Menu
Log in
Forgot password

Association for Conflict Resolution - Greater New York Chapter↑

Search
Log in
Forgot password

AI Arbitrator

Wednesday, October 01, 2025 2:21 PM | Anonymous
International Arbitration • Aceris Law

The AAA-ICDR announced in September that it will “release an AI Arbitrator to deliver fast, cost-effective, and trusted dispute resolution. The AI arbitrator will first be available in November 2025 for documents-only construction cases, a high-volume area where efficiency and speed are essential.”

What are your thoughts about the use of an AI arbitrator?

Comments

  • Thursday, October 02, 2025 1:08 PM | Robert L. Arrington
    As a lawyer, I would not use an AI Arbitrator.
    Link  •  Reply
  • Thursday, October 02, 2025 1:12 PM | Former judge of
    I think it would be a very valuable option. I wonder whether, at least until people become accustomed to it, there should be an ‘appeal’ to a live arbitrator with the appellant paying all costs and fees if the appeal returns a result substantially the same as the AI outcome.
    Link  •  Reply
  • Thursday, October 02, 2025 1:16 PM | David Vaughn
    It is the ethical obligation of neutrals to conduct their own analysis of disputes presented to them. Reliance by an arbitrator on an AI analysis would, in my opinion, violate that obligation.
    Link  •  Reply
  • Thursday, October 02, 2025 1:20 PM | Heather Shore, KC MO
    I would review the case of La Paglia v Valve Corp in the context of this discussion. My personal experience with AI is that it is very superficial and lacks the sophistication to appreciate or understand how certain issues and factors (including those that may have subjectivity) may or should play into the ultimate decision.
    Link  •  Reply
  • Thursday, October 02, 2025 1:23 PM | Anonymous
    I confess I do not fully understand how AI works. However, I have read fairly recently that all or most AI systems suffer from fantasizing and it has not yet been possible to eliminate that flaw. In the light of that concern, I must agree with those who have expressed reluctance or opposition to the idea of an AI arbitrator
    Link  •  Reply
  • Thursday, October 02, 2025 1:23 PM | Mel Simburg
    AI cannot think or evaluate nuances of particular cases. What it can do is aggregate past cases (if there is an available database) and decide based on the most common result for similar cases. That does not mean that it is giving the best result for the situation presented. It should be an advisory opinion only; best use is to assist settlement negotiations.
    Link  •  Reply
    • Thursday, October 02, 2025 1:48 PM | Nancy Greenwald
      I share this concern. I am curious to know what data the AI arbitrators are training on in light of the fact that awards are confidential and, in any event, rarely include detailed findings of fact and conclusions of law.
      Link  •  Reply
  • Thursday, October 02, 2025 1:26 PM | Steven Skulnik
    The AAA will have a human arbitrator review the Gen AI output and the human will issue the award. The diligence of the human is crucial.
    Link  •  Reply
    • Thursday, October 02, 2025 1:59 PM | Lisa Renee Pomerantz
      On one hand, arbitration permits the parties to agree on the procedures for resolving their disputes, provided there is compliance with the FAA or state analogs. I do have concern about how effective and efficient it will be to have a human arbitrator review and issue the AI arbitrator award. The human will have to review the actual documents to make sure they have been properly interpreted and applied, and there may be a lingering concern that the AI arbitrator missed something. Will there be an opportunity for the parties to review and comment on the AI arbitrator's draft award which could be of assistance to the human tasked with reviewing and issuing it?
      Link  •  Reply
  • Thursday, October 02, 2025 1:43 PM | Rick Flake
    If speed and costs are the driving factors in getting a dispute closed, (e.g. in a documents only, smallish amount in controversy case) I'm sure there are some parties who would agree to what I suspect would be the "rough justice" outcome using an AI arbitrator. For any nuanced dispute of substance however, I'm not sure who would want anything other than an experienced human to decide the issue. Then again, I am a Boomer and maybe my thinking is outdated....
    Link  •  Reply
  • Thursday, October 02, 2025 1:43 PM | Anonymous
    While AI can analyze data, I struggle to understand how AI can weight conflicting evidence and make a determination on what evidence to rely upon in arriving at a decision.
    Link  •  Reply
  • Thursday, October 02, 2025 1:47 PM | Sally Higgins
    I wouldn't even consider recommending this to a client unless I understood that it had been rigorously tested and there were some sort of appeal. For example, if I learned that AI and a human arbitrator had both considered some very significant number of cases and the results compared favorably, I might be more open to it. AI can be great, but it does go wrong and why should parties in a dispute take that chance?
    Link  •  Reply
  • Thursday, October 02, 2025 1:56 PM | Robert M. Gippin
    An AI arbitrator would not, in my judgment, be an "arbitrator" within the meaning of statutory or common law, nor as carried into standard arbitration agreements. I know of no reason why parties could not agree to it expressly, however. That said, I would be unlikely to advise a client to agree to it. If the result would be obvious enough for AI to determine, the case should be very amenable to settlement.
    Link  •  Reply
    • Thursday, October 02, 2025 2:09 PM | Margarita Echevarria
      I think it has a role in high volume non-complicated disputes. I have yet to get a response from AI that involves hallucination and all platforms include a disclaimer to review and make your own judgment. And if as someone responded a human will ultimately sign off, I don't see an issue for the type of case I described.
      Link  •  Reply
  • Thursday, October 02, 2025 2:18 PM | Richard Ziegler
    It seems to be our inevitable future, and AAA gets great credit for innovating, BUT this raises several challenging issues: (1) since current AI models are often unreliable, the human arbitrator's review will need to be searching and thorough, so there may be little cost savings; (2) because the AI's output will be a function of the database of rulings on which it was "trained," AAA should disclose the steps it took to ensure that the training opinions were (i) well-reasoned, (ii) thorough and (iii) balanced, including whether the database was skewed by relying solely on public opinions (or whether confidentiality obligations were satisfied if the database includes non-public opinions); and (3) AAA should consider the disclosures the AI arbitrator function should make to enable Parties to make bias challenges or explain why there is no basis for concern over possible bias. In short, I'm wary, as this seems to be a key step on the slippery slope towards replacing (human) arbitration as we know it!
    Link  •  Reply
  • Thursday, October 02, 2025 2:21 PM | Raoul East Drapeau
    I use AI practically every day as a useful advisor that has quick access to vast amounts of data such as state case law and statutes as well as technical practices and standards. But because of all its known faults I cannot imagine using it as a decider in something as nuanced as an arbitration or mediation.
    Link  •  Reply
  • Thursday, October 02, 2025 2:28 PM | Anonymous
    I can't tell whether this is a joke or not. Using a computer to decide lawsuits is nuts.
    Link  •  Reply
  • Thursday, October 02, 2025 3:00 PM | Peter Rundle
    A bit of a lengthy post - bear with me.

