The future of justice is Watson

Part 1 of 2

Client Executive, IBM

Imagine you are a judge at some point in the future. You face writing a decision, something you’ve done hundreds, maybe thousands, of times. The facts are clear or muddled as the case may be. But you’re convinced that your decision hinges on one critical point of law. Your line of reasoning is clear, except for that one troubling, elusive fact. What do you do? Call IBM Watson.

Or, imagine you are a single mother, and you’re doing the best that you can. The community center has been a lifeline, helping you apply for a job through a company website and offering free checkups for your baby. You had an interview and it went well. If you get the job, you will have almost enough income for daycare. If only the father would help. He’s got a good job, but his child support has been sporadic. You know there is something you can do, but where do you even start? Call Watson.

Consider one more scenario. Your business has been very successful since you opened last year. Everything was going great until you received a shipment of bad filters. Now you have customers calling to complain, warranty service claims have gone through the roof and, worst of all, after you tried to talk about some kind of cost sharing, the supplier is threatening to sue. You’re meeting with your partners to lay out the options and discuss just how much this situation could damage the bottom line, but first there’s one more thing you need to do. Call Watson.

Wanted: A multitalented legal assistant

The subjects in all three of the foregoing scenarios could really use a talented, industrious legal assistant. They need a self-starter able to multitask thousands of research requests simultaneously and respond within fractions of a second to general legal questions and specific requests for precedent cases and statutes. And if the ideal candidate is fluent in Arabic, Chinese, English, Portuguese, Spanish and many other languages, all the better.

Those qualifications may seem like an extremely tall order, but one candidate is up to the task. IBM Watson is a cognitive computing system that uses sophisticated language-recognition software and machine learning to instantly understand questions and respond with the correct answer. Watson can listen to a question, recognizing words regardless of accents or less-than-perfect audio quality, and then parse the words contained in the question to draw out subtle context and meanings.

Watson then drills down through thousands or even millions of documents looking for patterns and begins to rank the results. It then looks for evidence to either confirm or refute the results and adjusts its rankings until it arrives at the best answer to the question. A cognitive game show contestant

In the sheltered world of big data and machine learning experts, Watson had been big news for some time before it was finally introduced to the world. On 14 February 2011, an avatar representing Watson appeared on a special exhibition of the game show Jeopardy!, where Watson challenged two of the game show’s greatest champions in its history. Over two nights the computer systematically overpowered its human competitors.

Watson’s victory for IBM and the system’s architects was huge. More importantly, it marked a paradigm-shifting demonstration of what was possible when machine learning and intuitive software were unleashed on mountains of data.

Watson is quite impressive, but it can make mistakes. For example, when answering a question under the category, US cities, on that infamous episode of Jeopardy!, Watson responded, “What is, Toronto?” Many factors—both in the structure of the question and in Watson’s programming—required it to respond even when the confidence level was far below the minimum threshold. Despite this gaff, several aspects of how Watson makes mistakes point to its potential in the judicial system.

Capability: A future legal career

When Watson does make errors, a human user or system operator can easily recognize them to be absurd. In addition, its errors are random. Watson is not capable of making systematic errors based on political ideology, gender, upbringing or any number of factors that can creep into legal decisions made by humans.

And Watson is continually learning to give better answers. With each question and answer, Watson is able to adjust its algorithm based on feedback from a human operator—helping to prevent mistakes from happening again. In the early stages of Watson’s development, human operators trained the system, teaching it the specific language, basic principles and rules of law. From there, Watson learned by example. The more data, documents and questions that were thrown at it, the more Watson learned. And while Watson won’t replace judges or lawyers anytime soon, its potential to serve a legal assistance role in judicial scenarios is invaluable.

Are there other scenarios where you think Watson can be helpful in supporting judicial services? Tell us what you think in the comments. And learn how analytics in government settings have the potential to improve the efficiency of the judicial system.