More than two years after the launch of ChatGPT, AI is starting to deliver for legal aid and self-represented litigants. On June 20th, researchers and lawyers from around the world showcased that growth at an academic workshop co-organized by the LIT Lab.
The solutions are diverse. Around the world, people are using AI to:
- Help people access legal information and create their own documents using chatbots and other easy-to-use tools
- Organize and visualize complex data to support work on issues like sentencing reform and the school-to-prison pipeline
- Use text messages to spot legal problems early and share helpful legal tips, resources, and templates for self-advocacy
- Predict custody decisions to support better understanding and fair negotiations
- Automate the review of lower risk legal aid requests, court fee waivers, and repeated filings to save time and improve consistency
- Direct employees looking for help to the best resource for disputes with their employer
- Help legal professionals manage large amounts of legal text by building structured maps of information for easier research and analysis
- Make legal and government documents easier to understand by improving how they’re written
Authors also discussed the broader AI policy landscape, addressing topics like unauthorized practice of law and a novel viability scoring approach to measuring true access to just outcomes. Spriha Bhandari, from India’s Jindal Global University, shared a new participatory design toolkit for AI builders.
We accepted 22 papers from around the world for the workshop, from a competitive pool of 30 submissions. Participants came from Australia, Canada, Czech Republic, Finland, Germany, India, Ireland, Italy, Netherlands, Singapore, Taiwan, United Kingdom, United States, and Vietnam.
One of the attendees, a long-time AI and law practitioner, said the conference revealed that we are “drowning in low-hanging fruit.”
The workshop’s growth over time and the proliferation of new tools demonstrate how LLMs empower legal tech development and make it easier for people to create new tools. While in the past computer science academics dominated the conference, AI has opened new doors for practitioners and lawyers, and frameworks like Streamlit and Gradio make building AI solutions quick and easy.
Drew Jackson of the People’s Law School in British Colombia, along with their tech partner, walked us through a benchmark dataset of legal advice as well as a detailed exploration of how they used it to test their tool.
Data scientist and law student Brian Rhindress brought us away from large language models into the world of image-generation for access to justice with a provocative exploration of seven ways that AI-generated images could improve legal communication, from comic books for court volunteers to inspiration for redesigning court forms, grounded by examples and court-form design guidance.
There were other fun moments. The team from Singapore Management University—Jonathan Su, Lydia Tan, Nicole Daduya, Jayme Loo, Isaac Lu, Bryan Liew, and How Khang Lim—shared their prompting technique for their automated legal aid case law summarization and research tool, which included telling the LLM that it had been abducted by aliens.
LLMs didn’t dominate every discussion. Hsuan-Lei Shao and Sieh-Chuen Huang shared their traditional machine learning-powered algorithm for predicting custody outcomes in Taiwanese family law disputes.
Workshop attendees showed that AI could improve empathy and psychological safety for litigants, not just make work more efficient. Mattia Falduti and Anca Radu in Italy developed an AI tool to help survivors of image-based sexual abuse (IBSA) draft initial reports, directly addressing a crime type that has “the lowest reporting rate” and helping to overcome common barriers like “blank page anxiety.” Zoe Dolan, who ran an AI-training workshop for self-represented appellants, reported participants mentioning “healing trauma” and emotional support. Participants felt that having the AI to consult eased the isolation and anxiety of going through an appeal alone. This idea of taking AI-safety training directly to in-person clinics was a striking one for me.
As a document automation enthusiast, I was also excited to see the work that Marc Lauritsen and Hannes Westermann are doing to experiment with integrating large language models into otherwise rigid document automation workflows for family disputes in Washington State.
There were so many great papers that I feel irresponsible to stop here, but the great thing is that you can read all of the papers online, for free. If you’re feeling nostalgia for the Fall of 2024, you can even listen to an AI-generated audio overview of the conference made by uploading all of the papers to NotebookLM.
