Mediation, with an AI co-pilot for the mediator
LLMediator is an experimental online dispute resolution (ODR) platform that explores how large language models (LLMs) can support mediation. It brings together two parties and a human mediator in a shared online environment while providing the mediator with an AI-powered co-pilot.
Cyberjustice Laboratory · Faculty of Law · Université de Montréal
What the platform does
Co-pilot draft
Let's separate the deposit from the repair costs and look at each in turn.
Suggested interventions
The co-pilot drafts a mediator intervention and proposes when to step in. The mediator edits it, approves it, or sets it aside.
Draft
You're lying — that deposit was mine and you know it!
Reformulated
I believe the deposit should be returned, and I'd like to explain why.
Message reformulation
Emotionally charged messages are reformulated into more constructive language before they reach the other party.
Rising tension — consider a caucus
Conversation dynamics
The platform monitors the dynamics of the exchange, so a conversation that is heading off course is visible early.
Clause 3 · Security deposit
− refund of $1,200 within 30 days
+ refund of $950 within 14 days
Agreed by both parties · PDF / Word
Settlement drafting
Points of agreement become a structured settlement draft, assembled from the parties' consensus and exportable to PDF or Word.
Two parties, one human mediator
A shared mediation room
Two parties and a human mediator work in one shared online environment, with a full record of the exchange.
Consumer
Landlord–tenant
Small claims
Common, high-volume disputes
Everyday disputes
The platform is aimed at common, high-volume conflicts — the disputes where access to justice is hardest to obtain.
Every AI-generated suggestion remains under the control of the human mediator, who decides whether and how to use it.
About LLMediator
The platform is designed to assist human mediators. It can suggest mediator interventions, help reformulate emotionally charged messages into more constructive language, monitor the dynamics of a conversation, and assist in drafting settlement agreements based on the parties’ consensus.
LLMediator focuses on improving access to justice by supporting the resolution of common, high-volume disputes, such as consumer, landlord–tenant, and small claims conflicts. By exploring how AI can augment human expertise, the project aims to make online dispute resolution more accessible, efficient, and fair.
About the research
LLMediator is an active research project investigating the capabilities and limitations of large language models in mediation and negotiation. It serves as a platform for evaluating how AI can responsibly support dispute resolution while preserving human oversight, transparency, and fairness.
The project has received funding from IVADO and the Canada First Research Excellence Fund through the research project “Exploring the Capability of an LLM to Act as a Mediator in Complex Mediation Cases,” led by Professor Karim Benyekhlef.
Current research includes multimodal dispute analysis (text, images, audio, and documents), AI-assisted drafting of settlement agreements, and new approaches to AI-supported mediation. Findings from the project have been published and presented at leading conferences in artificial intelligence, computational law, and online dispute resolution.
Publications and talks
A companion platform
Mediation Arena
Mediation Arena is an experimental research platform that explores how AI agents can participate in mediation through real-time voice communication. Developed at the Cyberjustice Laboratory, Faculty of Law, Université de Montréal, the platform investigates how conversational AI can facilitate dialogue between parties engaged in a dispute.
arena.llmediator.org
purpose · bridge → both parties
“Name the one point they have just agreed on.”
About the Cyberjustice Laboratory
Founded in 2010 by Professor Karim Benyekhlef at the Université de Montréal, the Cyberjustice Laboratory is an interdisciplinary research centre dedicated to studying how technology can improve access to justice and transform dispute resolution.
The Laboratory brings together legal scholars, computer scientists, engineers, and social scientists to study the impact of digital technologies on justice systems and to develop practical tools for courts, tribunals, and online dispute resolution. Its research covers areas including cyberjustice, artificial intelligence, online dispute resolution, digital courts, and legal technologies.
Beyond its permanent team at the Université de Montréal, the Cyberjustice Laboratory collaborates with an international network of universities, research centres, and institutional partners. Its research infrastructure includes experimental courtrooms, software development facilities, and simulation environments that enable researchers to design, prototype, and evaluate innovative justice technologies.
LLMediator builds on more than a decade of the Laboratory’s research into technology-enabled dispute resolution and reflects its mission of developing practical, evidence-based solutions that improve access to justice.
Mediation Arena builds on the Laboratory’s long-standing research in online dispute resolution and artificial intelligence, extending this work by exploring how autonomous conversational agents can participate in mediation through natural, real-time voice interactions.