    It seems to me that an important first step in rolling out an AI arbitrator would be to take 10 - 20 (or more) moderately sophisticated arbitration files (Demand, Answer, Memorials, Witness Statements, Exhibits, Transcripts, Etc.), input them into the AI arbitrator tool, and request the output of an award based upon the materials. Then, for each test case, issue anonymized copies of the original tribunal's award together with the AI arbitrator generated award. I think this sort of transparent review by counsel and parties is the only way to achieve buy-in and "informed consent." Let's see the good, bad and ugly of the process as a first step.

    Taking a slightly different approach to the issue, having a human arbitrator review and approve of the AI output may be putting the proverbial cart before the horse. If the human arbitrator is editing what has already been generated, confirmation bias is built into the process. How diligent or willing will the human arbitrator be to contradict the machine? In a recent article in Arbitration International, "Reasons and reasoning in arbitral awards," 2025, Vol. 41, Issue 1, the authors (E. Thong & Y. Guo) discuss the crucial role that drafting plays in rendering a well-reasoned award that satisfies the parties' interest in both outcome and due process. Editing is not the same as crafting one's own product.

    Finally, to summarize what I've conveyed at several recent ADR conferences, every talented barrister, trial attorney and party advocate - much like the Sophists of the Roman law courts - hold dear the belief that through persuasive rhetoric and emotional engagement with the trier of fact, even a weak case can be won. If a completely detached and unemotional algorithm sits on the bench, or in the jury box, these rhetorical tools are lost. Whether that is a good or bad thing often depends upon which side of the courtroom - or conference table - you are seated. However, arguing a contrary point of view and playing devil's advocate, one might find support for the notion that an AI arbitrator eliminates the risk of deciding a case based upon inclination, feel or emotion - just as the Athenians attempted when the First Sophists were banned from their law courts.

