Back to Arbitration Committee publications

Sadaff Habib
Beale & Company Middle East, Dubai
S.Habib@beale-law.com

In 2017, Price Waterhouse Coopers (‘PWC‘) published a report on the value of artificial intelligence (‘AI’) on different businesses. They anticipate that by 2030, AI could contribute up to $15.7 trillion to the global economy. This is more than the current output of China and India combined. Of this, $6.6 trillion is likely to come from increased productivity and $9.1 trillion is likely to come from consumption-side effects. The PWC report concluded that developing countries will experience a modest increase in GDP because the rate at which AI technology is adapted will be lower. In fact, the total impact on the GDP of Africa, Oceania and other Asian markets is a humbling 5.6 per cent.    

As a continent, resource-rich Africa is lagging behind in development of AI because of the lack of trained and qualified AI professionals. To get a commendable and decent education in Information Technology, candidates often travel abroad and seldom return home. As African countries continue to suffer this exodus of talent, the AI industry remains largely under-developed. The issue remains – Africa must harness and/or retain its talent to take advantage of AI.

AI and arbitration

AI refers to the simulation of human intelligence in machines that are programmed to think like humans and mimic their actions. The term may also be applied to any machine that exhibits traits associated with a human mind, such as learning and problem-solving. At first glance, it is difficult to see how AI may correlate with arbitration. However, AI may be a tool that lawyers could use to transform their legal services. Arbitrations are known to be document intensive and often require paralegals to review, filter and compile relevant documents to support the case. This step could be undertaken by AI where the AI system could filter out relevant documents and arguably reduce billable hours chargeable to a client.

But can a robot simulating human intelligence make a decision? What about emotional intelligence in arbitration? Most arbitration institutional rules are silent on AI. At most, they allow parties and the arbitral tribunal to agree to have virtual hearings and case management conferences. However, taking a literal look at the wording of most institutional rules, they do not define an arbitrator as human ­– is there room to argue that an arbitrator could be a machine?

The case for AI

In 2018, academics at Oxford University evaluated the possibility of job automation based on nine skills that a robot would be required to perform: social perceptiveness, negotiation, persuasion, assisting and caring for other, originality, fine arts, finger dexterity, manual dexterity and the need to work in a cramped work space. Therefore, it is predicted that humanoids will have the capacity to undertake jobs ranging from clerk to arbitrator.

There are many benefits to incorporating AI in arbitration. The use of machines could speed up contract review and establish the most suitable arbitration agreement; assist with appointing arbitrators; assist with scrutinising arbitral awards; transcription; assist with sifting through thousands of pages of documents to highlight those relevant to the proceedings. AI has assisted in resolving disputes faster and allowing the parties to settle. For example,Ebay’s on-line processes resolve 60m disputes a year, and China has ‘No Court Room’ internet courts. However, these benefits come at a cost – that is the cost of using the machines. Parties would need to assess the time and cost factors in using AI in an arbitration.

An appetite for AI in arbitration in Africa

Recent years have seen a number of promising developments in Africa. These developments can be traced back to the Organization for the Harmonization of Business Law in Africa (‘OHADA’) reforms. In late 2017, OHADA (a group of 17 African countries geared towards unifying business laws to boost investment) enacted the Uniform Mediation Act, the Uniform Arbitration Act (‘UAA‘) and Arbitration Rules for the Court of Justice and Arbitration (‘CCJA‘), all of which were introduced with the intention of attracting investors and boosting confidence in OHADA seated arbitrations.

For AI to take a formative presence in arbitration it needs to be crystallised into the arbitration system to avoid and reduce uncertainties. Article 5 of the UAA requires that the duties of an arbitrator may only be performed by a natural person. Therefore, even if the parties to an arbitration agreement were to agree to refer their disputes to an AI system and were subject to the UAA, such an arbitration agreement could be unenforceable. On the other hand, and rather interestingly, the Lagos State Arbitration Law 2009 (the ’Act‘), allows parties to freely agree on how their dispute is resolved, subject to those safeguards necessary to public interest. The Act does not expressly require an arbitrator to be a natural person. This begs the question, if parties with the seat of arbitration in Lagos agreed to resolve their arbitration through an AI system, would such a referral be recognised under the Act or would it be against public interest? The Act is largely modelled on the UNCITRAL Model Law. Notably, the UNCITRAL Model Law is widely adopted by many African countries and across the globe with States often making certain amendments to fit the law with their overall framework.

The UNCITRAL Model Law defines an arbitral tribunal as consisting of a sole arbitrator or a three- member tribunal. It does not specify that the arbitral tribunal should be a natural person. Does this give scope to an AI arbitral tribunal? Is this enough to ignore the reference to person under the UNCITRAL Model Law when discussing challenging and appointing an arbitral tribunal, or is there scope to argue that person can and does include a natural and artificial person? If the latter interpretation is adopted, the Model Law and the Act would be thought-provoking and forward-thinking legal instruments, albeit might not have been within the contemplation of the drafters at the time.  