Here is the full list of papers:
- Testing a GenAI Legal Chatbot – Drew Jackson, David Kandestin, Spencer Keene and Chris McGrath
- In With the Old: Enhancing ‘Classical’ Document Automation with Generative AI – Marc Lauritsen and Hannes Westermann
- Drawing the Line: Legal Information vs. Legal Advice in AI Tools for Justice – Shivani Sathe, Avanti Durani and Abhishikta Mallick
- Harnessing Generative AI to Enhance Access to Online Dispute Resolution in Employment Law – Sarah Nason, Ebube G. Ogbonda, Jonathan C. Roberts and Peter W. S. Butcher
- AI Program for Self-Represented Litigants: Methodology, Outcomes, and Insights – Zoe Dolan and Aiden An Ai
- Juxtaposing Legal AI Tools and their Practical Use – Juraj Vladika, Stephen Meisenbacher, Benedikt Thiess and Florian Matthes
- Leveraging LLMs for Efficiency in Family Law Research – Jonathan Su, Lydia Tan, Nicole Daduya, Jayme Loo, Isaac Lu, Bryan Liew and How Khang Lim
- PONK: Tool for Client-Oriented Legal Writing in Czech – Tereza Novotná, Jan Černý, Ivan Kraus, Ivana Kvapilíková, Jiří Mírovský, Arnold Stanovský and Barbora Hladká
- Building Knowledge Graphs from Legal Documents with LLM-based Entity Recognition – Doan Thai Binh Phan, Hien Nguyen and Hung Ngo
- Visualizing Legal Documents with Multimodal Large Language Models (MLLMs) for Self-Represented Litigants (SRLs) and their Lay Legal Assistants” – Brian Rhindress
- Bridging Gaps with AI: Dismantling the Cradle-to-Prison Pipeline – Chandra Harsha Bhamidipati, Katherine Stathulis and Erin Stewart
- Building Transparent AI Systems for Justice: Criminal Reform to Legal Access – Aparna Komarla, Arman Aydemir and Nick Colvin
- AJusticeLink: Mobile AI System for Legal Risk Detection – Chlece Walker-Neal-Murray
- An AI-Based Approach to Early Reporting and Justice Initiation in Image-based Sexual Abuse. A Pilot Study – Mattia Falduti and Anca Radu
- Access to Justice for All – A Novel Roadmap – Bruce Cahan and Yen Kha
- Place Matters: Comparing LLM Hallucination Rates – Damian Curran, Vanessa Sporne, Lea Frermann and Jeannie Paterson
- Rabula Benchmark: Measuring Legal Reasoning in Brazil – Eduardo Caruso Barbosa Pacheco and Glauber Machado Pinto, along with mention of additional papers, Scaling Access to Justice through AI-Assisted Precedent Mechanisms and Automated Screening of Legal Aid and Court Fee Waiver Requests
- “Stack Overflow for Lawyers” – Tex Pasley
- Bridging the Justice Gap in Custody Disputes –Hsuan-Lei Shao and Sieh-Chuen Huang
- Designing with Dignity: A Participatory Toolkit for Ethical AI – Spriha Bhandari
The event on June 20th was a part of the broader International Conference on AI and Law. This year’s edition was hosted at Northwestern University’s Pritzker School of Law. I organized the workshop along with Margaret Hagan of the Stanford Legal Design Lab, Hannes Westermann of the Maastricht University Law and Tech Lab, Marc Lauritsen of Capstone Systems, and researcher Jaromir Savelka of Carnegie Mellon University. It was the fourth AI and Access to Justice Workshop that the Lab has participated in at an international AI and Law conference, and we hope to hold another one at the upcoming Jurix conference in Turin, Italy in December of 2025.
The first AI workshop, at ICAIL 2023 in Braga, Portugal, was organized by Karl Branting, who sadly passed away earlier this year. He was recognized at the workshop with a message from one of his closest collaborators.