Daily Archives: November 22, 2017

AI Can Certainly Identify Relevant Case Law, But Can It Predict Judge’s Decisions?

We are all familiar with the use of AI in contract (lifecycle) management to scan contracts, decompose them into clauses, identify the type of clause, index the contract against the clause types, and identify entity and attribute types in those clauses. This is common practice for best-of-breed CLM providers like Icertis and Exari and standalone best-of-breed contract analytic providers like Counselytics and Seal Software.

But not all of these are designed to evaluate a contract, and more precisely, a contract dispute against existing law and case law to determine the likelihood of a ruling in a party’s favour. But that is apparently what the new AI entrants are shooting for.

As per this article in The Globe and Mail, a new AI entrant is offering simulation software that predicts how a court might rule in a given case involving tax law. The software, which takes client details, arguments, previous case law, and current law into account, crunches this, and other inputs, and delivers a simulated judgement claimed to be 90% accurate.

And now the company in question is turning its attention to labour law and hoping to offer a product soon that will offer the same type of predictive accuracy in this legal area. This will obviously be more difficult than tax as there are more grey areas as to what is a violation, but it may be achievable.

But is 90% accuracy enough? That’s 1 in 10 wrong predictions. But maybe it’s enough. If the software predicts that the plaintiff will lose, with 90% or more reliability, a case that could easily cost hundreds of thousands of dollars, and the damages for pleading guilty are less than a million, the right legal strategy would obviously be to seek out a good deal and not risk losing it all in court.

Plus, if both sides of a pending legal battle had access to simulation software that always showed one side losing with significant confidence, then at least one side will always be up for negotiation, and that’s almost always cheaper than a costly court case. Plus, it gets a resolution quicker, and sometimes the biggest cost of a dispute is the lost time.

So while it will never be truly accurate, as you can’t truly predict how a judge will rule in any given situation (as a human interpretation of a grey area can always be different than what might be the typical interpretation), this might just be the software that lawyers, tax accountants, and, soon, HR professionals need to come to their senses and resolve the dispute before wasting time in court.

And if it works in these areas, maybe soon it can ingest global law and treaties and be applied to Procurement disputes …