The Neutral in the Machine? Emerging Designs of Dispute Assessment in Arbitration

The pressure to resolve international disputes faster and at lower cost has driven a significant shift in institutional thinking. Alongside formal arbitration and traditional ADR, a new mechanism has emerged: dispute assessment provided by arbitration institutions, offering parties an early, structured evaluation of their disputes. Although such tools share similarities with early neutral evaluation and non-binding expert determination, they are fundamentally different. They are more closely connected to arbitration: designed and administered within established arbitral frameworks, either reflecting the institution’s own procedures or built upon its accumulated arbitral data. This lends the output procedural and substantive grounding in actual arbitral reasoning. Furthermore, the results have a significant influence on arbitration, either forming the basis of arbitral proceedings or operating within them, influencing party strategy and shaping the course and outcome of the arbitration.

Yet despite this common purpose, the emerging tools reflect two fundamentally different design philosophies, taking shape in the SCC Express Dispute Assessment and the AAA’s AI Resolution Simulator.

Two Models, One Goal

1.     Model A: Institution-Administered, Human-Neutral Evaluation

The first model situates the assessment within a structured institutional framework. The process is consensual, requiring the agreement of all parties, and is administered by the institution, which appoints a qualified neutral expert and oversees proceedings from referral to delivery of findings. The neutral expert conducts the evaluation broadly like an arbitrator.

The SCC Express exemplifies this model. Upon receipt of a request, the SCC appoints a neutral applying the same criteria used for arbitrator appointments. The neutral holds a case management conference, establishes a timetable, and delivers a reasoned non-binding assessment within 21 days of referral. The procedure also permits oral hearings where appropriate. The entire process is governed by the SCC Express Rules 2023 and is strictly confidential. Parties, counsel, and the arbitration community can scrutinise the framework, anticipate its operation, and hold the institution to its published standards.

Parties may agree to make the findings contractually binding or appoint the neutral as arbitrator to confirm the assessment in an enforceable award, transitioning the matter into arbitration under the applicable SCC Rules. Whilst the SCC Express mirrors several procedural steps of SCC arbitration, it is a markedly more compressed process offering fast and cost-predictable dispute assessment. In SCC arbitrations under the Arbitration Rules or Expedited Arbitration Rules, arbitrators have significantly more time, can receive extensive evidence, and conduct full merits hearings before issuing a binding award.

To facilitate adoption, the SCC offers three model clauses for incorporation into contracts, as well as a standalone model agreement enabling parties to refer an existing dispute to SCC Express on an ad hoc basis, regardless of whether such a clause was included in the underlying contract.

2.     Model B: Institution-Designed, AI-Powered Simulation

The second model employs AI to generate a non-binding, AI-driven simulated decision. It is generally designed for single-party use: the party submits its materials and receives an AI-generated simulated outcome.

The AAA’s AI Resolution Simulator, launched in March 2026, exemplifies this model. It builds on the AAA’s AI Arbitrator, a tool initially developed to handle document-only construction cases by automatically evaluating case merits, generating recommendations, and preparing draft awards for human arbitrator review. The Resolution Simulator extends this reasoning engine into the early stages of arbitration in response to demand for practical, evaluative tools.

The Simulator operates in four stages: submitting dispute materials; reviewing summarisation and key issues generated by AI; receiving a structured case analysis with supporting citations; and receiving a simulated decision. This indicates that the tool is used after arbitration has commenced, when dispute materials from both sides are available.  While no official timeline has been published, the process is likely slower than a typical AI chatbot, as the second stage requires user interaction. That said, it remains considerably faster than human-led assessment. As the process is neither overseen by a human arbitrator nor initiated by mutual consent, the simulated decision is non-binding, a limitation expressly acknowledged by the AAA.

Design Differences: Where the Paths Diverge

1.     Bilateralism vs. Unilateralism

For SCC Express, both parties bear procedural obligations and enjoy procedural rights, and the outcome can be made contractually binding or serve as a strong reference point. The fundamental premise is that both parties participate on equal footing. Such bilateralism closely aligns with due process in formal dispute resolution: both parties can present their views, the resulting assessment possesses greater legitimacy, and it is more consistent with institutional traditions of arbitration, making it more readily relied upon in subsequent proceedings.

