for the recognition and enforcement of
a foreign award (15) and prima facie right to ask for said recognition and enforcement.
(16) Indeed, the provision requires the applicant to supply evidence that an award has
been rendered and that the arbitral tribunal founded its jurisdiction on an arbitration
P 124 agreement. Article IV, nonetheless, seems to refer to written arbitration clauses or
P 125
agreement and written awards, (17) thus making it interesting to analyse the meaning of
these requirements in the context of AI tribunals and proceedings.

2
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 First, we should consider the meaning of the requirement of Article IV.1(a). Although the
New York Convention does not provide for a definition of what is intended with the terms
‘authenticated’ or ‘certified’, case law and commentators agree that authentication
means confirmation that the signatures of the arbitrators are authentic. (18) Certification,
instead, is the process by which a copy of a document is attested to be a true copy of the
original document. (19) Both processes require the involvement of an external body that
is capable of confirming the said characteristics.
The application of these characteristics to an algorithmic arbitration award presents
some issues. First, the algorithmic arbitrator might not be able to sign its final decision.
This is generally the case when the decision is made by a software. For example, one can
consider the case of blind-bidding software, commonly used in online consumer
disputes. (20) In this scenario, the decision is taken automatically by a computer, without
any specific formal requirements such as signatures. The final award, therefore, may lack
the necessary characteristic that is generally under scrutiny when the authenticity of a
document needs to be confirmed.
However, not recognizing that an award was made by the arbitrator could prevent the
decision from being enforced under the New York Convention and as a consequence
nullify all the activities carried out by the parties to that point. (21) In order to avoid this
negative result, technology should come to help. Indeed, a signature is the equivalent of
a mark or a stamp. Its main characteristic is the ability to link the signature, i.e., that
mark or stamp, with a determined person. (22) Therefore, it is easy to imagine that
technology could be able to provide any software acting as an arbitrator with a
predetermined and unique mark or stamp that is uniquely and undisputedly linked to a
specific author.
Another issue is represented by the fact that, in algorithmic arbitration, it may not be
P 125 clear what constitutes the origin of the award. Indeed, in this kind of proceedings, the
P 126
final award might be sent by the algorithmic arbitrator directly to the parties, for
example as an email attachment. As such, the attachment is the original award. The
situation may change if we consider what happens when the award is further saved by the
parties. Both parties, in fact, are able to store, save and reproduce that award an
indefinite number of times: by way of example, a party may decide to save the award in
his personal computer and, as a backup copy, in a separate hard drive. Furthermore, the
award may be sent via email to a third party: the document is now stored also on the
email providers’ servers of both the sender and the recipient. As it is now clear, this
process multiplies the copies of the award but, contrary to what happens with copies of
paper awards, all the documents appear the same – even if they are not the same
document – thus making it harder to determine which one constitutes the original and
which one constitutes the copy, that, therefore, cannot be filed without certification.
The distinction is not one without relevance. Indeed, national courts have been variously
interpreting the provision and, although the lack of a duly certified copy may not cause,
tout court, the rejection of the application for the recognition of the foreign award –
generally resulting in the request to supply further documentation, in spite of the
temporal requirement that Article IV also imposes – the impact on the proceedings in
terms of time and, consequently, cost is evident. (23)
Even in this regard, technology advancements should provide, in the future, the
substantial elements that the algorithmic arbitration award lacks. For instance, an
electronic document could be identified as the original with the use of a system similar
to a serial number or based on the blockchain technology; these types of solutions may,
on one hand, limit the multiplication of originals and, on the other, provide clarity on the
situations that entail the certification of a copy by an external body.
In order to complete the discussion over the recognition of an algorithmic arbitration
awards for the purposes of recognition and enforcement under the New York Convention,
the focus will now shift to Article IV.1(b). This article requires the party to submit, at the
time of the application, the original or a certified copy of the arbitration agreement,
allowing courts to have a view on the arbitration procedure. This requirement, indeed,
has opened the floor for the courts to analyse the elements of a valid arbitration
agreement under Article II of the Convention, the extent of which has been variously
interpreted. (24) It is not within the scope of this chapter to define algorithmic
P 126 arbitration agreements nor to investigate the applicability of Article II to the
P 127
algorithmic arbitrator. However, quick reference will be made to possible scenarios, in
order to identify additional issues that algorithmic arbitration may face and encourage
further consideration and analysis on the current and future challenges of the