    Thanks for sticking with me on this one. Food for thought.
    Link  •  Reply
  • Thursday, October 02, 2025 3:01 PM | Anonymous
    I thought this was a joke, but it seems it's real. I can't imagine how any competent lawyer could or would recommend this to a client.
    Link  •  Reply
    • Thursday, October 02, 2025 3:27 PM | Ruth Samas
      I have been involved in what is known as AI hallucinations. It made up a case and a holding. There are articles now about this phenomenon. Until AI is more reliable I would not trust it to make decisions.
      Link  •  Reply
  • Thursday, October 02, 2025 3:34 PM | Robert E. L. Wright
    I think it would be unwise to comment on the AAA's AI Arbitrator initiative without hearing the thoughtful, considered approach being taken by AAA which will have humans involved in the process. (Why is my comment being centered instead of left aligned or justified?)

    That said, I am interested to see how the implementation works. While I can envision issues arising, I think it is inevitable that AI will become a tool for litigants to use in resolving their disputes. Just as I fought (and lost) the war against having a computer on my desk, the war against AI will eventually be lost. So I always ask AI politely and never curse at it. Right Siri?
    Link  •  Reply
  • Thursday, October 02, 2025 3:37 PM | Peter Kaufman
    This is inevitable soon, but premature now.
    What datasets will it be trained on?
    What’s the point of having a human “supervising arbitrator(?)” review a documents only case after the AI arbitrator preps the Award? If that’s what’s required, you might as well have a human arbitrator handle the matter in the first place.
    Link  •  Reply
  • Thursday, October 02, 2025 4:41 PM | Gary L. Benton
    I agree with all the comments above: it’s good, it’s bad, it’s wonderful, it’s awful. It certainly draws attention to AAA, but the real test will be in the details. In my view, the concept is sound if the process ensures that skilled arbitrators are reviewing the inputs, the reasoning, and the outcome. Time and cost savings are commendable, so long as they do not come at the expense of an increase in wrongly decided awards.
    Link  •  Reply
    • Thursday, October 02, 2025 7:28 PM | The Honorable Neil Brown KC
      Obviously, there must be a human element involved sooner or later. But we should explore the whole issue of AI and, for example, intellectual property disputes. AI companies in Australia are asking for an exemption from copyright law so that they may have access to content in publications so that they can use it to "train" their AI machines. Publishers are not very keen on this being allowed. My experience in domain name arbitration has encouraged me to think that there could well be scope for the option of an arbitration on the papers with strict time limits and (comparatively) low costs to resolve such issues which will only increase in number as AI advances on its inevitable way. It would probably have to be supported by legislation or consent by the parties to using such a process, but it would be worthwhile to avoid the hight cost and delay of litigation. Does anyone else think this is worth pursuing?
      Link  •  
      Author
      Comment
       
  • Thursday, October 02, 2025 7:51 PM | Alan Levy
    AI has been known to hallucinate and not always give the correct answer. I find it difficult to believe that the current state of AI technology can judge subtle dishonesty and exaggeration. Maybe someday for small matters, but not yet for any arbitration.
    Link  •  Reply
  • Thursday, October 02, 2025 7:57 PM | Peter Scott Caldwell
    AI, in essence, is pattern recognition software. It can spot pattern in data that humans often cannot see. If the data it is analysing is flawed, the output will also be flawed.
    When people choose arbitrators the choice is based on trust. Humans have an innate ability to assess the trustworthiness of other humans. Our assessments can be wrong but they are more often right than wrong.
    We have no innate ability to assess the trustworthiness of an AI system. This seems to be the factor that is lacking in adopting AI as a quasi-arbitrator
    Link  •  Reply
  • Friday, October 03, 2025 6:12 AM | Denise Presley
    I’ve taken MANY continuing legal ed courses on AI, and would not recommend it for adjudicating party disputes. Regardless of what the marketers say, the technology simply isn’t ready to perform the work.
    Link  •  Reply
  • Friday, October 03, 2025 12:35 PM | Anonymous
    I believe that a neutral arbitrator must be a person. A computer AI is not a person and cannot be an arbitrator. If a lawyer consents to the use of a substitute for an arbitrator, in many states they would be subject to sanction for not properly representing their client. The process would not be more efficient in time or costs since there would be a right of review so instead of one step, there would be two with the time it would add to the process. For a party to get human review they could have to pay. Even if they want human interaction by a neutral but loose, they would have to pay for the proper desire to afford their client human involvement in the decision. The proposal fails on all aspects of being fair, proper, efficient, faster, and less costly for the parties. To upload my comment now, I must certify that I am not a robot and AI cannot ever give that assertion without halucinating which is another issue to contend with using AI
    Link  •  Reply
  • Saturday, October 04, 2025 1:49 AM | Prof Colin Roberts FCIArb
    AI is a useful tool for us to use; however, I am against replacing the intellect of an arbitrator with AI at this stage (or perhaps, ever). In the eyes of the parties, I would be horrified to think that "Computer said 'no'"
    Link  •  Reply
  • Saturday, October 04, 2025 10:59 AM | Jack Neal
    I would not recommend it to a client.
    Link  •  Reply
  • Sunday, October 05, 2025 10:26 AM | Leslie S Marks
    I asked AI your question. This is the response:
    The introduction of an AI arbitrator by the AAA-ICDR for documents-only construction cases marks a significant evolution in dispute resolution, with potential to increase efficiency, reduce costs, and address high-volume caseloads in areas where speed is critical. However, its adoption raises important considerations about trust, procedural fairness, transparency, and the ability of AI to interpret both the law and industry nuances in complex factual scenarios.

