If we want to meet our society’s need for legal help at their most vulnerable times, and not only the needs of our wealthiest or largest corporations, we need to apply technology to the problem. One of the technologies that deserves more investment is the humble expert system. Expert systems, step-by-step wizard-like questionnaires that deliver advice tailored to the user’s situation, are not new. A lot of modern AI talk focuses on statistical learning approaches. Yet expert systems, which were part of the first wave of AI and date to the 1970s, have a lot of life left in them, especially when it comes to the law. See Susskind, Richard (1986). Expert Systems in Law: A Jurisprudential Approach to Artificial Intelligence and Legal Reasoning. Modern Law Review. 49 (2): 168–194.
The Legal Services Corporation survey showed that our current legal system is only helping meet about 14% of the needs of the poor. Legal Services Corporation, Justice Gap Report: Measuring the Civil Legal Needs of Low-income Americans. 2017. Available online at https://www.lsc.gov/media-center/publications/2017-justice-gap-report.
An astounding, overwhelming 86% of those needs are never met. Many more people in the large middle class who don’t qualify for legal aid are still unable to hire a lawyer.
What does that mean? The biggest areas of need are in housing and family law. Right at the top of Maslow’s hierarchy of needs. But the needs are diverse. This means tenants losing their homes, victims of domestic violence being kept in harm’s way. Veterans losing benefits, persons with disabilities being left without financial support or in vulnerable situations.
I have worked in legal aid for almost 12 years, and more resources for legal aid would certainly help, but increasing by a factor of 7 might not be realistic. Part of the solution might be more lawyers, or simply more pro bono. James D. Abrams and Ann Hancock – February 14, 2017. The Justice Gap and Pro Bono Legal.
But others have made compelling arguments that pro bono alone cannot be the solution. The problem is simply too vast. See, e.g., Gillian Hadfield. Why legal aid and pro bono can never solve the access to justice problem. (2016). Helping everyone who currently can’t afford a lawyer with one-on-one assistance would require a vast number of new lawyers, and still might not meet the affordability test.
What’s more, we need to recognize that one-to-one legal assistance by an attorney might not be what every person wants or needs. If the rich are receiving the equivalent of tailored suits, perhaps some less well-off members of society are left unclothed for want of someone who can deliver a quality mass-manufactured shirt. Many people who receive no legal assistance right now need relatively simple legal advice or slightly customized forms.
Each week for years I have sat at Rosie’s Place, a women’s shelter in Boston, taking the cases that come in one at a time and offering the advice that I can. My role as an attorney at Greater Boston Legal Services focuses on housing law advice, but the questions that came in through Rosie’s Place often are hugely varied. Yet, a few minutes of research, a kind ear, and sometimes a letter, a call, or a bit of advice are enough to leave the client satisfied and happy at the end of the interaction. As I participated in this clinic over the years, I realized that one form that was really too long to be completed in these brief sessions could be automated in a way that would serve our clients better.
Evictions in Massachusetts are civil cases with many tenant protections. This has led to a long checkbox Answer form, with more than 180 questions to cover all of the more likely defenses and counterclaims a tenant might have. Greater Boston Legal Services runs a weekly clinic that generally takes about 4 hours to help a tenant through the forms. One-on-one, a lawyer still takes at least an hour to fill out the paperwork by hand.
In response to this long, complex form and clinic process, I created Massachusetts Defense for Eviction, or MADE (gbls.org/MADE), a guided online expert system that walks a tenant through the forms in as little as 20 to 30 minutes. That made it possible for me to use in our Rosie’s Place clinics. And it’s allowed us to cut down the length of our group clinics to an hour and a half without making the forms less protective of tenants’ rights. Tenants can work at their own pace; the questions are translated into 5 languages and target a 6th grade reading level. Tenants also receive automated reminders of follow-up forms and court dates by text message.
MADE is now helping an average of 200 people a month respond to an eviction case in court. Almost all of those are tenants who GBLS would never have seen. By filing a response, each tenant gains at least two weeks to work out a solution with their landlord. Beginning in March 2020, I will be starting a new position at the Suffolk University Law School’s Legal Innovation and Technology Lab. MADE will continue to grow and perhaps serve new states through my work there.