3
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 development of this technology.
Article II requires the arbitration agreement to be in writing. Some courts have been
flexible in interpreting the requirement, for example by imposing a prima facie standard
or considering additional elements to the agreement, such as participation in the
negotiation, to invoke its existence and validity. (25) However, when imagining a full AI
arbitration procedure that involves algorithms and software in the entire process (i.e.,
from the conclusion of an arbitration agreement to the appointment of an arbitrator, to
the submission of documents, to the evaluation of the claims and the rendering of an
award) issues may arise as to how the agreement was concluded and whether it
constitutes an agreement in writing.
Further issues may arise in the analysis of Article II.2 with regards to its applicability to
the AI arbitration procedure described above. The provision, in fact, allows the
arbitration agreement to be concluded also with an exchange of letters or telegrams.
Given the evolution of modern communication (as indicated further in the following
paragraphs), however, the exchange of letters has been extended and interpreted as an
exchange of communication in general, (26) thus implying that the said communications
can apply to email and, possibly, in the next future, to an online environment such as an
AI arbitration platform. Indeed, the platform could provide a communication system
between the parties and between the parties and the algorithmic arbitrator (i.e., a
messaging system or an internal email box). The said exchange of communications should
be stored within the platform and also for the purposes of Article IV.1.b.
In addition, in the context of an AI arbitration process, multiple concerns may originate
as to the validity of the agreement and of the procedure itself that can be investigated
by the enforcing courts with the application of Article II.3. Both the algorithmic
agreement and the AI arbitration procedure indeed will need to be designed and
performed so as to comply with public policy and with formal and substantial
requirements of applicable laws, in order to avoid to undermine the applicability of the
AI procedure and therefore deny the recognition and enforcement of an award so made.
A related issue is the recognition of an electronic arbitral award, regardless of whether it
was issued by an algorithmic arbitrator or a human. The UN Convention on Electronic
P 127 Communications in International Contracts (27) was drafted with the view of removing
P 128
obstacles for the acceptance of electronic communications, agreements or arbitral
awards to be accepted in lieu of their written relatives. (28) The UN Convention on
Electronic Communications in International Contracts attempts to do so by providing the
rules on when an electronic record or electronic contract is functionally equivalent to a
contract that is required to be ‘in writing’. This is important to the New York Convention
as the New York Convention requires agreements to be ‘in writing’, pursuant to Article II.
(1), and such agreements need to be supplied for recognition and enforcement of an
arbitral award as per Article IV.(b). (29)
However, it has been highlighted by academics such as Reinmar Wolff that the UN
Convention on Electronic Communications in International Contracts does not allow
electronic arbitral awards to be recognized, despite its intention to do so. (30) The
International Chamber of Commerce also states that the use of a ‘fully enforceable, fully
electronic award, remain as barriers still to be conquered’. (31) Considering this debate,
electronic arbitral awards issued by algorithmic arbitrators should in theory not be
denied recognition on the sole ground that the arbitral award was issued in an electronic
form; however practice does not seem to have yet adjusted.

§8.05 ARTICLE V
Article V of the New York Convention lays down the possible grounds that parties can rely
on to refuse recognition and enforcement of the arbitral award. The following sections are
then dedicated to understanding whether the use of an algorithmic arbitrator to issue an
arbitral award will be a possible ground for refusal of recognition and enforcement of the
arbitral award under Article V.

[A] Article V.1(a)

According to Article V.1(a), a party resisting to the recognition and enforcement of the
award can oppose the proceedings if it can prove that ‘the said agreement is not valid
under the law to which the parties have subjected it or, failing any indication thereon,
P 128 under the law of the country where the award was made’. (32) The following paragraphs
P 129
will investigate this article and will consider how it can be applied in the context of
algorithmic arbitration.