In the wake of the Belt and Road Initiative, Chinese and African stakeholders agreed to establish the China-Africa Joint Arbitration Centre (‘CAJAC‘) in Johannesburg and in Shanghai to which trade and investment disputes arising between nationals of China and Africa could refer their disputes. The CAJAC Johannesburg Rules define a Tribunal similarly to the UNCITRAL Model Law, that is consisting of either a sole arbitrator or more arbitrators. Whilst the Rules do not make mention of the arbitrator being a natural person, Article 15.1 addresses replacement of an arbitrator ‘ […] if an arbitrator dies or becomes unable to perform his or her functions due to any reason beyond his control…’ It is unlikely that a machine can be assigned gender or would die. Therefore, the CAJAC Johannesburg Rules may be seen to refer to a natural person. However, Article 2.2 of the Rules allow parties to vary the Rules by agreement in writing. Arguably, a party can have its AI arbitrator after all, subject of course to any mandatory laws applicable.

More recently, the Arbitration Foundation of South Africa (‘AFSA‘) launched its new international arbitration rules for public comment. The rules do not define an arbitral tribunal to be a natural person but Article 6 (5) refers to gender of the arbitral tribunal when being notified of a potential challenge. This could mean that the new arbitration rules do not leave room for an AI system as arbitrator.

Arguably, whilst there may be scope for parties to agree to AI as an arbitral tribunal, such scope may be limited under mandatory laws or due to public policy. Ultimately, whether or not AI is used, a party’s goal is to obtain an enforceable arbitral award. A key instrument used in enforcing arbitral awards is the 1958 Convention on the Recognition and Enforcement of Foreign Arbitral Award (the ’Convention‘).

The ever-gnawing issue of public policy

Out of the 54 countries that make up the African continent, 33 have ratified the Convention. Ethiopia is the most recent African country to accede to the Convention and did so in early 2020. A country that is signatory to the Convention is attractive to foreign investors and investment because of the perceived ease in enforcing a successful arbitration award between signatories to the Convention.

A signatory to the Convention should implement it through its national legislation. This has typically been achieved amongst African countries through (i) adopting the Model Law, (ii) the UAA and (iii) domestic legislation.

Article V (2) (b) of the Convention allows the State where enforcement of the award is sought to refuse enforcement where it contravenes its public policy. Public policy is often described as an unruly horse with its scope being vague and difficult to define. Public policy usually consists of principles that reflect the needs and values upheld by a society at a given time. A difficulty that often arises with countries adopting the Convention is the different interpretations of the Convention’s public policy. Some countries, for example Rwanda and Mozambique, may give it a wider interpretation while others such as Zimbabwe may give it a narrow interpretation. Also since the Convention does not make express reference to AI (arguably nor does it exclude it) a party could argue that the use of AI itself is against public policy. Of course, it would be difficult for a party to raise this where it agreed to the AI process to begin with.

Caution should also be heeded where a State has implemented the Convention through the UAA. As discussed above, the UAA expressly requires the arbitral tribunal to be a natural person. There is no room to agree an alternative AI system. A general point to note, only 11 of the 17 member states are signatory to both the UAA and Convention. So as a matter of enforcement, parties should consider which States are signatory to both to benefit from the ease of enforcement.

Some States which are signatories to the Convention still have in place old archaic arbitration laws that may not fit in with the provisions of the Convention. For example, Botswana and Namibia have arbitration laws based on the UK arbitration laws dating back to the 1950’s. These would most likely pose further hurdles to a party seeking to enforce an AI award.

Conclusion

Every case is unique on its own facts. How can you expect a machine to decide on a jurisdictional challenge that turns on capacity? Or to measure up and evaluate a security for costs application? Is artificial intelligence capable of self-learning? Can AI help avoid cognitive bias? These are just some of the many questions that arise when considering AI as an arbitral tribunal.

The ideal scenario may be for AI and humans to work hand in hand in an arbitration. There are unarguable advantages to using AI, particularly in time-consuming and costly tasks such as document disclosure or dealing with large volumes of expert evidence. With these tasks being machine handled, lawyers and arbitrators can better focus on the more human aspects of the case, that is the client’s needs and effective dispute resolution. However, whether such a balance can be achieved remains to be seen.

Regardless, to keep up its momentum in international arbitration, what Africa as a continent needs to do is to invest a certain portion of its resources in developing IT programs to retain talent to be ready if AI kicks off in arbitration. Systems such as DRExM was recently used in Egypt to resolve construction disputes. The system is able to recommend the most suitable dispute resolution technique, depending on the nature of the dispute, the evidence and the relation between the parties.

Should machines take over arbitration entirely? I remain sceptical.

Back to Arbitration Committee publications