International Arbitration and Artificial Intelligence

Yash Sharma*

With the digitisation of the world, technological advances have significantly eased the way we perform tasks. Artificial Intelligence (hereafter as AI) is one of the most prominent advances of the 21st Century, which has eminently reduced the burden of labour from humans. Automation is not only present in the mainstream industries such as the automobile industry and information & technology industry but is disrupting other mainstream sectors, and the legal industry is not one to shy away from it. In his book in 1990, Richard Susskind was of the view that e-mails, e-filings would soon impact the legal industry[1]. Later in 2017, he claimed that the outlook of the legal sector would change significantly as disruption of technology will assist lawyers in breaking down tasks that would help firms to achieve maximum cost-efficiency.[2] Well, he is not wrong in making such claims as AI has revolutionised arbitration with various products. AI products are built on the cornerstone of four algorithms such as blockchain, big data, data mining and machine learning. This paper aims to assess AI in its capacity to assist parties/ arbitrators and its ability to replace arbitrators altogether.


AI assists by providing due diligence in assessing contractual matters, predictive technology in evaluating outcomes, legal analytics in finding relevant case laws/topics, document automation and intellectual property filings.[3] However, the present paper will only delve into the first three categories. Due diligence is the most crucial part of the proceedings for both the arbitrators and the parties as they might miss out on an exceptionally essential clauses/facts, which might potentially affect the trajectory of the case. ‘Kira’ is one such software that assists parties in due diligence of contracts and lease deeds. It scans the document and highlights the relevant portion, parallelly it runs another scan wherein the user can look at different clauses that are embedded in the software itself to strengthen the document. The Magic Circle itself relies on software such as Kira in their daily practice.[4]

‘ArbiLex’ is another program that provides dual services for International Arbitration. The first being that of an analytical nature, by looking at the relevant documents and highlighting important parts, the second feature is that of a predictive nature, where it predicts the winning percentage of claims[5]. Since International Arbitration involves high-value cross border disputes, such predictive tools give parties a chance to assess the probability of success and offer insights on whether they should pursue such claims. For e.g., Company A belongs to France, and Company B belongs to the U.S.A., there arises a dispute regarding the works contracts agreed between the parties. Now if Company A runs the check and finds a probability of 72% to win the case, most likely they will proceed forward with such huge claims, whereas if the test shows a win probability of 32% they want to settle the dispute amicably without pressing any hefty claims.

The Columbian government has introduced a ‘Siarelis’ robot which poses questions to the judges with respect to the claims and provides probable answers for the same with supporting judgements.[6] A tool of a similar nature can provide immense help to arbitrators in the long run. Another tool that provides tremendous support to in-house counsels in discovering relevant e-documents is predictive coding. A counsel who is well-versed with the facts reviews the reference set and codes it according to relevance. The computer, once the algorithm is applied copies the same decision throughout the remaining documents. This process significantly removes the irrelevant documents leaving the counsels only the assorted relevant document for inspection.[7] Ross Intelligence is another tool that eases the work when it comes to finding applicable case laws and other legal papers. The search engine is so advanced with nuanced algorithms that users can practically type their queries as questions and still receive a detailed analysis of it.[8]

Lastly, one of the most prominent problems that parties face is choosing an unbiased arbitrator, which takes up a significant amount of time. Such biases, in the view of the author, can be removed by utilising the Arbitration Intelligence Questionnaire, it which consists of two parts and must be filed by the parties and arbitrator after every arbitration. The questionnaire covers a wide range of questions which later builds a portfolio for the arbitrator highlighting the type of arbitrations he has been in. This answer is readily available for parties to look upon and decide the arbitrator. Notably, no specific legal framework of arbitration in itself provides rules to enforce artificial intelligence; however, Article 19 of the UNCITRAL (Model Law) states that the parties are free to follow procedures of their own, failing which, the arbitrator can demand his way to be followed for the expedient disposal of cases.[9] Therefore, there exists no bar on the employment of AI-centric tools, provided that a clause pertaining to the same finds its mention within the contractual agreement. On account of the convenience they bring along with them, AI instruments are gaining significant popularity in their employment as supportive tools to the arbitration process, but their usage in the same requires the fulfilment of certain legal mandates such as the aforementioned.


With such rapid changes, it is foreseen that AI can remove replace arbitrators and would be able to adjudicate matters on its own. Some jurists are of the view that with proper precision AI can be the perfect impartial and independent arbitrator, free of bias which human arbitrators possess. The process behind building such a complex system is machine learning. It can be referred to as a system that learns from its experience and are is capable of changing their behaviour to enhance their performance. The algorithm detects patterns in order to automate tasks or make predictions. Such algorithms need to be in consonance with the four V’s of Big Data such as Volume, Velocity, Variety and Veracity.

Although AI as an arbitrator makes for an attractive alternative, the same cannot be implemented in its present state. There exist multiple limitations on the functionality of AI-based on the Four V’s mentioned above. These limitations exist in the form of pre-requisites which are required to be fulfilled for the functioning of the machine learning software.