    Potential Benefits
    Efficiency and Cost Savings: AI arbitrators can quickly review documents and standardize outcomes, potentially accelerating case resolution and reducing party expenses compared to traditional arbitration or litigation.

    Consistency and Predictability: Algorithms can promote predictability in straightforward, repetitive disputes (such as those common in construction), reducing variability in awards.

    High-Volume Case Management: The deployment in construction disputes addresses a recognized need for rapid case turnarounds due to the time-sensitive nature of such conflicts.

    Key Concerns and Limitations
    Due Process and Fairness: There are questions about an AI’s capacity to fully comprehend context, nuances, and equity in individual cases—especially where written submissions may be ambiguous or incomplete.

    Transparency and Accountability: Stakeholders may challenge the “black box” quality of AI decision-making and seek assurance of unbiased, explainable outcomes.

    Legal and Ethical Considerations: Issues may arise relating to the enforceability of AI-driven awards, protection of confidential information, and compliance with relevant laws and arbitration standards.

    Suitable Scope and Future Impact
    The initial focus on documents-only cases suggests caution—AI is well-suited to standard, document-based disputes where interpretation is relatively straightforward and less reliant on live witness credibility. As the technology matures, its role may expand, but effective oversight, appeal mechanisms, and transparency standards will be essential to sustain trust and legitimacy.

    Overall, the move by AAA-ICDR could drive significant innovation, but careful implementation will be crucial to ensure that efficiency gains do not compromise core values of arbitration, such as impartiality and substantive fairness.
    Link  •  Reply
  • Sunday, October 05, 2025 12:48 PM | Steven Schwartz
    On September 19, I went to Comerica Park to watch the Detroit Tigers game. After the game, I Googled the attendance at the game. AI told me there was no Detroit Tigers game on September 19 and that their last game was in May 2025. That’s my opinion on whether AI should make arbitration decisions.
    Link  •  Reply
  • Sunday, October 05, 2025 2:55 PM | Linda F. Spiegel
    I have looked at the AI summary AAA offers in extremely simple construction matters and have found errors. If there are errors in the summary, there will be errors in the award. I strongly oppose using AI to decide any level arbitration. That being said, if the parties have an agreement to use AI then the parties are taking full responsibility for the outcome.
    Link  •  Reply
  • Sunday, October 05, 2025 4:50 PM | Josette Belvedere
    I like AI as an information gathering & dispensing tool. I do not think AI should be entrusted to resolve cases, however. AI is not accountable in the same way as a human arbitrator. Certainly, there would need to be a mechanism for "appeal" to a human if it were being used.
    Link  •  Reply
  • Sunday, October 05, 2025 7:01 PM | David W. Ichel
    I agree with Richard Ziegler's comments. I think that AI can have a role but it must at least at this stage have meaningful human review. As helpful as AI is, I have seen the hallucinations and the desire to please that can turn into bias. So the devil is in the details and the disclosure.
    Link  •  Reply
  • Tuesday, October 07, 2025 4:17 PM | Anonymous
    FAQ: Understanding the AAA AI Arbitrator
    By Jeffrey Zaino

    When I first posted that the AAA-ICDR would launch an AI arbitrator, the reaction was exactly what was hoped for: spirited, curious, and deeply thoughtful. Several respondents expressed skepticism. We welcome and appreciate all comments, and I look forward to sharing our progress as we work with clients and AAA arbitrators excited to be a part of the inaugural and voluntary roll out.

    That said, I would like to clarify what our AI arbitrator will be at launch, what it won’t be at launch, and how we’re making sure that the trust built over nearly a century of arbitration remains at the heart of this new process.

