AI and Online Mediation and Conciliation (ODR)

This page is a translation of the original Portuguese text, provided for convenience. It deals with Brazilian law. If there is any discrepancy, the Portuguese version prevails. Read the original in Portuguese

Artificial intelligence is transforming online mediation and conciliation, creating new regulatory challenges that call for immediate attention from legal professionals and lawmakers.

The Rise of AI-Powered ODR Platforms

Online dispute resolution, known by the acronym ODR, is no longer a distant promise. It is now an established part of the justice system in Brazil and abroad. Looking at how these platforms have evolved in recent years, it is clear that the addition of artificial intelligence tools has been a major leap in how conflicts are managed, sorted and, in many cases, resolved without traditional court intervention.

AI-equipped ODR platforms work on several fronts. Natural language processing algorithms can read court filings, categorize disputes by type and complexity, suggest settlement amounts based on similar precedents, and even carry out the preliminary stages of automated negotiation between the parties. In Brazil, the National Council of Justice (CNJ), the body that oversees the country’s judiciary, has encouraged the adoption of technological solutions to relieve the burden on the courts, and a number of private and public platforms already build some degree of intelligent automation into their mediation and conciliation workflows.

This transformation is not limited to consumer disputes. Areas such as family law, debt collection, condominium conflicts and even less complex labor matters already have experience with AI-assisted mediation. The potential for scale is enormous: while a human mediator handles one session at a time, automated systems can manage thousands of negotiations at once, continuously adjusting proposals and identifying points of agreement between the parties.

Brazil’s Regulatory Framework and Its Gaps

A look at Brazil’s regulatory landscape for the use of AI in online mediation and conciliation reveals a patchwork of rules. They are relevant, but they do not yet provide a specific, complete framework for this type of dispute resolution. The Mediation Law (Law 13.140/2015) sets out the general principles of mediation, including impartiality, party autonomy and confidentiality. But it was written before artificial intelligence tools became widespread, and it does not expressly address the challenges these technologies bring.

The Legal Framework for Artificial Intelligence, which moved through the National Congress amid intense debate over algorithmic governance, contains general provisions on transparency, explainability and accountability for AI systems. Applying it to the specific context of dispute resolution, however, requires careful interpretation. The overlap between AI regulation and procedural rules creates zones of uncertainty that must be confronted. When an algorithm suggests a settlement amount, is that suggestion a decision? Who is responsible for an unfair outcome produced by algorithmic bias?

CNJ Resolution 358/2020, which regulates the creation of technological solutions for resolving conflicts within the judiciary, is a step forward because it recognizes the legitimacy of digital platforms. Even so, its provisions do not spell out technical requirements for the AI algorithms used on those platforms, such as minimum standards for explaining automated decisions or protocols for algorithmic audits. This regulatory gap is worrying, because opaque systems can reproduce and amplify structural inequalities without the parties or the system’s own operators noticing.

Regulation of artificial intelligence on online mediation and conciliation platforms must balance technological innovation with basic procedural guarantees, especially the parties’ right to understand how automated decisions are reached.

Internationally, UNCITRAL (United Nations Commission on International Trade Law) and the European Union have been producing ODR guidelines that increasingly address AI concerns. The European AI Act, for example, classifies AI systems used in the administration of justice as “high-risk,” subjecting them to strict compliance requirements. This approach offers a valuable point of comparison for improving Brazilian regulation.

Algorithmic Transparency and Procedural Guarantees

One of the most sensitive points where AI and ODR meet is algorithmic transparency. In a traditional mediation, whether in person or by videoconference, you can observe the mediator’s conduct, question their suggestions and understand the reasoning behind the proposals. In an automated environment, that dynamic changes radically. If the algorithm suggests that Party A accept a specific amount as a settlement, what criteria was that suggestion based on? The parties have a right to know.

The principle of explainability runs through both Brazil’s General Data Protection Law (LGPD), in Article 20, which deals with automated decisions, and the broader debate over AI regulation. It requires systems to be able to give understandable explanations of their decision-making. In the ODR context, this means that platforms that use AI to suggest settlements, prioritize disputes or assess risks must be able to show, in plain language, the factors that shaped each recommendation.

Confidentiality, a cornerstone of mediation, faces added challenges when the parties’ data feeds machine learning systems. If information from thousands of earlier mediations is used to train predictive models, how can we ensure that sensitive data is not indirectly exposed, or that discriminatory patterns are not built into the model? Compliance with the LGPD requires that this data be processed on an appropriate legal basis, for specific purposes and with robust security measures.

Another aspect worth highlighting is informed consent. Parties who use ODR platforms with AI components must be clearly told what role the technology plays in the process, which decisions are automated, which are merely assisted by an algorithm, and at what points human intervention is guaranteed. This information must be presented clearly and in advance, not buried in lengthy terms of use that are rarely read.

Algorithmic Bias and Access to Justice

A central concern in applying AI to online mediation and conciliation is algorithmic bias and its impact on access to justice. AI systems are trained on historical data that inevitably reflects existing social, economic and regional inequalities. If a model learns that disputes involving a certain socioeconomic profile tend to end in lower settlements, it may perpetuate that disparity by systematically suggesting lower amounts for parties with similar characteristics.

This risk is particularly serious in Brazil, where deep structural inequalities can be encoded in training datasets. An ODR platform that uses AI without adequate mechanisms for detecting and correcting bias may, paradoxically, worsen the very imbalances that mediation is supposed to help reduce. Sector-specific regulation needs to establish mandatory bias audits, with periodic testing to assess whether the system’s results show unjustified disparities between different groups.

At the same time, AI has the potential to significantly expand access to justice. Automated ODR platforms can serve remote communities, run outside business hours, operate in multiple languages and dramatically reduce the cost of resolving conflicts. For lower-value disputes, where the cost of a traditional lawsuit would be out of proportion to the expected benefit, automated mediation may be the only workable path to resolution. The regulatory challenge is to capture these benefits without sacrificing fundamental guarantees.

Outlook for Sector-Specific AI Regulation in ODR

Looking ahead to the regulation of AI applied to online dispute resolution in Brazil, a few guidelines stand out as essential. First, regulation should be proportionate to risk. Not every use of AI in ODR is equally sensitive: a chatbot that helps fill out initial forms calls for different controls than a system that actually suggests settlement terms or evaluates the merits of claims.

Second, we advocate creating sector-specific certification standards for ODR platforms that use AI. These standards should cover algorithmic transparency, independent audit protocols, ways to challenge automated decisions, data protection policies, and performance indicators that make it possible to verify the fairness of results over time. The CNJ, working with the data protection authority and any AI regulatory bodies that may be created, would be the natural body to lead this effort within the judiciary.

Third, meaningful human intervention at critical stages of the process is indispensable. Fully automated mediation, with no human oversight at all, raises ethical and legal questions that current technology cannot satisfactorily resolve. A hybrid model, in which AI acts as a support tool for the human mediator (enhancing their capacity for analysis and speeding up operational tasks), seems better suited from both a regulatory and a practical standpoint.

Finally, the training of mediators and conciliators needs to be updated to include digital skills and a basic understanding of how algorithms work. Professionals who work on ODR platforms must be able to critically assess AI-generated suggestions, spot possible distortions and step in when necessary. Regulation can and should encourage this training as a requirement for working in technology-assisted dispute resolution environments.

This subject is directly related to online mediation platforms for disputes, which we cover in a dedicated article.

This subject is directly related to cookies and online tracking, which we cover in a dedicated article.

As informações deste artigo são de caráter informativo e não substituem consulta jurídica individualizada. Cada caso possui particularidades que exigem análise específica.