Artificial intelligence, in its brief period of integration of just over a decade, has already had a massive impact on the traditional approach of doing business. This has revolutionised the legal sector as well, particularly international commercial arbitration, where the constant development in data management technologies and predictive analysis is changing the conventional approach to dispute resolution. With capabilities far beyond human comprehension in the realm of data analysis, predictive analysis, and automated research, it begs the question of whether AI can, in fact, replace a human arbitrator. To answer that question, we need to analyse the current AI tools, their integration in the dispute resolution process, and the future of AI and its development trajectory.
What AI integration boils down to is the tools currently used by legal professionals. This includes ROSS Intelligence, Kira Systems, and Lex Machina, which have changed the spectrum of legal research and data analysis, having provided professionals with access to even newer tools. Tools that can predict the outcome of cases based on the documentary evidence submitted by the parties. In arbitration, AI has begun assisting with e-discovery, document drafting, transcription, language translation, and even arbitrator selection based on past decisions. This has resulted in a twofold benefit.
First. It has reduced costs by minimising the working hours billed by professionals for labour-intensive tasks such as legal research, drafting, documentation, etc.;
Second. As an aftereffect of the first one, saves everyone involved in the process time and energy.
However, only looking at the upside that AI brings to dispute resolution would result in a false narrative. While the capability is undeniable, it has its own shortcomings that need to be addressed. Moreover, even if an AI system is found capable of replicating it, it begs the question of whether such an award would be enforceable under legislation like the New York Convention (1958)
After the pandemic in 2020, major international arbitral institutions like the London Court of International Arbitration, Singapore International Arbitration Centre and International Chamber of Commerce brought about major changes in their operations to accommodate the need for digital proceedings. Nonetheless, even after the pandemic, the use of AI tools, virtual hearings, and digital submissions has become common practice due to their inherent benefit in convenience and brevity for both the parties and the institutions. Some of these modernised practices include:
First. E-discovery software, these types of software can process millions of documents, provide well-drafted documents, and even help attorneys prepare their arguments.;
Second. ArbiLex: ArbiLex is at the forefront of predictive analysis that can analyse the facts of any particular case and predict the likely outcome.; and
Third. Speech recognition and translation tech: The cross-cultural nature of international commercial arbitration inherently poses a linguistic barrier for parties. Modern software can transcribe and convert languages at near instantaneous speed, boosting the efficiency of the entire process.; and
Fourth. AI-powered arbitrator analytics: These kinds of tools help parties analyse past decisions of various tribunals and identify behavioural patterns of arbitrators to assist in the selection of a suitable arbitrator for their dispute.
That said, there is a constant challenge for these tools as they rely on openly accessible data. Since the process of international commercial arbitration is private in nature, the data regarding the selection of the tribunal, proceedings, etc., is not publicly accessible and hence renders these AI tools helpless. Nevertheless, the growing adoption of transparency in investment arbitration might not only pave the way forward for such AI tools but also encourage the development of more advanced technology in this field.
AI’s strength definitely lies in its ability to process large volumes of data and give precise analysis within seconds. In the legal field, especially in international commercial arbitration, where there are enormous amounts of legal data to process in terms of precedents of both countries, precedents of international courts, as well as legal contracts and treaties signed by parties and other documents used as evidence during the process, AI’s abilities are definitely useful.
However, with this strength, there is also a certain weakness of AI. While it can process data and possess enormous legal knowledge, it cannot apply that knowledge with legal reasoning, as legal reasoning is inherently the output of innate human qualities such as morality, empathy, and contextual understanding. It is not yet capable of understanding the nuances of an argument, reading between the lines, or perceiving the emotional and cultural context of a certain issue. The process of arbitration is not merely a factual analysis of evidence but an important means of delivering justice.
Empathy is one of the core principles of natural justice. When arriving at an award, arbitrators do consider cultural nuances and human context so as to make sure that the fundamental nature of justice delivery does not get blinded by minute factual intricacies. This ensures that the decisions arrived at are well-reasoned and understood. AI is not yet capable of replicating that.
Empathy alone cannot function without the existence of discretion. Unlike traditional courts, the process of arbitration allows flexibility. Thus, arbitrators have to consider the existing norms and expectations as well as other implied conditions existing at the time of the contract between the parties. UNCITRAL Model Law recognises this by encouraging party autonomy and flexible procedure. AI, which is driven by data input but lacks real-world human experience, fails to navigate this grey zone.
Lastly, when talking about moral judgment, AI cannot replicate it since AI systems are not trained to have a moral compass. Consequently, they are unable to deduce the morality of a situation, relying merely on bare facts. This is even more important in international arbitration, where party dynamics can vary a lot depending on the strength of the nation, the access to resources that a particular party has, and the position of parties in global politics. Hence, an arbitrator needs to look further than just merit.
Although AI is not yet capable of acting as an independent arbitrator, the future is unpredictable, and thus we need to look at the existing legal framework to see if, eventually, AI is able to overcome all the shortcomings, will our legal system allow the enforcement of such an award.
According to the New York Convention, an award needs to be “duly reasoned” and rendered by a “tribunal” agreed upon by the parties. An award by an AI arbitrator does not fit this definition. Article V(1)(d) allows refusal of enforcement if the procedure was not in line with the party agreement or applicable law. Since India follows the New York Convention, the Arbitration and Conciliation Act, 1996, lays down a similar principle. There are no provisions for an AI-rendered arbitral award. Even if the parties consent, it would not guarantee a hundred per cent chance of enforceability of such an award.
The CIArb’s guidelines on AI emphasise transparency and constant human oversight to make sure due process is followed. Until provisions for an AI-rendered award are crafted, the use of AI in arbitration will stay in the grey zone.
When talking about transparency, the biggest issue is the “black box” problem. This arises from the fact that the process of reaching an output by an AI program cannot be explained, even by the very developers who develop it. This is due to the machine learning nature of AI. Thus, it would make it impossible for humans to comprehend how an AI arbitrator reached an award, and such a lack of transparency would create distrust, leading to constant abuse of Article V(1)(d) of the New York Convention by the party against whom the award is passed.
This is because such an award would undermine a core principle of arbitration: reasoned awards. If an arbitrator cannot explain the reasoning behind an award, such an award cannot be considered legitimate. Parties have the right to know which arguments prevailed and what was the reasoning behind one party’s failure and the other's success. AI is incapable of doing this right now.
Even with its advanced abilities, currently, AI is not capable of acting as a sole arbitrator. However, this does not mean that AI should not play any role during the arbitration process. A hybrid approach, where the emotional intelligence of a human arbitrator and the data processing ability of an AI software, seems like the best way forward. Such an approach helps fill in the gaps that both parties have and together make for a stronger, more efficient, and accurate system of dispute resolution.
Advancements in Explainable AI and sentiment analysis show promise to one day be able to replicate the human instinct coupled with its unmatched data processing capabilities, but till then, it will best serve as an assistant to an arbitrator and not the arbitrator itself.