Beyond human judgment: A critical examination of Artificial Intelligence in Arbitration via the LaPaglia v. Valve Corp. Case

1.    Introduction

Arbitration, as a favoured method of resolving disputes outside traditional courtrooms, has always balanced efficiency with the impartiality and nuanced judgment of human arbitrators. In recent years, AI tools have begun to permeate legal practice[1], offering the promise of enhanced efficiency, consistency, and cost-effectiveness. However, when AI is used in core adjudicative functions, questions regarding human accountability, transparency, and the reliability of decision-making processes arise. The LaPaglia v. Valve Corp. case serves as a focal point for this discussion, by questioning when AI moves from a supportive role to one that compromises the very essence of human adjudication.

2.    Overview of the LaPaglia v. Valve Corp. Case

In LaPaglia v. Valve Corp., the claimant challenged an arbitral award on the basis that the arbitrator allegedly excessively relied on AI - specifically mentioning the use of ChatGPT - to draft critical portions of the award. Key allegations include that the arbitrator reportedly admitted to employing AI to expedite the drafting process, citing an impending trip as a motive for fast-tracking the decision. There were also allegations of factual inaccuracies as the award allegedly incorporated statements and facts not present in the trial record, suggesting that AI-generated content influenced the outcome. Finally, the due process concerns were at stake as the undisclosed use of AI raised serious questions about whether the parties received a fully reasoned, human-authored decision in line with the arbitration agreement.

The claimant’s petition invokes Section 10(a)(4) of the Federal Arbitration Act (FAA), arguing that by “outsourcing” his adjudicative role, the arbitrator exceeded the powers granted under the arbitration agreement.

3.    What are the legal implications?

Under the FAA and longstanding jurisprudence, arbitrators are vested with the sole responsibility of rendering decisions based on a diligent examination of the record. The claimant’s invocation of Section 10(a)(4) draws parallels with cases where awards were vacated due to misrepresentation or unauthorized delegation of decision-making power. If AI is used beyond a supportive role, it arguably disrupts the balance of human oversight and breaches the parties’ contractual expectations.

A fundamental element of fairness in arbitration is that awards must emerge from a process that is both transparent and amenable to review. Failure to disclose the use of AI and the incorporation of unverifiable factual assertions not only clouds the rationale behind decisions but also violates the parties’ right to a reasoned and accountable decision. Judicial intervention could therefore be warranted if AI is found to undermine the integrity of the procedure.

4.    Pros and cons of using AI in Arbitration

AI’s ability to handle large datasets and expedite routine drafting tasks has undeniable advantages. In fact, AI tools can quickly process large volumes of data, organize evidence, and generate drafting templates, significantly reducing the time required for routine tasks[2]. By expediting the process, AI may lower the overall costs associated with arbitration, benefiting both arbitrators and parties. AI can help create consistent formats and language in drafting awards, reducing variability that might arise from human inconsistency. AI systems can effectively organize complex datasets and highlight relevant evidence, supporting arbitrators in making informed decisions. When used strictly as an analytical assistant, AI can provide a base for further human review, potentially contributing to a more thorough evaluation of facts and legal precedents.

However, these efficiency gains must be rigorously balanced against the risk of inaccuracies - commonly known as AI “hallucinations”- which can inadvertently contaminate the record with errors or fabricated information risking the inclusion of errors in the final award. Ensuring that any AI-derived content is meticulously verified remains a pressing practical challenge. The insistence on human oversight is compelling since legal adjudication inherently involves ethical, contextual, and interpretative nuances that current AI systems cannot replicate. Analogies to cases where awards were vacated due to misrepresentations in the decision-making process lend additional weight to the claimant’s argument. The legal tradition prioritizes the integrity of judicial reasoning - a principle potentially jeopardized by AI over-reliance.

Excessive reliance on AI might lead to a reduction in the essential human element of legal reasoning, where ethical and contextual nuances are best assessed by a trained arbitrator. When AI is used beyond administrative tasks, it might overstep this crucial boundary. Failure to disclose the extent or nature of AI involvement can undermine parties’ expectations of transparency and due process in a contested award. Non-disclosure or undisclosed reliance on AI can deprive parties of the opportunity to fully understand or challenge the basis for decisions, potentially compromising procedural fairness.

We understand that the rapid progression of AI technology may soon render some of today’s concerns obsolete. With improved verification mechanisms and more refined outputs, future AI systems might complement rather than compromise human decision-making. In the meantime, if used strictly as a drafting aid with rigorous human oversight, AI’s benefits can coexist with judicial integrity. There is need of clear and enforceable guidelines to overcome these challenges.

5.    Assessing the key principles in the emerging guidelines on AI in arbitration

In response to the rapid evolution of AI technologies, the arbitration community has developed emerging guidelines to define acceptable AI use without compromising human judgment. Two prominent frameworks are the Silicon Valley Arbitration & Mediation Center (SVAMC) Guidelines and the Chartered Institute of Arbitrators (Ciarb) Guideline.  These guidelines collectively underscore the view that while AI can serve as an auxiliary tool, its use should never compromise the independent judgment that is essential to the arbitration process.

