By Kyle Mittan,
If you died tomorrow, would all your affairs be in order, laid out in a will?
If you answered no, you’re not alone. According to a 2021 survey by the analytics firm Gallup, only 46% of Americans have a will, and other surveys suggest it could be less. The reasons for this vary, experts say, but it often comes down to the fact that it’s expensive and time-consuming to hire a lawyer to help draw up the document.
A team of University of Arizona researchers made up of experts in artificial intelligence, law and computer science, has an idea to make creating and executing wills easier: Let software do it.
Clayton Morrison, an associate professor in the School of Information in the College of Social and Behavioral Sciences is principal investigator of a new project designed to make that possible. The project also includes Derek Bambauer, a professor in the James E. Rogers College of Law, and professor Saumya Debray and associate professor Mihai Surdeanu, both from the Department of Computer Science in the College of Science.
The effort is being funded by $750,000 from the National Science Foundation’s Designing Accountable Software Systems program.
“If we can build systems that can help people plan their own affairs more accurately and give them certainty and comfort that when they’re gone, the things that they’ve gathered over a lifetime will go to the people and places that they want, that seems like an enormous achievement for technology,” Bambauer said.
Anatomy of a will
When a will is created, a person decides how to organize their affairs when they die, including who gets their personal belongings or who will take care of their children.
People typically work with attorneys to draw up a will, following laws that vary by state, to ensure the document accurately represents their wishes. As part of that process, they must prove that they are legally competent and not being forced into the agreement. A witness also is required to sign off on the will.
In most cases, Bambauer said, wills sit in drawers for years until someone dies. At that point, the will enters a process called probate, when a family goes to court to execute it. Most of the time, Bambauer said, the court does not need to get involved unless someone challenges the terms of the will.
Wills may lend themselves well to software automation because they often follow templates, said Bambauer, who worked in the computer science field as a systems engineer before his legal career.
A quest to make software more accountable
The NSF’s Designing Accountable Software Systems program supports projects that create “accountable” software, defined by the NSF as software with three key traits:
- It follows the rules society has set, such as laws and social norms.
- It can adapt to those laws and other rules as they change.
- Average people – not just software programmers – can look under the hood and understand what the software is designed to do.
“We’re trying to see if we can get computers to understand the natural language that’s used to describe legal documents, like wills, and have our understanding be translated into code that could be executed,” Morrison said. “We also want to make sure we understand what was compiled into that executable code and be able to ask, ‘Does this accurately represent what was intended, and how does it stand relative to the existing law?'”
Accountable software designed to write and execute wills, Morrison added, would first need to be able to take the terms of a will in plain language and turn them into computer code. Then, when it’s time to execute the will, the software would need to be able to automatically verify the will’s terms against current law to ensure that it remained legally binding.
“At minimum, what the system could do is identify issues that could potentially be a problem by comparing the text of a will to the probate code,” Bambauer said. “That might help out people drafting their own wills, or even attorneys, because attorneys make mistakes.”
In a best-case scenario, Bambauer said, the software would not only identify issues with a will – it would fix them, too, by rewriting the will as laws or facts change.
“You can imagine a world where a state makes this available – a will-checking service – or an attorney’s office has a software service that does it,” Bambauer said.
Translating legalese to ones and zeroes
Of course, wills are typically written in a language better suited for humans.
Teaching a machine to understand the language of a document such as a will often involves feeding the machine many examples of a type of document so it can learn to understand them and represent them in code, said Surdeanu, the team’s natural language processing expert.
The researchers have already done this with hundreds of actual wills, taken from public records in Tennessee and Idaho. They now have a database that takes thousands of legal statements from those wills and matches them with corresponding computer code. A previous award of $55,000 from the UArizona Provost’s Investment Fund provided the funding for this database, and helped the team earn the larger NSF award.
Even with a clear vision of their final product in mind, the researchers have many unanswered questions, including: How will this software track down the information necessary to ensure a will is legally valid? When will technology provide the infrastructure to prove, in a digital format, that physical items have been transferred to the right person?
When the researchers set out to define the scope of the project, these hard-to-answer questions were what got all four researchers on board.
“It was just fun,” computer scientist Debray said. “Who wants to solve easy problems?”
The work has also included three graduate students: Alice Kwang, a doctoral student in linguistics; Guy Forte, a third-year student in the College of Law; and Jacob Israelsen, who graduated with a law degree in May.
Ali Bridges, interim assistant dean for external communications and marketing in the James E. Rogers College of Law, contributed reporting to this article.