4
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 There are two scenarios that the article proposes: according to the first one, the parties
may have chosen a law specifically governing the arbitration agreement. To be
considered valid, an arbitration agreement must fulfil the requirement of the said
applicable law that may be different. Article 1507 of the French Code of Civil Procedure,
(33) for instance, does not impose any formal requirements to the arbitration agreement;
section 5 of the English Arbitration Act, (34) instead, states that the agreement should be
in writing.
The algorithmic arbitrator should, therefore, follow the indications of the law chosen by
the parties in evaluating the requirements of Article V.1(a), even when a software is
involved in the conclusion of the said arbitration agreement. For the case of English law,
for example, the AI entity involved in the provision of the arbitration process must be
able to analyse and evaluate, according to precedents and case law, written evidence of
the agreement provided by the parties. If the agreement is instead governed by French
law, any manifestation of the intention to devolve the dispute to an arbitral tribunal,
however made, could be a valid arbitration agreement.
Imagining a scenario where AI is involved in the whole arbitral process, as it could be in
an AI arbitral body, the procedure put forward by the AI arbitral body could provide for
an initial stage assigned to the conclusion of an arbitration agreement itself. Parties
coming to the algorithmic arbitrator to solve their dispute, therefore, could find a
preliminary process in which they are presented with an arbitration agreement drafted
by the AI arbitral body according to the requirements of the applicable law chosen by
the parties, and that can be then signed and used as the basis for the subsequent
algorithmic arbitration.
A different situation opened to the parties when no specific choice of law was made with
relation to the arbitration agreement. In these circumstances, Article V.1(a) requires the
courts to consider the law of the place where the award was made. This provision has
been variously interpreted as to indicate the law of the seat of the arbitration, (35) the
mandatory laws of the place where recognition and enforcement was sought, (36) the law
where the arbitral award was signed, (37) the law chosen by the parties to govern their
substantive relationship or that with which the arbitration agreement has the closest and
most real connection. (38)
P 129
P 130
Although the arbitral tribunals have been involved in numerous decisions to determine
the meaning of such provision, many issues may arise in the scenario of an algorithmic
arbitration. Indeed, in the failure of the indication of a specific law governing the
arbitration agreement, most of the reasoning in the decisions above mentioned resort to
considering aspects of the arbitral procedure itself or the arbitral award.
When an algorithmic arbitrator is involved in the decision of a dispute, identifying the
place where the algorithmic arbitral award was made may not be straightforward, when
considering that the parties may access the algorithmic arbitration procedure from
different countries, the algorithmic arbitrator itself may be stored and powered by
numerous computers located in different countries and additional issues arise as to the
place where the algorithmic arbitral award was signed. While the scope of this chapter is
to highlight the challenges of the involvement of AI in the context of arbitration, further
analysis is required to address the said issues.
In all the aforementioned circumstances, in determining whether or not an arbitration
agreement constitutes a valid arbitration agreement, and specifically if an algorithmic
arbitration agreement could aspire to be considered a valid one, the public policy
argument comes forward. Its application will be investigated in the following paragraphs.

[B] Article V.1(d)

Part (d) of Article V.1 provides parties with the ground for refusing recognition and
enforcement if ‘the composition of the arbitral authority or the arbitral procedure was
not in accordance with the agreement of the parties, or, failing such agreement, was not
in accordance with the law of the country where the arbitration took place’. When a party
decides to use an algorithmic arbitrator, there are several possible situations that can
arise. First, this can occur when both parties expressly agree to use algorithmic
arbitrators. Here, there is no dispute that an algorithmic arbitrator can be used; however,
there may be complications if parties elect a specific permanent arbitral body to
administer the arbitration and such permanent arbitral body has no rules in
administering such an arbitration. In any case, for the purposes for Article V.1(d), this