In contrast, the AAA’s Resolution Simulator is designed for unilateral use. The party using the tool is required to submit its own case materials —  including claims, counterclaims, supporting evidence, and legal materials —  as well as the opposing party’s submissions where available. The inclusion of the other party’s submissions does not alter the fact that outcomes remain grounded in unilateral consent and have no legal effect on the other party. However, precisely because they are unilateral, the process may be considerably faster and more accessible.

2.     Human Judgement vs. Algorithmic Output

The distinction between the two models becomes particularly pronounced when examining who — or what — stands behind the assessment. In SCC Express, the human neutral is a named, qualified legal professional appointed through a transparent institutional process. Their reasoning is set out in a written, reasoned assessment. They can be challenged on grounds of impartiality or independence, replaced, and held professionally accountable, though the involvement of a human neutral inevitably carries higher costs.

In the AAA’s Resolution Simulator, the output is generated by an institution-designed AI system. It cannot be challenged in the same manner as a human decision-maker, nor is it subject to equivalent professional accountability or procedural control. While users may interrogate the output, they cannot meaningfully test its underlying methodology through adversarial means. However, what stands behind such an assessment is the accumulated legal reasoning of arbitrators whose decisions have been recorded through the institution’s case history. The Simulator’s design is also guided by principles of governance, transparency, and auditability, reflecting the AAA’s commitment to responsible AI adoption.

Procedural Risks Worth Watching

1.     Confidentiality and Privilege

Any dispute assessment mechanism raises questions about how submissions are stored, protected, and potentially used in subsequent proceedings. Where the process is governed by published institutional rules like the SCC Express Rules, the confidentiality regime is established and known to all parties in advance. By contrast, where submissions are made to an AI platform, parties must understand the applicable data governance framework, including how materials are stored and processed. For the AAA’s Resolution Simulator, a single party uploads materials after arbitration has commenced; these remain subject to the confidentiality obligations of the arbitral proceedings. It is, however, unclear whether the AAA’s AI system would be considered a third party if the parties commenced arbitration with the AAA, and parties must take care not to inadvertently include privileged communications concerning the opposing party. Such concerns are particularly acute in cross-border disputes where privacy and data protection laws vary across jurisdictions.

2.     Strategic Distortion: AI Bias vs. Human Bias

The predictive value of any assessment depends on the quality and completeness of submissions and the assessor’s capacity for independent, impartial judgement. AI carries a structural risk of bias, e.g., the AI may reinforce the submitting party’s narrative. In the AAA’s Resolution Simulator, such bias is further compounded by the tool being grounded in AAA’s internal case history, reflecting systemic patterns in past decision-making that do not account for evolving legal standards or the particularities of an individual dispute.

By contrast, the neutral expert in SCC Express can exercise independent judgement, while ensuring both parties have an equal opportunity to present their case. Nevertheless, human neutrals are also susceptible to biases, and AI tools can in some respects offer greater consistency by applying the same analytical framework across all cases. The key distinction lies in accountability: human neutrals are procedurally and professionally accountable, whereas AI outputs cannot be independently interrogated or challenged in the same manner, leaving parties reliant on the integrity of the system itself.

3.     Settlement Pressure: Procedural Abuse and Inequality of Arms

A non-binding assessment carries no formal legal consequence, but this does not render it strategically neutral. A dispute assessment can pressure a counterparty into settlement on unfavourable terms, particularly where one party commands greater resources, risking entrenchment of inequality of arms. SCC Express mitigates this risk through its bilateral consent requirement. By contrast, the AAA’s Resolution Simulator lacks the same structural safeguard. Using simulated decisions to force settlement during arbitration could raise due process concerns. However, this does not diminish the Simulator’s utility as a tool to inform strategy and identify the best path for resolution. Its effective use will depend on careful interpretation by counsel and awareness of its limitations.

Conclusion

Dispute assessment tools designed by arbitration institutions represent an innovative approach to dispute resolution. The SCC Express, grounded in institutional rules, produces outcomes with high credibility and a clear pathway to binding effect, but comes with higher costs and demands mutual commitment. The AAA’s Resolution Simulator offers accessibility and immediacy through unilateral engagement, yet raises legitimate concerns over AI bias, confidentiality, and strategic misuse. As these tools diversify and AI becomes increasingly embedded in practice, the arbitration community must engage critically with how they work, their advantages, and their limitations. Their value will ultimately be measured not by technological sophistication, but by the discernment with which practitioners employ them.