The Pew Charitable Trusts is turning to Suffolk University Law School’s Legal Innovation & Technology (LIT) Lab to take on a challenging problem: people in real legal need who find their online searches for advice end up with search results that are neither relevant nor helpful. The LIT Lab and Stanford Law’s Legal Design Lab invented a mobile game, Learned Hands, to help address the problem.

Suffolk’s Lab, powered by a group of law students and their director David Colarusso, assists non-profits, law firms, courts, and government offices with data science and legal technology projects.

Learned Hands asks lawyers and non-lawyers alike to label real people’s legal questions. The LIT Lab and Stanford teams collected thousands of such questions: “I got kicked out of my apartment because of my dog. What can I do?” for example. Players label the questions, assigning them to discrete areas of the law, from housing to family law and bankruptcy.

Why do all of this labeling? Machines struggle to understand the context of speech, says Ericka Rickard, leader of Pew’s new Civil Justice Innovation Project and a teacher in Suffolk Law’s online Legal Innovation & Technology Certificate Program.

A human would understand the meaning of “kicked out”, she says, but machines generally wouldn’t, and need to be trained to understand that the person actually was seeking information about eviction law.

The game will do that training. Crowdsourced labels provided by the game’s players will teach an artificial intelligence (AI) tool to make sense of human speech, Rickard explains. Lots of correct labels will improve the results of online searches—and thus the legal advice available to everyone on the web. She wrote about the partnership with Suffolk Law in a Jan. 24 article for the Pew website, “How Artificial Intelligence Could Improve Access to Legal Information."

What happens if some of the players mislabel a question? Once enough players are in agreement about the answer, the outliers getting the label wrong are disregarded by the game’s algorithms.
Once there are AI tools that better understand the legal questions humans are asking, says Colarusso, the long-term goal is to connect low and moderate income people with the legal services they need, be it an attorney in the right practice area or particularly relevant resources on a court service center website.

When building these tools, it’s important to ask if they’re an improvement over the existing solution, he says. For too many seeking legal advice, he adds, the existing solution is no help at all. “Our goal isn’t to build a robot lawyer, but tools that can help the public better navigate their legal issues,” he says.