5
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 means that there should be no grounds for refusal where parties contractually agree for
an algorithmic arbitrator.
The difficulty arises when one party wishes to have an algorithmic arbitrator, and the
agreement is silent on whether this is permitted. Considering that the use of algorithmic
arbitrators is not widespread, and at the present moment, still rather controversial due
to the inability of an algorithmic arbitrator to provide the rationale for its decisions other
than based on precedents, it would not be surprising that, if contested, a decision against
the use of an algorithmic arbitrator would be rendered unless parties contractually
agreed for an algorithmic arbitrator. This, in turn, leads to the non-enforcement and
P 130 recognition of the arbitral award pursuant to Article V.1(d). In light of such
P 131
circumstances, it is suggested that either parties contractually permit it or if parties are
uncomfortable in solely using algorithmic arbitrators, a two-tiered arbitration clause
permitting the use of algorithmic arbitrators depending on the claim amount can be
drafted. Permanent arbitral bodies can consider preparing such model clauses for future
adoption and use by commercial parties if the use of algorithmic arbitrators becomes
more popular in the future.

[C] Article V.2(b)

If a party intends to use Article V.2(b) to invalidate an arbitral award by an algorithmic
arbitrator, such party must prove that the use of an algorithmic arbitrator would be
against the public policy of that country. Considering the current interest and
development in the use of AI to make decisions, it is unlikely that a country would ban
the use of AI judges; however, there are a few countries such as France that effectively do
so by limiting the definition of an arbitrator to that of a natural person. (39) Certain
legislation or arbitration procedural rules also make reference to arbitrators using
gender pronouns such as ‘he’ or ‘she’, (40) although this can be said to be a question of
semantics rather than a purposeful attempt to restrict arbitration to only human
arbitrators.
For countries that do not have such an outright restriction, AI arbitrators appear to be
theoretically permitted, although it would be understandable if a country decides to
restrict the kind of cases that AI judges can make or introduces additional checks on the
decisions made by AI judges. However, in the case of an AI arbitrator or an algorithmic
arbitrator, it is difficult to extend the argument that this would be contrary to public
policy, unless explicitly expressed so, as parties did contract to pursue arbitration, let
alone contracting to pursue arbitration and agree to a decision made by an algorithmic
arbitrator.
Yet, the use of an AI system for judicial decision-making, be it on behalf of a judge or an
arbitrator, comes with several concerns. Key concerns include ethical concerns, such as
whether the algorithmic arbitrator will make decisions fairly and ethically, (41) which
may be a possible ground for not enforcing the judgment as an unethical and unfair
decision is arguably against public policy. Furthermore, it has been suggested that the
use of an algorithmic arbitrator ‘should be set aside for defying the international public
order, as it lacks key human characteristics such as human emotion, empathy and the
ability to explain its decision’. (42) Yet, this remains to be seen whether such an
argument would be sufficient under the public policy ground of Article V.2(b).
P 131
P 132

§8.06 OTHER ISSUES ON ARTIFICIAL INTELLIGENCE AND ARBITRATION

The difficulty in recognizing algorithmic or artificial intelligence powered arbitrators lies
in defining who or what can serve as an arbitrator, either solely or in a tribunal. If
legislation permits for the use of an algorithmic arbitrator, how such permission should
be granted is an important consideration. First, should governments recognize the
company that has developed the algorithm or the algorithm itself? At present, most
countries recognize companies as separate legal entities but do not recognize individual
algorithms as separate legal entities. If the former approach is taken, then, governments
face the issue of potentially opening the floodgates to allow companies to become
arbitrators, which is undesirable. (43) One solution would be to use an ex ante approach
to define and limit which software companies are entitled to provide algorithmic
arbitrator solutions, although this will impact competition and development of this field
and also runs counter to the general spirit in International e-commerce laws where an ex
post approach is preferred. (44)
However, if the alternative method is taken, i.e., by recognizing the algorithm,

