Man beats machine
Earlier in the week the Financial Times’ (£) ‘Big Read’ ran a familiar scare story predicting the legal profession was ripe for ‘Uberisation’. If the nation’s cabbies could be so readily outmanoueuvred, surely it was only a matter of time before the lawyers were killed off.
The usual suspects were rolled out — the Susskinds et al. to say the usual things: Lawyers are toast making for a better big read than lawyers continue to prosper as have done since Magna Carta. It is doubtful that the top QCs and magic circle partners raised an eyebrow of concern — how can we be undone by desiccated calculating machines when we are ourselves desiccated calculating machines?
Good question. But might it be possible that even though man could beat machine on interpreting the law, machine might beat man in predicting how judges might decide the law. And if you could predict your chances of winning a case then you would be at a significant advantage in calculating how much to spend in pursuing that case.
In a footnote to the piece reference was made to an experiment over a decade ago in which a machine beat a panel of experts in predicting US Supreme Court decisions racking up a score of nearly 70%. The implication being that Moore’s Law being what it is if computers were that good then they must be exceptional now. Oddly, though, computers have hardly improved at all and only won because the experts — a random bunch of academics — were not as expert as had been presumed, scoring under 60%.
Satisfyingly, when the state of the art legal predictor {Marshall}+ algorithm entered into an online Supreme Court fantasy league last year it was bested by the wisdom of the human crowd. The group was better than the computer.
And really satisfyingly the best predictor within the group, the only three time winner of the competition, the only person to score over 80% regularly is a thirty something bloke who lives in Queens and has no legal training. He plays under the name Melech — ‘king’ in Hebrew — and his real name is Jacob Berlove. Since he picked up a constitutional law text book when he was at high school and read it cover to cover Berlove has taken an interest in the Supreme Court. He says:
As someone long interested in Supreme Court cases, it was very frustrating to be unable to do anything substantial in relation to this interest, given that I will never serve on the Supreme Court. I had long felt that I might have a pretty good understanding and feel for the Justices’ thought processes and decided to set out to test just how much I did.
He has certainly passed that test. Berlove puts his success down to trying to get into the minds of the justices and getting down to the essence of what a case is about. He also leaves his own beliefs to one side and remains dispassionate:
It doesn’t matter what I think about it; it matters what those nine people think about it.
Simple, really. But, apparently, too complex for the machine.