Study helps define how to use AI for dispute resolution

A project by two researchers from the Universitat Oberta de Catalunya (UOC) has analysed the impact of the new European Regulation on Artificial Intelligence (AI Act) on alternative and online dispute resolution (ADR and ODR) systems. The study concludes that AI can significantly improve access to justice and bring numerous benefits, but it also suffers from a series of significant limitations and risks if it is not used correctly. To ensure success, the authors propose a new categorization of alternative AI-assisted dispute resolution systems according to their degree of autonomy and risk. This categorization, which has received the Elías Campo Villegas Award from the Catalan Association of Notaries, will help in the design of public policies and more precise regulatory frameworks in line with technological advances, and will be a useful tool for providers of resolution services, ensuring that they choose the most appropriate system.

The study, published as open access in the journal Computer Law & Security Review, was conducted by two members of the Faculty of Law and Political Science: Professor Aura Esther Vilalta, a researcher in the Law, Internet and Digital Transformation (DIDT) group; and Marian Gili Saldaña, a researcher in the Taxation, Labour Relations and Business (TAXBUSINESS) group. "When the EU AI Act was approved in 2024, we realized that the impact of AI on alternative methods for resolving disputes in the courts of justice was not clear. It was necessary to assess whether the new regulations really protect people when AI is used in resolving disputes, because, with the level of development and application we see in these technologies – particularly generative AI – the new AI agents could become de facto judges or arbitrators," said Vilalta.

The two researchers, who are affiliated to the Digital Transformation and Governance Research Centre UOC-DIGIT, have focused their work on ADR (ways of resolving disputes outside the courts, through mediation, conciliation, arbitration, adjudication, etc.) and ODR (a tech-mediated version of the previous approach, which has gained wider significance in the post-pandemic digitalization process). It involves tools such as legal chatbots, systems for analysing and generating documents and decision-making assistants that are already being used successfully in specialized contexts. Another example that many people may have interacted with without being aware of it is the automated solution systems that platforms such as Amazon or eBay use when there is a dispute with customers.

Classification into four levels

From the results obtained, the UOC researchers have concluded that while not all AI in justice is high risk, some tools that seem innocuous, such as those offering assistance in negotiations, can influence the parties' decisions. To dispel doubts, the authors propose a new categorization of AI-assisted dispute resolution mechanisms according to their degree of risk and autonomy. In it, the category of minimum risk would include the use of AI for exclusively auxiliary administrative activities that do not affect the individual outcome of cases, such as text translation or audio-to-text transcription. At a second level, that of limited risk, there would be AI that interacts with the parties and generates synthetic content, provides information or performs support tasks, for example complaint triage and classification systems or chatbots that generate content during the dispute resolution process.

Moving up, the high-risk category corresponds to AI systems that assist the parties and produce results with possible legal effects, such as those that prepare drafts of legal documents, automatically review agreements or give advice based on precedents. The final category corresponds to cases of unacceptable risk, which includes wholly autonomous systems, with AI that generates automated decisions without human supervision that have legal effects on the parties. There should be few cases of this type, given that the AI Act emphasizes the imperative need for meaningful human participation in decision-making processes.

"It is clear that AI brings benefits such as greater speed, cost reduction, greater accessibility and the elimination of human errors due to fatigue. But there are also threats such as the "black box" effect (not knowing how AI has reached a conclusion), loss of empathy, and the risk that the system is biased or does not take human factors into account," said Gili. According to the researchers, there is a real risk that AI could compromise the right to a fair legal outcome. "This includes the right to be heard and the right to a reasoned decision. If an AI system makes a decision based on historical or discriminatory data or information that are not transparent, effective judicial protection is being violated, even if we are outside a conventional court," Vilalta added.

Transparency, auditability and human agency are key

To avoid possible algorithmic bias in decision-making, the experts stress that the solution should not be exclusively technical, but also human. They propose measures such as external audits that constantly review training datasets, and ensuring the quality of the underlying data and of the results, and that a human expert will always be in charge of making the final decision. There must also be algorithmic transparency, allowing the parties to know if AI is being used and under what guidelines.

The study includes a compliance guide both for the administration of justice and for those responsible for resolving disputes,  which provides a legal basis to avoid future appeals and proceedings that are null and void. "For example, before using chatbots or case triage systems, an assessment must be carried out of the impact they will have on fundamental rights," said Gili. "Moreover, operators must be properly trained in the field of justice, so that their interactions with AI do not violate people's rights, as any information provided via an AI chatbot can be public and accessible to anyone, if appropriate measures are not taken," Vilalta said.

Vilalta sees the constant updating of legislation and protocols regarding AI as one of the most important challenges today, "given that the tech industry is advancing at great speed." The work of the two researchers from the Faculty of Law and Political Science, however, lays the foundations for progress in this direction, which is why it has captured the attention of the Catalan Association of Notaries. It has been declared joint winner of the third Elías Campo Villegas Award in the "doctrinal" category as one of the most significant legal works on alternative methods of dispute resolution.

This study aligns with the UOC's Ethical and human-centred technology mission and contributes to the UN's Sustainable Development Goals, (SDGs), especially 16, Peace, Justice and Strong Institutions, and 9, Industry, Innovation and Infrastructure.

Reference article

Vilalta, E. and Gili, M. (2025). AI-driven alternative and online dispute resolution in the European Union: An analysis of the legal framework and a proposed categorization. Computer Law & Security Review, Volume 57, 106145: https://doi.org/10.1016/j.clsr.2025.106145

Transformative, impactful research

At the UOC, we see research as a strategic tool to advance towards a future society that is more critical, responsible and nonconformist. With this vision, we conduct applied research that's interdisciplinary and linked to the most important social, technological and educational challenges.  

The UOC’s over 500 researchers and more than 50 research groups are working in five research centres focusing on five missions: lifelong learning; ethical and human-centred technology; digital transition and sustainability; culture for a critical society, and digital health and planetary well-being.

The university's Hubbik platform fosters knowledge transfer and entrepreneurship in the UOC community.

More information: www.uoc.edu/en/research