*This is an AI-powered machine translation of the original text in Portuguese
** Image resource obtained from Freepik.com
The effective measurement of the use of Artificial Intelligence (AI) in arbitrations is challenging due to the inherent confidentiality of the proceedings. Despite this difficulty, AI has various potential applications in this context, such as conducting discovery and identifying relevant documents, document translation, transcription of hearings, grammatical correction, text summarization, arbitrator selection, conflict analysis, case law research, generation of drafts, predictive analysis, and contract term review. However, due to the lack of standardization in the application of these tools and the recent widespread dissemination of AI solutions, there are generally no specific guidelines on their use in arbitral proceedings.
Depending on the application, the use of AI can be more or less controversial. For example, grammatical corrections or translations performed by AI tend to be less discussed issues, while the possibility of artificial intelligence drafting documents for arbitrators is a more controversial topic, raising questions similar to the controversy regarding whether there is a delegation of the arbitrator's decision-making power to the secretary. In these cases, it is relevant to establish rules for the use of AI in each procedure. To determine whether AI should be used in the procedure, it is crucial for lawyers to be vigilant and work to ensure that the use of the tool respects fundamental principles such as due process, adversarial proceedings, and the arbitrator's freedom of conviction. Additionally, it is important to be well-advised by professionals who not only master the law but also have the necessary knowledge to understand, in real-time, how the application of AI can influence the resolution of the dispute.
One of the challenges faced when incorporating AI into arbitrations is understanding its operation, given a certain opacity in its learning process, often referred to as the "black box." Therefore, it is relevant for AI systems to adhere to ethical and legal principles, meeting technical criteria of reliability, traceability, and accountability as a way to mitigate these risks. Currently, there is a discussion about addressing issues related to the difficulty of understanding decisions made by algorithms, which can be incorrect and/or biased, as well as the responsibility arising from the use of AI. Thus, it is important to ensure that there is human review of AI applications used in the procedure to promptly identify errors and potential discriminatory conclusions.
Another challenge is the scarcity of arbitration data available for AI training. AIs depend on a large volume of information to be accurately trained for specific purposes. However, the confidentiality surrounding arbitration proceedings makes it difficult to access data that could optimize AI applications in specific arbitration-related skills. For example, while audio transcription would likely not be affected by this limitation (given its similar applicability to transcribing speeches on different topics), the quality and accuracy of drafts generated by AIs may be compromised by the lack of extensive data on the subject, as they depend on data from a particular segment to allow accurate draft generation. Additionally, the specialized nature of arbitration cases adds complexity to the use of AI for this purpose.
It is important to clarify that AI does not replace the arbitrator but can act as a tool to enhance the efficiency of the arbitration process. The arbitrator remains the person chosen by the parties, while AI provides assistance without taking on the primary role. It is essential to ensure that the use of AI respects the constitutional guarantees of due process, including the right to participation and adversarial proceedings. Therefore, the application of AI to arbitration proceedings is a topic with various nuances. AI tools have the ability to support the development of activities to be completed more quickly, but the limits of their application must be clearly determined by the parties, regardless of how the technology is used.
