Bot Challenges Lawyers To Beat It At Predicting Case Outcomes

The UK company Elexirr (which rebranded in July from LawBot) has thrown down the gauntlet, challenging lawyers to a competition to determine whether its bot can better predict case outcomes than can they.

They are calling it the Elexirr Lawyer Challenge and they describe it as the first event in history to directly pit machines against lawyers. More than 50 UK lawyers have confirmed their participation and will use the same information as the Elexirr system to predict the outcome of complaints alleging mis-selling of payment protection insurance (a type of credit insurance).

Rebecca Agliolo, Elexirr’s marketing director, says that the company has confirmed Magic Circle partners, barristers, and in-house counsel for the challenge.

As I reported here in July, Elexirr — which gained notoriety for developing a legal chatbot to help crime victims — launched a new bot that predicts whether a legal claim would be won or lost if it went to court. The bot predicts the outcome of a claim with 71 percent accuracy, Ludwig Bull, managing director, told me at the time, and he hoped further development would enable the bot to achieve 80-90 percent accuracy.

Here’s how it will work:

From Oct. 20 to Oct. 27, participants will login to a website with facts of actual PPI mis-selling complaints received by the Financial Ombudsman Service. They will predict whether the complaint was upheld or rejected. The Elexirr system will receive the same facts and make its own predictions. The side with the higher accuracy wins the competition.

A technology judge and a legal judge will verify the fairness of the competition. Ian Dodd, UK director of Premonition, will announce the verified results during an event on Oct. 27 at Kennedys Law Firm in London. A second, public event the same day at the Google campus in London will showcase the Elexirr system.

Lawyers interested in taking part in the challenge may register here.

And may the best person … errr … intelligent entity win!