Artificial Intelligence in arbitration: use, challenges and limitations

How AI operates

AI works by using algorithms and large amounts of data to accumulate knowledge and learn from previous experiences. Over time, the system improves, allowing it to identify patterns of behaviour and information to help verify the output it generates.

Often, however, the information fed to the AI tool from which it learns and acquires knowledge is not easy to verify, making it extremely difficult to audit or review the database used by the AI developer.

The international community has already pointed out cases where AI has presented erroneous conclusions, especially when used for more complex and sophisticated issues, which raises questions about its potential use by legal practitioners, particularly concerning complicated legal matters.

Considering the lack of AI regulation in many parts of the world, AI’s use must comply with the ethical and legal principles of each jurisdiction where it is used. AI regulations must also promote reliability and well-defined criteria for information tracking to provide greater security to those who use it.

Use of AI in arbitration 

Today, we can envision the use of AI in conducting both national and international arbitrations. The tool promises to improve efficiency, increase the speed of procedures, reduce costs, and optimise the time spent by lawyers, arbitrators, experts and Arbitration Chambers.

The following are examples of the possible use of AI in arbitral proceedings:

• Document review: AI is capable of reviewing and categorising large volumes of documents and information, helping to quickly identify key information relevant to a party’s claim and reducing time and cost in conducting the review.
• Review of pleadings and testimonies: AI can assist lawyers in reviewing requests and testimonies, as well as helpfully organising the content received.
• Preliminary analysis: AI can analyse arbitration precedents, predict results or the possibility of negotiations, and assist parties in deciding whether to proceed with litigation or negotiate.
• Case management: AI case management can assist arbitrators and lawyers in conducting procedures more effectively, through automation of scheduling, deadline control and management of information and documents.
• Legal research and precedent analysis: AI can help lawyers find laws and precedents applicable to a specific case, improving the quality of the work developed.
• Transcription and translation of texts: In international arbitrations, AI can be an important tool for translating texts into other languages, as well as transcribing testimonies in a foreign language.
• Choice of arbitrators: AI is capable of researching the academic and professional background of arbitrators, and their performance in previous cases. Since many arbitrations are confidential, the information available may be limited.
• Compliance and due diligence: AI can assist parties and lawyers in identifying potential conflicts of interest, as well as in the applicability of laws and regulations to the specific case.
• Information security and privacy: AI can improve the security of confidential or sensitive information in arbitration, detecting and preventing violations of data protection regulation.
• Decision support: AI can assist arbitrators in analysing evidence and identifying patterns, as well as presenting relevant information brought by the parties, helping the Arbitral Tribunal to make decisions more swiftly.
• Virtual arbitrators or mediators: Some AI tools have been developed to act as arbitrators or mediators, aiming to assist in preliminary negotiations and fact gathering.

As seen, there are several possibilities for the use of AI in arbitral proceedings.

The implementation of AI, however, is not so simple, and the process poses risks and challenges that legal practitioners must overcome.

Challenges for the use of AI in arbitration

In a preliminary and theoretical analysis, the use of AI in arbitral proceedings could be extremely advantageous, as demonstrated above. However, whether due to inherent peculiarities of arbitration or non-negotiable legal principles, the practical application of this tool deserves attention.

The first and primary challenge in using AI in arbitration is the confidential nature of arbitral proceedings, which prevents the underlying algorithm from learning and processing new data due to the scarcity of information.

Scarcity of information is further exacerbated by the absence of repetitive patterns since arbitral proceedings do not generate precedents and arbitrators are not obliged to follow binding precedents.

Confidentiality in arbitration also hampers the algorithm's ability to ensure transparency of information due to data protection and privacy regulations now prevalent in many jurisdictions.

Beyond the issue of confidentiality, another question arises: would it be appropriate or even ethical for decisions to be based on or even decided by algorithms without human supervision?

The lack of transparency over the underlying data used by the AI is a concern to the international community as parties must understand the origin of the information used in important legal decisions.

Moreover, it has been observed that AI often presents fabricated or untrue information, which is highly problematic in the context of use in arbitral proceedings.

Another concern is the difficulty that AI could have in understanding the broader context of an arbitral dispute, which may involve applicable legislation, cultural aspects of each jurisdiction and the profile of the parties involved. All these factors influence the resolution of a given dispute.

Regarding the selection of arbitrators, it can be questioned whether parties, by mutual agreement, could realistically appoint a programmer behind an AI system as an arbitrator.

In arbitral proceedings, parties are free to choose any arbitrator they trust. Thus, if the parties mutually decide to appoint a programmer with an AI tool that they trust as an arbitrator, then, theoretically, there would be no impediment to choosing an arbitrator created by AI.

However, the principles of impartiality, independence, and neutrality of an arbitrator in arbitrations are equally essential conditions for the validity of the procedure. Currently, AI does not guarantee the application of these principles because, as mentioned above, it is not possible to verify the database used by AI, which evidently complicates compliance with these principles.

The same reasoning applies to the guarantees of due process and the arbitrator's impartiality in arbitral proceedings. The failure of an AI arbitrator to observe these guarantees could result in the nullity of an arbitral award, which would be a detriment to the security and effectiveness of the procedure.

As for the use of AI in arbitral hearings, this is also a controversial issue.

AI can transcribe in real-time what the parties and arbitrators say in a hearing, and seek arguments and evidence contrary to the claims of the opposing party, witnesses, and experts. In these cases, however, AI cannot not guarantee that the responses and information obtained are accurate or even correct, potentially leading the Arbitral Tribunal into error.

Conclusion

From all that has been discussed above, AI clearly has much to offer legal practitioners, especially in arbitral proceedings, since AI can facilitate and expedite the work of lawyers, arbitrators, experts, and the Arbitration Chamber.

However, until AI is broadly regulated in jurisdictions around the world, the technology’s application remains limited. If professionals decide to use AI in arbitral procedures, they should be aware of the risks mentioned above, adopt a cautious approach, and respect the principles and guarantees required by law.