By D. James Greiner, Michelle Blouin, & Eric Krebs
The OpenJustice project, now just a year since inception, has gained interest due in part to its hot topic: combining access to justice and artificial intelligence. A2J Lab Faculty Director Jim Greiner spoke with Mandy Mobley Li, Assistant Director of Research Innovations at the A2J Lab, in a recent Proof Over Precedent episode about study developments.
The project addresses whether AI allows volunteer pro bono attorneys and staff to provide legal information and advice (without a traditional attorney-client relationship) more effectively and efficiently than status quo operations, which do not use AI assistance. Researchers at Queens University’s Conflict Analytics Lab are developing the chatbot. Will AI help improve attorney/staff conversations with clients? Longer term, the question may shift: Will AI effectively support individuals in self-represented litigation?
The Staff, the Funders, and the Field Partners
The Conflict Analytics Lab at Queens University (Ontario, Canada) created the OpenJustice platform, a generative AI platform focused on law and customizable to clients’ needs. Study development began when A2J Lab Director of Research and Strategic Partnerships Renee Danser connected Mandy, a former computer scientist, with Conflict Analytics Lab Director Samuel Dahan. If all goes smoothly, field partner Pro Bono Ontario will provide the study testing ground for OpenJustice, with the case type to be determined, although attention is currently focused on landlord/tenant litigation. Canada’s Social Science and Humanities Research Council awarded the Conflict Analytics Lab an Insight Grant in April, enabling it to implement the Open Justice platform and the A2J Lab to do the evaluation.
The Setting
Business as usual in pro bono settings may involve volunteer attorneys and staff working in fields less familiar to them, sometimes requiring in-depth conversations and follow-up questions with specialists, other volunteer attorneys who have received more extensive training, or lawyers from another organization. What if AI could provide the expertise?
Generative AI platforms work by taking in information, processing it, drawing from additional background sources, and generating feedback—all based on prompts. The OpenJustice platform is customizable. Partners prime it by feeding it with rules of law and procedure based on the jurisdiction. The goal is to combine the best of a generalist AI platform with the specialized and localized application that a good lawyer can provide.
The Study
Several factors contribute to the study’s timeline. First, much like an attorney needs training when transferring to a new field of law, the OpenJustice platform will need training in the field of law chosen for the study. This customization may take up to a year. Attorneys and staff providing advice and information to clients must be trained to use the platform well. Second, the standard Institutional Review Board (IRB) process is necessary. Third, Pro Bono Ontario will need to enroll participants.
Once a client presents a matter matching the study’s legal type (perhaps landlord/tenant), the attorney or staff member will use an A2J Lab-programmed website to randomly assign the interview to one of three conditions:
- Use the OpenJustice platform to guide and supplement the conversation
- Use ChatGPT (for comparative analysis) to guide and supplement the conversation
- Proceed sans artificial intelligence
“The attorney will absolutely . . . have the ability to override anything that comes back from OpenJustice or ChatGPT if it is incorrect,” Mandy said. “[W]e do want to make note of that because that information will help us determine whether the AI is assisting the attorney or causing the attorney more work.”
A panel of experts will review a transcript of the conversation, scoring the advice/information provided on accuracy, efficacy, and time to respond. The transcript will be masked to prevent the panel of experts from knowing the conversation’s treatment condition.
The labs expect the study to take two to three years to get results.
Challenges to AI
The potential for unauthorized legal practice could present a stumbling block for greater AI adoption in the legal field. Likewise, the need for regulatory adaptation could slow its growth. In the area of access to justice, Mandy pointed out, private lawyers and law firms have been the first to embrace AI, potentially increasing the access to justice gap. And the legal services industry has sometimes resisted tech-based innovation. Having credible evidence on the utility of AI in a classic lawyer-client interaction will, hopefully, inform these debates.
Potential for AI
OpenJustice has room to grow, whether to a different jurisdiction within the same area of law, a new area of law within the same jurisdiction, or other legal aid organizations beyond Pro Bono Ontario.
The study, as it currently stands, focuses on testing OpenJustice as a tool for attorneys only. But what if it could be used as a tool for clients? The attorneys’ role in the study is to translate legal jargon into digestible insight for clients. Perhaps future iterations of OpenJustice could address clients’ need to translate that jargon into plain language, enabling more wide-spread client self-representation and, hence, more access to justice.
Like TurboTax, through which users can fill out their taxes online and if desired pull in an accountant to review them before submission, future OpenJustice or similar AI platforms may enable users to consult a lawyer for review after they’ve done most of the legwork.
“That is the situation today,” Mandy said. “But tomorrow, I believe the goal is to make the prompting so easy that a 5-year-old could ask a question and get back an accurate legal response.”
If that happens, the next study may be whether AI outperforms legal professionals altogether.
—–
If you’re interested in more details of this project, listen to our Proof Over Precedent podcast episode on the topic.