5.1.        The key principles of the guidelines

A.    Non-delegation of adjudicative functions

Both the SVAMC Guidelines and the Ciarb Guideline emphasize that arbitrators must retain ultimate control over the decision-making process. Specifically, the SVAMC Guidelines (Guideline 6)[3] and the Ciarb Guideline (Article 8)[4] clearly state that while AI tools may assist with administrative or drafting functions, they must not replace the arbitrator’s independent analysis of facts, law, or evidence. This principle serves to protect the core responsibility of the arbitrator and maintain party expectations regarding a well-reasoned, human-authored award[5]. By firmly establishing that AI can only serve as a support tool, the guidelines help maintain the sanctity of human judgment in arbitration. This principle is crucial for preserving party confidence and ensuring that ethical and contextual considerations remain central to decision-making.

B.    Transparency and disclosure

Transparency is another cornerstone in these guidelines. The SVAMC Guidelines (Guideline 7)[6] and the Ciarb Guideline (Article 9)[7] mandate that any AI-generated information used in the arbitration process must be disclosed to the parties.[8] This requirement is intended to preserve due process by ensuring that all parties are aware of, and can comment on, the methods used to generate or support substantive conclusions. In practice, this could involve preemptive consultations or detailed disclosures within the award documentation.

Mandated disclosures foster an environment of openness that allows parties to understand and challenge the use of AI in the process. This transparency can prevent surprises during the post-award phase and reduce disputes regarding the source or accuracy of information relied upon by the arbitrator.

C.    Verification and accountability

A recurring concern with AI is its tendency to produce “hallucinations” or generate information that appears accurate but is ultimately flawed. To counter this, both sets of guidelines obligate arbitrators to independently verify any material produced by AI. By emphasizing robust human oversight, these guidelines aim to prevent the incorporation of erroneous or unsubstantiated material into the final award, thereby reinforcing the accountability of the arbitrator.

The guidelines offer a framework that can standardize AI use across arbitrations, potentially reducing the risk of variable practices among arbitrators. Uniform standards could simplify training, oversight, and ultimately improve the enforceability of awards.

5.2.        Challenges of the current guidelines and future considerations

The guidelines remain soft law instruments and are not uniformly binding across jurisdictions. Without formal regulatory backing, ensuring consistent adherence to these principles poses a significant challenge, potentially allowing for disparities in how AI is deployed and disclosed. As AI technology continues to advance, the current guidelines may quickly become outdated. Future iterations of AI might incorporate more reliable verification mechanisms, thereby challenging the rigid frameworks established today. Arbitrators and institutions will need to continuously update these guidelines to keep pace with technological improvements.

The requirement to verify every AI-generated output is a demanding standard that may place additional burdens on arbitrators. Adequate training, resources, and perhaps even new disclosure technology might be required to ensure that AI outputs are meticulously checked against the record, a process that could erode some of the efficiency gains achieved through AI. The line between administrative assistance and substantive decision-making can sometimes blur in practice, making it challenging to determine when AI’s use oversteps permissible boundaries. This ambiguity could invite disputes over the correct interpretation of the guidelines and necessitate judicial intervention to resolve such conflicts.

Looking ahead, arbitration institutions will likely need to collaborate with technology experts, legal scholars, and practitioners to create more enforceable, adaptive frameworks for AI use. Potential strategies include establishing binding regulations that integrate the guidelines into institutional rules and national legislation could standardize practices and enhance accountability; leveraging innovative tools to verify AI outputs may reduce the manual burden on arbitrators and ensure greater reliability in AI-assisted processes; continuous professional development for arbitrators on the latest AI technologies and their potential pitfalls will be key to ensuring these guidelines remain effective and relevant.

The advent of AI in arbitration necessitates the development of clear guidelines and protocols for its use. Arbitral institutions may be compelled to update their rules and procedural frameworks in response to these technological advancements. This evolution is crucial for ensuring that the efficiencies provided by AI are harnessed responsibly while safeguarding the integrity and fairness of the arbitration process.

Conclusion

The LaPaglia v. Valve Corp. case underscores the tensions at the intersection of technology and human judgment in arbitration. While AI has the potential to streamline processes, reduce costs, and enhance consistency, its integration raises serious questions about the erosion of human adjudicative authority, transparency, and due process. The emerging guidelines from SVAMC and Ciarb represent an important first step in setting boundaries for AI use in arbitration. However, challenges related to enforceability, technological evolution, and practical implementation remain considerable.

A sustainable future for AI-integrated arbitration hinges on developing binding standards that adapt to technological progress while preserving the core principles of human reasoning and accountability. Continued dialogue among legal scholars, practitioners, and technology experts will be essential to refining these frameworks and ensuring that AI remains an aid rather than a substitute for human judgment. The dialogue between technology and human judgment in arbitration is just beginning, and its evolution will undoubtedly influence the future of dispute resolution.

 By Prof. Joseph Yav,

Arbitrator, Attorney at Law and Professor of Law [www.yavassociates.com]