6
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 governments may be according a separate legal personality to such algorithms, if not in
form but at least in substance. This poses tricky questions as governments will have to
consider whether they should formulate a specific policy for algorithmic arbitrators (and
by extension AI judicial decision-making) or consider an umbrella approach in
formulating a general AI policy that would cover other industries, such as the automobile
industry where autonomous vehicles have already been used on the roads. (45) The
recognition and adoption of algorithmic arbitrators should thus be considered in totality
and in conjunction with other AI policies that the government may want to roll out in
P 132 other industries; (46) a quick fix by recognizing certain providers or recognizing
P 133
specific algorithms for the purposes of arbitrators may have a snowball effect on how AI
laws and policies will be considered in other sectors. (47)
In the same vein, another issue would be whether the use of algorithmic arbitrators
should be classified under online dispute resolution. The development and use of
algorithmic arbitrators can arguably be a new form of online dispute resolution. Online
dispute resolution, if understood broadly, refers to the use of technology to facilitate
parties in resolving their disputes, (48) where such technology can either speed up the
process of existing dispute resolution measures such as arbitration, mediation or court
litigation or replace physical dispute resolution altogether, such as the use of online
arbitration for small claims or the UK’s online court for small claims. (49) In this sense,
the use of algorithmic arbitrators can be considered as a type of online dispute
resolution, although such taxonomic issues are best discussed substantively in a separate
article.

§8.07 THE FUTURE

The developments in the field of AI judicial making or AI decision-making by algorithmic
arbitrators should not be ignored. Even if there is no expressed restriction in the use of
an algorithmic arbitrator or nothing in the New York Convention that would outright
prevent the use of an algorithmic arbitrator, it is suggested that laws or procedural
amendments for the use of such technologies in arbitral proceedings should be made.
While the use of such algorithmic arbitrators may be seen as disrupting traditional
arbitration and competing with human arbitrators, the commercial future, however,
appears to be headed that way, and arbitral bodies should note these developments.
For a start, some clarity on whether an algorithmic arbitrator is acceptable under the
New York Convention would be appreciated, even if the New York Convention does not
expressly limit the use of algorithmic arbitrators. Next, permanent arbitration bodies can
assist in the adoption of algorithmic arbitrators by providing model arbitration clauses –
for instance, these clauses can allow for a two-tiered arbitration depending on the
amount of claim; that is, parties may be more comfortable allowing for algorithmic
arbitrators to arbitrate smaller claims and leave bigger claims to be resolved by a
human arbitrator. If such suggestions are to be implemented to facilitate the use of
algorithmic arbitrators, then care should be taken to ensure that parties, who rely on
third-party software to supply an algorithmic arbitrator, may have to relinquish their
P 133 right to have the said third-party software algorithmic arbitrator provide the machine
P 134
learning algorithm, if parties so chose to question the reasoning behind the algorithmic
arbitrator’s decision.
If governments and arbitration centres decide to integrate the use of AI arbitrators in the
decision-making process, one alternative way of integrating them is to have a human
checking the decisions of the AI arbitrator once the AI arbitrator has rendered a decision.
This may reduce the time and effort needed for the human arbitrator to draft an arbitral
award. In this case, the AI arbitrator is replacing certain functions of the human arbitrator
and not the functions of the human arbitrator entirely. However, considering the current
problems surrounding the ‘black box’ of AI, whereby software developers themselves are
unable to explain why the AI made a certain decision, (50) and that the AI arbitrator may
have to draft in natural language to enable human arbitrators without prior coding
background to understand, such an implementation may require some time to develop.
(51)
While there is still some way to go before algorithmic arbitrators fully replace human
arbitrators, (52) or that parties become comfortable with arbitrating large claims with
algorithmic arbitrators, it is never too late to ponder how the future of arbitration would
be, and to start providing the necessary legal framework to embrace such a future.
P 134