    First and Most Importantly: Every Case Has a Human Arbitrator

    Despite what the term “AI arbitrator” might suggest, no case is decided solely by AI.

    Here’s how it works:

    1. Parties submit their claims and evidence.
    2. Parties validate that the AI arbitrator has accurately summarized their submissions.
    3. AI arbitrator then parses claims, analyzes evidence, applies law, and drafts a proposed award with record citations.
    4. An AAA-trained human arbitrator reviews, revises as needed, finalizes, and issues the award.

    This “human-in-the-loop” structure allows for every decision to be grounded in real legal judgment and transparent reasoning—not algorithms alone.

    Furthermore, party participation is entirely optional. Both parties must agree to use the AI arbitrator. If one or both prefer a traditional AAA arbitration, that’s exactly what they’ll get.

    We’re starting with two-party, documents-only construction disputes—cases where there are no live witnesses and where speed and efficiency are paramount. As the technology matures, we’ll expand to other case types, always with human oversight built in.

    For nearly a century, the AAA has helped people and businesses resolve disputes fairly and efficiently. The AI arbitrator is our next step in carrying that mission forward—using technology to scale fairness, not to substitute for it. Take a look at the below selection of FAQs crafted to help address some common inquiries.

    1. Is the AI arbitrator actually deciding cases on its own?

    No. Every decision made by the AI arbitrator involves a human arbitrator. The AI arbitrator uses legal reasoning to draft a recommended award—not a final decision. An AAA-trained human arbitrator reviews the AI’s analysis, revises if needed, and issues the final, binding award.

    2. Does the AI replace human legal judgment or ethics?

    No. Human legal judgment remains central at every step. The AI arbitrator acts as an analytical tool—it organizes and reasons through case documents, but only a human arbitrator can finalize an award. Every output is validated for logic, fairness, and legal soundness by an experienced AAA arbitrator trained in AI arbitrator oversight. Carefully crafted AAA frameworks and standards are baked into every layer of the AI arbitration platform’s development to allow for transparency and accountability.

    3. What data is the AI arbitrator trained on?

    For its initial launch, the AI arbitrator was trained on real AAA construction awards—more than 1,500+ expert-labeled awards—and refined through expert human calibration. This training gives the system domain-specific expertise while protecting confidentiality.

    4. Can parties review the AI’s draft before a final award?

    This is not a black box. While the parties do not see the AI’s draft award, during the filing and response process, they do have the opportunity to review and confirm that the AI correctly summarized their claims and submissions before their information is submitted.

    5. Is participation in AI arbitration mandatory?

    No. The AI arbitrator is completely opt-in. Both parties must agree to use it; otherwise, the case proceeds under traditional AAA arbitration.

    6. How does this affect arbitrators?

    This innovation is designed to empower, not replace, human arbitrators. The AI arbitrator increases the speed and volume of cases, which we think can provide more opportunities for panelists. Arbitrators working on AI-led cases receive specialized training in reviewing AI-generated analyses.

    7. Is this suitable for complex or subjective cases?

    Not yet. The AI arbitrator’s initial deployment focuses on documents-only construction disputes—cases without live witnesses or complex factual issues. As the system matures AAA may expand its use to other case types.

    8. How do we ensure fairness and trust in this process?

    AAA’s reputation for fairness is built on nearly a century of trusted arbitration. Every AI-led case follows AAA’s ethical frameworks, due process principles, and disclosure requirements. Parties will always know who their arbitrator is, how the case is handled, and how the decision is made. AAA maintains a dedicated AI Governance Committee overseeing compliance, ethics, and model outputs.

    9. What are the time and cost benefits?

    Early testing shows 20–25% faster resolution times and 35% or greater cost savings in documents-only construction disputes.

    10. What if the AI “gets it wrong”?

    The human arbitrator is ultimately responsible for detecting and correcting any deficiencies before issuing the final award. Every award remains the human arbitrator’s award, not the AI’s. The AI assists, but the decision—and accountability—stays human.

    For more information or for a demo, please contact me at ZainoJ@adr.org.
    Link  •  Reply
Association for Conflict Resolution - Greater New York Chapter

© ACR-GNY

Contact Us

Email us at questions@acrgny.org

ACR-GNY's mission and programming are generously sponsored by:

ADR Notable: Dispute Resolution Management Made Easy

Discounts on ADR Notable platform available for ACR-GNY members!

Powered by Wild Apricot Membership Software