Remember the 18-year-old who created that awesome website last summer to help people appeal unfair parking tickets?
Well he’s back, and this time he’s gone one step further and actually created a robot lawyer to help the public deal with a whole range of different legal issues.
“Immediately after you wrote about DoNotPay, users began sending me emails ranging from questions about how to use the site (‘I got a parking ticket — which appeal should I choose?’) to general questions about consumer law (‘What happens if I can’t pay my court fine?’),” wrote Joshua Browder, who is now 19 and studying Economics and Computer Science at Stanford University, in an email to Mashable.
“Although I tried to respond to every single one, as the site gained popularity, it became harder to respond in detail to thousands of these emails a month,” he said. “To solve this problem, I realised that the best way to help people would be to create a computer program that could talk to users, generate appeals and answer questions like a human.”
Enter the robot lawyer.
Basically, the robot works by asking the user a series of questions about their situation. When it has enough information — and assuming the person does have legal grounds for an appeal — the robot will generate a letter that the person can use.
“The robot can currently handle parking ticket appeals, payment protection insurance (PPI) claims and delayed flights/trains,” Browder explained. “It can also answer some general legal questions (“I can’t afford my ticket. What do I do?”). I am ultimately looking to give it as much functionality as possible in the spirit of trying to replace the large group of exploitative lawyers.”
When it came to actually developing the robot, Browder told us it was a lot trickier than the website.
“I was fortunate to receive some advice from my professors,” he said.
“Unlike the creation of the basic site, which took three weeks, making the robot seemed impossible.”
“Initially, I thought the best way to go about it was to create lots of individual rules for it to follow (e.g if one says ‘I got a ticket’ the robot says ‘What was wrong with your ticket?’),” Browder said. “However, I quickly failed with this approach because there are thousands of ways to say the same thing and it would be impossible to catch every one.”
“The breakthrough came when I learned how to create a way for the robot to learn and compare phrases itself, so that it doesn’t matter how the user phrases his or her requests,” he said.
Browder went on to explain that the robot uses a text comparison which includes keywords, word order and pronouns. The more people who use the robot, the more the algorithm improves. Like his DoNotPay website, the information in the appeals generated by the robot is backed up by Freedom of Information data and legal advice gathered by Browder.
“If the robot can’t answer, it provides a generic and helpful message offering the user some sample phrases or the option of contacting me directly,” he said. “On the backend, whenever the robot can’t answer, I get notified and I work as quickly as possible to add functionality for any future requests of a similar nature.”