7
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 References
1) See generally, Maud Piers & Christian Aschauer, Arbitration in the Digital Age: The
Brave New World of Arbitration (CUP 2018). Also, for instance, the use of technology in
arbitral hearing rooms has been discussed: see Whitley Tiller & Timothy Foden, The
Effective Use of Technology in the Arbitral Hearing Room,
[Link]
technology-in-the-arbitral-hearing-room (accessed 25 August 2018).
2) In the Netherlands, the Stichting Transport and Maritime Arbitration Rotterdam-
Amsterdam offers TAMARA E-arbitration, which allows parties to share documents on
a secure part of their website. See also TAMAR E-Arbitration, E-Arbitration: Faster and
More Efficient Arbitration, [Link]
arbitration/ (accessed 25 August 2018).
3) See also FairClaims, Be Heard. Resolve. Move On., [Link]
(accessed 25 August 2018).
4) While there has been much talk about artificial intelligence and AI, none of the large
arbitration chambers have yet implemented such ‘robot’ arbitrators.
5) See also the Keynote Address by Sophie Nappat, Algocracy in Arbitration, which was
reported in Pratyush Panjwani’s post for Kluwer Arbitration Blog. See Pratyush
Panjwani, The Present and Near Future of New Technologies in Arbitration: If Not Us,
Who? If Not Now, When?,
[Link]
future-of-new-technologies-in-arbitration-if-not-us-who-if-not-now-when/ (accessed
25 August 2018).
6) Adam Liptak, Sent to Prison by a Software Program’s Secret Algorithms,
[Link] (accessed 25 August 2018).
7) UCL, AI Predicts Outcomes of Human Rights Trials, [Link]
articles/1016/241016-AI-predicts-outcomes-human-rights-trials (accessed 25 August
2018).
8) Tyler technologies, Modria, [Link] (accessed 25 August 2018).
9) José Maria de la Jara, Alejandra Infantes & Daniela Palma, Machine Arbitrator: Are We
Ready?, [Link]
arbitrator-are-we-ready/ (accessed 25 August 2018).
10) CIArb, Become an Arbitrator, [Link]
train-with-ciarb/become-an-arbitrator (accessed 25 August 2018).
11) Convention on the Recognition and Enforcement of Foreign Arbitral Awards (New
York, 1995), Article IV.1.
12) See Adam Liptak, supra n. 6. Also refer to Andrew Griffin, Robot Judges Could Soon Be
Helping With Court Cases, [Link]
tech/news/ai-judge-robot-european-court-of-human-rights-law-verdicts-artificial-
[Link] (accessed 25 August 2018). In Griffin’s article, reference is
made to an algorithm that does text analysis was used to predict cases at the
European Court of Human Rights, but this algorithm remains a secret.
13) Nanette Byrnes, Why We Should Expect Algorithm to Be Biased,
[Link]
be-biased/ (accessed 25 August 2018).
14) Convention on the Recognition and Enforcement of Foreign Arbitral Awards, supra n.
11 at Article IV.
15) Julian M.D. Lew & Loukas A. Mistelis & Stefan M. Kröll, Comparative International
Commercial Arbitration, 704 (Kluwer Law International, 2003).
16) See Yukos Oil Co. v. Dardana Ltd., [2002] EWCA Civ 543.
17) With reference to the requirements of most national laws, see Julian M.D. Lew, supra
n. 15 at 130-131.
18) Albert J. van den Berg, The New York Convention of 1958: An Overview, Enforcement of
Arbitration Agreements and International Arbitral Awards – The New York Convention
Practice, in Emmanuel Gaillard and Domenico Di Pietro (eds), 12-13 (Cameron, 39 May
2008).
19) Ibid. See also O Limited (Cyprus) v. M Corp. (United States) and others, Supreme Court,
Austria, 3 September 2008. Also Emmanuel Gaillard & John Savage, Fouchard Gaillard
Goldman on International Commercial Arbitration, para. 1675 (Kluwer Law
International, 1999).