Volume refers to large amounts of data, which will be fed into the algorithm for the recognition of patterns. Via machine learning, these patterns will be studied and on the basis of these patterns, the software will then make informed decisions. Velocity refers to the speed and frequency of incoming data. In the legal world, this data is constantly changing. The software uses past data to study patterns and thus, this implies that the decisions to be made in the present and future are determined by data of the past. On account of the ever-changing nature of data in the legal context, such a system would be rendered ineffective and instead be called out for being conservative and incapable of keeping up with the constantly changing scenario. Variety refers to the repetition of certain patterns which have binary outcomes. Without repetition, the system would be incapable of identification and thereby fail in making predictions, let alone successful ones. Lastly, Veracity pertains to the accuracy and trustworthiness of data.[10] A plethora of studies has arrived at the common conclusion that humans are regular victims of irrationality being a part of their decision-making processes. Contrary to popular belief, the same can be said about machines as well. A computer program in the U.S. has reported being biased against black prisoners. The study showed that the program stated that black people were the ones to re-offend the same crime.[11] Google, in one of its algorithms, referred to gorillas as black people.[12]

Another important factor that plays an important role is who is to be held accountable if such systems default. Since these machines do not have are not legal entities enforcing liability on these machines will reap no benefits. Even if such a law is passed, there would be next to no justice serviced as the programmer will have the right to switch it off. In the same light, it would be tough to hold the programmer liable for the mistakes of the machine. Machines feed on data to come to conclusions, and the programmer would wave off liability on him by stating that the machine has its own algorithm to come to such conclusions. Another critical factor that such machines miss are emotions and empathy. Based on the foundation of principles of a sound justice system such emotions play a massive role in determining the facts of the case and lastly such systems are only equipped to provide answers and not reasons for the same as well. A lot of times, parties choose options for receiving the reason for the award; however, with the current state of development, such decisions will not be available to parties.


In the present scenario, with AI blooming in the legal sector, it does not provide for the means and resources for the machines to act as sole arbitrators. The Adjudicating matter is more than providing the winner of the suit; it is also about judicial activism and constant checking of the system. Insertion of AI into the field of law is no child’s play, due to the role which law plays in society. It establishes rules which govern all individuals, entities and their interactions. The pursuit of justice is not to be lost in modernising law and order. Thus, AI instruments require themselves to be extremely accurate and reliable before finding employment in the justice department. Significant changes to the current technology are to be made in order to make the inclusion of AI into law more trustable and viable. However, it is not to be understood that such a change will never see the light of the day. Moreover, in my opinion, AI should work in consonance with human beings rather than an individual itself independently. Algorithms could be developed where they provide extensive details about help in the drafting of the award rendered which when coupled with a human arbitrator will increase its effectiveness as the human arbitrator can understand the line of reasoning followed and can thus correct if the machine goes out of place to provide necessary inputs. With the upward shift of arbitration in India, much work is needed to inculcate AI in its system. Legal-tech firms with the Indian Council of Arbitration should formulate a plan which embarks India on a technological approach in the arbitration which in turn will make Indian a hot seat for International Arbitration.

*(3rd Year Law Student at O.P. Jindal Global University, Sonipat)

[1] Susskind Richard, Future of Law: Facing the Challenges of Information Technology, Oxford University Press, 1996. [2] ‌Susskind Richard et al., “Richard Susskind on the Future of Law” (, August 13, 2017) < > accessed July 23 [3] Faggella D, “AI in Law and Legal Practice - A Comprehensive View of 35 Current Applications” (Emerj, March 13, 2020) < > accessed July 23, 2020 [4] Elizabeth Huang, “Tech Wizards: Your Guide to AI and the Magic Circle, Cambridge University Law Society (CULS)” (, 2017) <> accessed July 22, 2020 [5] “ArbiLex - Predictive Analytics for International Law” (, 2020) < > accessed July 22, 2020 [6]Artificial Intelligence in International Arbitration: From the Legal Prediction to the Awards Issued by Robots” (, February 18, 2019)< > accessed July 22, 2020 [7] Wheater Michael, “Predictive Coding: The Current Landscape” (Dispute Resolution blog, July 21, 2016) <> accessed July 22, 2020 [8] “ROSS Intelligence” (ROSS Intelligence, 2014 < > accessed July 22, 2020 [9] Subject to the provisions of this Law, the parties are free to agree on the procedure to be followed by the arbitral tribunal in conducting the proceedings.” [10] Scherer Maxi, 'Artificial Intelligence and Legal Decision-Making: The Wide Open?', (2019), 36, Journal of International Arbitration, Issue 5, pp. 539-573, < > accessed July 22, 2020 [11] Buryani Stephen, “Rise of racist robots- how AI is learning all our worst impulses” (, August 08, 2017) <> accessed July 22, 2020 [12] Griffin Andrew, “Google photos tags black people as ‘Gorillas’, puts pictures in special folder” (, July 01, 2015) <> accessed July 23, 2020

  • Grey Google+ Icon
  • Grey Twitter Icon
  • Grey LinkedIn Icon
  • Grey Facebook Icon

© 2023 by Talking Business.  Proudly created with

  • Twitter Social Icon