In The Economist, a segment entitled "AI am the law" explores the impact of "smart software" in use in Australia that predicts whether clients have winnable cases in court and assesses applicants likelihood of receiving legal aid. In tests to date the software simulations are accurate 98% of the time in predicting the actual outcomes.
"What the systems still lack is the ability to exercise discretion, and that is not likely to change for the foreseeable future," the author muses.
But these tasks seem ripe for automation, or machine migration. Why send someone through the cost and effort of an application process or court proceedings if the outcomes are so predictable as can be mapped beforehand for reference. The assumption is that the discretion previously used by the gatekeepers to legal aid and court dates was frivolous and surplus. All they were really doing was exercising their right to either pass someone into the process or exclude them. So the starting line for machines in the legal advice industry just got one step closer to the claimants. The discretionary task of passing people on to the next step is gone. Could this herald a threat to all human professions with a corresponding shrinkage in the definitions of when discretion is a necessary element of a decision process.
Sitting in a Holborn Street cafe with Milverton Wallace (if ever there was a man who needed a blog), one of the superstars of the British digerati, he shook his head mightily: "You're talking about any professional activity. Perhaps. Anything without discretion."
But is there a danger that as discretion becomes less prevalent, hunted wherever it lies by thirsty software entrepreneurs and business process re-engineering exercises, that it could face virtual extinction, no pun intended. A good management consultant can process map anything: "Tell me about what you do and how you do it?". The question frames the instrumental terms. Break it down for me baby! I'll draw the charts. Later the charts feed into tables that feed into objects and algorithms whirl. Poof! We found out that we can automate what you do. Surprised?
Lawyers are undoubtedly caught off-guard by the technological developments in their industry, susceptible to early adoption by virtue of the structured syntax and tasks that constitute the majority of their professional activity. These are very early days in the intelligent legal systems game, as it is for virtually every industry, so it is difficult to predict where the discretion line will be drawn in the sand. Anyone for a computer judge?
Regardless we are all familiar with that feeling we get when our fate seems to hang in the hands of another. When for that moment in time it seems as if all hangs on the outcome of the decision. Most can recite experiences with a subjectively villainous or empathetic policeman, parking attendant, secretary, customer, or other. "He really did me a favour!" or "I was like, why is he screwing me?" That's the smoking gun, or at least the puff of smoke that comes with discretion. And life is richer with some of it in play.
Think of the golfer who decides to play the aggressive shot from the bunker even though he knows he is 98% likely to either top or pluck the ball. Or the shopper who, rattled with indecision, decides to buy both pairs of shoes.
More seriously think of the value of "the ability or power to decide responsibly" or the "freedom to act or judge on one's own", both dictionary definitions of discretion. How will decisions on where to use systems that implicitly have no discretion be made? How will we know if conservation services are needed due to the imminent extinction of discretion in professional life. Surely no one is immune. How easy would it be to create the perfect CEO by developing working models from studying all of the decisions of previous successful CEOs? Like Deep Blue beat Kasparov so one day will CEO SIMULATOR X wag a virtual finger menacingly at a reticent Donald Trump: "Your numbers stink. You're fired!"
See this Media Project post for an Introduction to Artificial Intelligence
Labels: computing