8
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 20) Blind-bidding software is commonly used in monetary claims and allow parties to
confidentially submit offers or demands. The software confronts the indicated
amounts and, where possible, settles the dispute; Cybersettle is an online dispute
resolution mechanism that uses blind-bidding software. See Cybersettle,
[Link] (accessed 25 August 2018).
21) On the role of courts at the time of enforcement, see Gary B. Born, International
Commercial Arbitration, 3396-3405 (2d ed., Kluwer Law International, 2014).
22) See Directive 1999/93/EC of the European Parliament and of the Council of 13
December 1999 on a community framework for electronic signatures, Article 2.1. Also,
among the characters of an Advanced Electronic Signature, Regulation (EU) No.
910/2014 of the European Parliament and of the Council of 23 July 2014 on electronic
identification and trust services for electronic transactions in the internal market
and repealing Directive 1999/93/EC, Article 26.b, indicates the capability of the
signature to identify the signatory.
23) In Rainstorm Pictures Inc v. Lombard-Knight [2014] EWCA Civ 356 (27 March 2014), the
Court of Appeal stated that lack of certification on the copies of the arbitration
agreement provided to the court for the recognition and enforcement of the arbitral
award should not result in the plain refusal of its enforcement. In Medson Co. Ltd. v.
Viktor (Far East) Ltd. [2000], Hong Kong High Court, CLOUT case No. 459 (accessed 25
August 2018), the Court concluded that enforcement should be granted if the
applicant was able during the proceedings to remediate the procedural defects.
However, in an older decision, Jassica S.A. v. Ditta Polojaz, Court of Cassation, Italy, 12
February 1987, 1526, XVII Y.B. Com. Arb. 525 (1992), the Italian Supreme Court denied
the applicant recognition and enforcement of the award, based on the failure to
provide the documents required under Article IV of the New York Convention at the
time of application for said recognition.
24) See, among others, Emmanuel Gaillard & Yas Banifatemi, Prima Facie Review of
Existence, Validity of Arbitration Agreement, 225: 125, N.Y.L.J. (2005).
25) Aloe Vera of America, Inc (United States) v. Asianic Food (S) Pte Ltd. (Singapore) et al.,
Supreme Court of Singapore, High Court, Singapore, 10 May 2006, OS 762/2004, RA
327/2005, XXXII Y.B. Com. Arb. 489 (2007); Chloe Z Fishing Co. Inc., et al. v. Odyssey Re
(London) Ltd., formerly known as Sphere Drake Insurance, P.L.C., et al., District Court,
Southern District of California, United States of America, 26 April 2000, 109 F. Supp.
2d 1236 (2000).
26) See Nigel Blackaby, Constantine Partasides, et al., Redfern and Hunter on
International Arbitration, 75-79 (6th ed., OUP, 2015).
27) United Nations Convention on the Use of Electronic Communications in International
Contracts (New York, 2005).
28) UNCITRAL, United Nations Convention on the Electronic Communications in
International Contracts (New York, 2005),
[Link]
[Link] (accessed 25 August 2018).
29) Convention on the Recognition and Enforcement of Foreign Arbitral Awards, supra n.
11, Article IV.1(b).
30) Reinmar Wolff, The UN Convention on the Use of Electronic Communications in
International Contracts – An Overlooked Remedy for Outdated Form Provisions Under
the New York Convention?, in Katia Fach Gómez & Ana Mercedes López Rodríguez
(eds), 60 Years of the New York Convention: Key Issues and Future Challenges (Kluwer,
2019); Richard Hill, On-line Arbitration: Issues and Solutions, 15(2) Arb. Intl. 199-207.
31) ICC Commission, Information Technology in International Arbitration – Report of the
ICC Commission on Arbitration and ADR, [Link]
technology-international-arbitration-report-icc-commission-arbitration-adr/
(accessed 25 August 2018).
32) Reference is made to the agreement referred to in Article II as indicated in the first
part of Article V.1(a).
33) Code de procédure civile, Livre IV, Titre II, Article 1507.
34) Arbitration Act, section 5 (1996).
35) Most commentators agree that the seat of the arbitration is the legal place where an
arbitration takes place, according to the arbitration agreement, and despite where,
physically, hearings or meetings are held. See Fraiser P. Davidson, Where an Arbitral
Award Is Made? – Hiscox v. Outhwaite, ICLQ 637 (1992).

