The Mediation Act, 2023 and the Question of Artificial Intelligence (AI): What the Legislature Did Not Say

I. Introduction

Parliament enacted the Mediation Act, 2023 at a moment when the backlog before India's courts had reached, by the National Judicial Data Grid's own count, something in the region of five crore pending cases, a figure which several generations of procedural reform had failed to reduce in any meaningful way. ¹It was against this backdrop that Parliament enacted a statute intended to make mediation a genuinely usable alternative to court proceedings rather than a preliminary formality that parties endured before returning to litigation. Section 27 of the Act provides that a mediated settlement agreement may be enforced directly as a decree of court, without any fresh application to a judge. This addressed one of the more persistent objections to choosing mediation over litigation. The Act also established a Mediation Council of India, with powers to accredit mediators, and created a class of registered mediation service providers.

The Act says nothing about artificial intelligence. By the time Parliament passed the Act, AI-assisted online dispute resolution platforms had already drawn the endorsement of the Supreme Court, and the NITI Aayog had published a detailed Online Dispute Resolution (ODR) framework in which AI figured as central to scaling dispute resolution in India. The Act defines mediation as a process in which a mediator assists the parties in arriving at a settlement that is the product of their own will,² which describes what a human mediator does. It says considerably less about what an algorithm does when it analyses the parties’ submissions, matches them against a database of prior outcomes, and proposes a settlement figure. That function is considerably closer to adjudication than to the facilitation of a mediation session, and it is a question the Act’s drafters do not appear to have turned their minds to.

II. The Algorithm in the Room

Katsh and Rabinovich-Einy, who wrote the most sustained academic account of online dispute resolution, describe the technology platform as the "fourth party" to any dispute resolved online. By this they mean that the platform shapes the process, the framing of the issues, and sometimes the range of outcomes available to the parties, without ever being formally acknowledged as a participant in the way a human mediator and two disputing parties are.³ Rule, writing from a practitioner’s perspective, put the same point more directly: the design of an ODR platform is never a neutral act, and whoever makes the design choices is, in a meaningful sense, making choices about how disputes will be resolved.⁴ Neither observation is an argument against technology in dispute resolution. Both are arguments for transparency about what that technology is actually doing, a requirement that the Mediation Act, 2023 does not currently impose. If the platform is a party in any meaningful sense, it is the only party whose interests, assumptions, and design choices the Act does not require to be disclosed.

The account of mediation that the Act implicitly adopts assigns the mediator a role that is deliberately limited. She facilitates communication, helps the parties identify their underlying interests, and creates conditions in which settlement becomes possible. She does not evaluate the merits of the parties’ positions, and she does not recommend outcomes.⁵ An AI platform that analyses a party’s written submissions, draws on a database of prior cases, and proposes that a settlement in the range of Rs. 4.5 to Rs. 6 lakh would reflect typical outcomes in disputes of this kind is not doing any of that. It is offering a form of non-binding adjudication based on historical precedent, conducted by software whose training data the parties have never seen and whose embedded assumptions they have no means to interrogate, at a time when the process is still being described to them as a consensual and party-driven one.

III. The Indian Picture

ODR is not new to India, though the ambitions attached to it have changed considerably. Its earliest deployments were in banking and e-commerce disputes, where the case for automation was largely uncontroversial: the disputes were small in value, large in number, and involved parties with no pre-existing relationship worth preserving and no interest in anything beyond a speedy refund or a corrected account statement. The Mediation Act, 2023 creates a framework for mediated settlement agreements to be enforced as decrees without further court intervention,⁶ and it establishes a Mediation Council with the power to recognise and set standards for online mediation service providers,⁷ but neither provision specifies what standards should govern AI use within those platforms, what training data providers must disclose, or how the Council is to determine whether an algorithm’s proposed outcomes reflect genuinely comparable cases or merely the historical bargaining disparities between the kinds of parties who have previously used the platform.

The negotiation literature has documented a pattern that makes this opacity consequential. Parties tend to place greater weight on proposals that appear to come from an algorithmic system, on the assumption that a technological process is, by definition, impartial in a way that a human mediator cannot be.⁸ That assumption is not warranted. An AI system is trained on historical data, and historical data encodes historical outcomes, including outcomes that reflect the relative bargaining power of the parties, the availability of legal advice, and the systemic pressures that have long favoured certain kinds of disputants over others. Szabo describes this as algorithmic sedimentation, the process by which historical inequities are encoded into systems that present themselves as neutral arbiters of the present,⁹ and it is a process that the Act’s silence on AI does nothing to address.

The access problem adds a further difficulty the Act does not address. The Law Commission of India observed in its 129th Report that the overwhelming majority of disputes in India involved parties with no access to legal counsel, no familiarity with formal dispute resolution processes, and no realistic prospect of enforcing any agreement they reached.¹⁰ Whether a party in rural Rajasthan or Jharkhand who is disputing a boundary or money claim can, in practice, navigate an AI-assisted ODR platform that requires a smartphone, a stable internet connection, and a degree of digital literacy that a large proportion of India’s population cannot be assumed to have,¹¹ is a question that neither the Act nor the rules framed under it have yet attempted to answer.

IV. Conclusion

The Mediation Act, 2023 addresses structural problems that have long prevented mediation from realising its potential in India, and the concerns raised in this article are not arguments against the Act or against the use of technology in the dispute resolution process. What the Act has not done is grapple with the question of what AI-assisted mediation does to the guarantees of neutrality, transparency, and party autonomy that section 2(h) and the Act’s broader architecture seek to affirm. The Mediation Council of India, in exercising the rulemaking powers that section 50 confers upon it,¹² has an opportunity to fill the gap that the legislature chose not to address. A mandatory disclosure regime for AI-assisted platforms, requiring the publication of training datasets, the specification of outcome categories the algorithm is designed to optimise for, and a clear statement of the categories of dispute for which AI assistance is suited, would go some distance toward ensuring that parties who use ODR under the Act’s framework understand what they are choosing. That, after all, is what the Act’s own definition of mediation requires.

References

1. National Judicial Data Grid, Pendency Statistics (Government of India, April 2025) https://njdg.ecourts.gov.in accessed 10 April 2025. ↩
2. Mediation Act 2023, s 2(h). ↩
3. Ethan Katsh and Orna Rabinovich-Einy, Digital Justice: Technology and the Internet of Disputes (OUP 2017) 12. ↩
4. Colin Rule, Online Dispute Resolution for Business (Jossey-Bass 2002) 58. ↩
5. Mediation Act 2023, s 17(1); Nadja Alexander, Mediation in a Nutshell (2nd edn, Thomson Reuters 2011) 3; Laurence Boulle, Mediation: Principles, Process, Practice (3rd edn, LexisNexis 2011) 7. ↩
6. Mediation Act 2023, s 30(2). ↩
7. Mediation Act 2023, s 33. ↩
8. Fernanda Herrera, 'Artificial Intelligence and the Future of Mediation' (2021) 37 Negotiation Journal 45, 52; Robert Mnookin, Scott Peppet and Andrew Tulumello, Beyond Winning: Negotiating to Create Value in Deals and Disputes (Harvard University Press 2000) 11. ↩
9. Imre Szabo, 'Algorithmic Neutrality: The Myth of Unbiased Automated Mediation' (2020) 11 Journal of Dispute Resolution 77, 82. ↩
10. Law Commission of India, 129th Report on Urban Litigation: Mediation as an Alternative to Adjudication (Government of India 1988) para 3.2. ↩
11. Telecom Regulatory Authority of India, Annual Broadband Report 2023 (TRAI 2024) 18. ↩
12. Mediation Act 2023, s 50(2). ↩