Tenants have used MADE at all hours of the night and in many different circumstances. One tenant literally leaves their home only with the help of an ambulance. A social worker with a cellphone was all it took to allow them to deliver their answer to court and access a reasonable accommodation for their disability. Another tenant, the first to use MADE to submit an online intake to GBLS, finished their Answer at 12:30 AM. We didn’t even know who we were leaving unhelped before we had this new way to give them assistance.
Rohan Pavuluri, a Harvard College undergraduate student, joined Harvard Law professor Jim Greiner’s class and learned about an odd catch-22 in bankruptcy: millions of Americans are too poor to declare bankruptcy. With the average bankruptcy costing more than $1,000, he helped found Upsolve (upsolve.org), a non-profit that uses technology to automate filling out most of the 100s of pages of bankruptcy forms on their own and allow debtors to declare bankruptcy for free. A lawyer comes in at various points to help answer questions and review the final forms before they are submitted to the bankruptcy court.
All across the legal space, there is room for expert systems like MADE and Upsolve to help expand access to justice. These systems take the knowledge embedded in a lawyer’s years of practice and distill it into a series of questions and answers. By no means are expert systems free to create. They take painstaking work and validation before they can be unleashed on the world. Just like a good lawyer, an expert system needs to be kept up to date with changes in the law and respond to feedback to improve its delivery of information. But millions of Americans will benefit from the help of an expert system when they are participating in a mostly static but confusing to navigate area of law.
Expert systems are not risk free. Expert systems are qualitatively different from the old form books, or other lay law books aimed at non-attorneys. Those systems try to translate the law into plain text, teach a person what that law says, and then asks them to apply the law to their own facts. Expert systems can guide a person to the correct application of the law in a realistic way that form books never can. They are both safer and more dangerous for this reason. Expert systems provide the appearance of certainty. Yet, they can also be tested in a rigorous, ongoing way that one-on-one legal assistance never can. Standards, with adequate disclaimers, mitigate this risk. Marc Lauritsen and Quinten Steenhuis. 2019. Substantive Legal Software Quality: A Gathering Storm?. In Proceedings of the Seventeenth International Conference on Artificial Intelligence and Law (ICAIL ’19). ACM, New York, NY, USA, 52-62.
As at attorney at Greater Boston Legal Services, and recently as an independent consultant, I have tried to apply expert systems and technology to automate not only my own work, but to help my fellow attorneys. I’ve also opened up the door to train other attorneys in the basics of building their own expert systems. Some of those attorneys have been inside Greater Boston Legal Services, and some have joined virtual training sessions from around the world using the wonders of video conferencing. Anybody can learn to create an expert system; I have helped dozens of non-programmers achieve that goal with Docassemble (docassemble.org), the platform that powers both MADE and Upsolve. As this community grows, we all need to work together to share our best practices and ensure we continue to meet substantive quality standards.
If expert systems have existed since the 1970s, why have they not taken over the law yet? Putting regulatory issues aside, the truth is that the old expert systems, locked up in a physical computer that was too expensive for a legal aid office to own, can’t compare to the modern, web-based expert systems that every consumer can access on their smartphones. We also have a new generation of consumers who are familiar with computer-based wizards and web forms and have come to expect simple guided help on the web. We’re poised for a chance at a very different wave of consumer-friendly legal help.
MADE has been embraced by the Supreme Judicial Court in regulatorily-cautious Massachusetts. See Ruth Adjartey v. Central Division of the Housing Court Department, 481 Mass. 830 n. 12 (2018). Upsolve too has seen success around the country. Yet regulatory change, including reducing the risk of running into unauthorized practice of law, will be an important part of any expert system revolution. Those changes might also trickle down into allowing the provision of more types of assistance by non-attorneys, including perhaps limited technicians who have the assistance of an expert system.
Artificial intelligence may be a bit of a misnomer when it comes to expert systems. I see their role as taking the very human intelligence of a lawyer, coaxing that knowledge into a form that covers all of the possible scenarios, and translating it into computer code. Yet the more we can repeat this process, the more easily we can deliver that very human intelligence at a scale that a human could never achieve on their own.