9
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
 36) In the famous Dalico Case [Comité populaire de la municipalité de Khoms El Mergheb v.
Dalico Contractors (1994) 121 J.D.I. 432 French Cour de Cassation], the court held that
the common intention of the parties to arbitrate a dispute determines the validity of
an arbitration agreement, provided that the mandatory laws of the country where
recognition and enforcement is sought and international public policy are complied
with.
37) See Hiscox v. Outhwaite (No. 1), [1992] HL 1 A.C. 562.
38) Sulamérica CIA Nacional de Seguros SA and others v. Enesa Engenharia SA and others
[2012] EWCA Civ 638.
39) French Arbitration Act.
40) UNCITRAL, UNCITRAL Arbitration Rules (with New Article 1, Paragraph 4, as Adopted in
2013), Article 11, [Link]
2013/[Link] (accessed 25 August 2018).
41) See Jan L. Jabobowitz & Justin Ortiz, Happy Birthday Siri! Dialing in Legal Ethics for
Artificial Intelligence, Smart Phones, and Real Time Lawyers, Texas A&M U. J. Prop. L.
(2018).
42) José Maria de la Jara, supra n. 9.
43) Arbitral rules such as the UNCITRAL Arbitration Rules and the ICC Arbitral Rules do
not stipulate whether a company may or may not be an arbitrator, although they do
require arbitrators to be impartial. See also UNCITRAL Arbitration Rules, Articles 11-
13; ICC Arbitration Rules, Article 14.
44) This is in line with the prevailing view of UNCITRAL’s E-Commerce texts, i.e., the
Model Law on Electronic Commerce (1996), the Model Law on Electronic Signatures
(2001), the UN Convention on the Use of Electronic Communications in International
Contracts (2005) and the Model Law on Electronic Transferable Records (2017) where
the UNCITRAL’s E-commerce texts adopted the approach to distil the key elements of
the specific electronic record and accord functional equivalence of the digital copy
to the physical copy. This is contrary to procedures where countries decide that a
certain electronic contract is valid only when it is issued by a specific committee or a
specific law firm. See also, the most recent e-commerce text: UNCITRAL, UNCITRAL
Model Law on Electronic Transferable Records (2017),
[Link]
[Link] (accessed 25 August 2018).
45) See generally, Mark A. Geistfeld, A Roadmap for Autonomous Vehicles: State Tort
Liability, Automobile Insurance, and Federal Safety Regulation, 105 Calif. L. Rev. 1611
(2017).
46) For instance, in Singapore, the Singapore Government has awarded a significant
grant to the Singapore Management University’s Faculty of Law to address issues
relating to AI and its implication to society. See also, Singapore Management
University, SMU School of Law Awarded Significant Research Grant to Address
Governance of AI and Data Use in Support of the Development of Singapore’s Digital
Economy, [Link]
significant-research-grant-address-governance-ai-and-data-use (accessed 25 August
2018).
47) See generally, Nicolas Zahn, Why Intelligent Machines Also Need Smart Policies,
[Link]
innovation/blog/page/[Link] (accessed
25 August 2018).
48) Leah Wing & Daniel Rainey, Chapter 2: Online Dispute Resolution and the Development
of Theory, in Mohamed S. Abdel Wahab, Ethan Katsh & Daniel Rainey (eds) Online
Dispute Resolution: Theory and Practice (Eleven International Publishing, 2012).
49) [Link], Make a Money Claim, [Link] (accessed 25
August 2018).
50) Will Knight, The Dark Secret at the Heart of AI,
[Link]
(accessed 25 August 2018).
51) Other industries such as the Chinese e-commerce product copywriting industry has
made use of AI, see Zen Soo, Alibaba’s AI Copywriter Can Produce Up to 20,000 Lines of
Copy a Second, Passes Turing Test, [Link]
tech/article/2153698/alibabas-ai-copywriter-can-produce-20000-lines-copy-
second-passes (accessed 25 August 2018).
52) See Christine Sim, Will Artificial Intelligence Take over Arbitration?, 14 Asian Int’l Arb. J.
1 (2018).

10
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.

Kluwer Arbitration is made available for personal use only. All content is protected by copyright and other intellectual property
laws. No part of this service or the information contained herein may be reproduced or transmitted in any form or by any means, or
used for advertising or promotional purposes, general distribution, creating new collective works, or for resale, without prior
written permission of the publisher.

If you would like to know more about this service, visit [Link] or contact our Sales staff at lrs-
sales@[Link] or call +31 (0)172 64 1562.

KluwerArbitration

11
© 2024 Kluwer Law International, a Wolters Kluwer Company. All rights reserved.