Predicting Human Behavior
Predicting Human Behavior

Predicting human behavior outcomes through artificial intelligence (AI) is tricky business, as any science fiction movie buff will tell you.  And that should be doubly true for humans in positions of technical authority, say, like the Supreme Court (SCOTUS). However, it appears that AI is now substantially exceeding the success rate of the top-notch legal experts at determining the ruling of the court on matters of the judiciary.

Because of the substantial financial and time commitments entailed in a Supreme Court case, many lawyers and businessmen want to have some prediction as to how SCOTUS will rule prior to investing heavily in getting heard. This makes sense, of course, and it’s led to the proliferation of prediction methodologies.

One such system is the tried and true method of using legal experts. These fellows are top of the line legal scholars who have substantial knowledge of legal precedence and the voting patterns and political leanings of each of the judges on the court. They’ve been successful in predicting human behavior outcomes 66% of the time. Another option has been differing algorithms using small data sets from previous decisions. These systems have had some accuracy as well, predicting court outcomes with about the same precision as the legal analysts.

Recently, though, a group of AI specialists used the Supreme Court Database, containing case files all the way back to 1791, to build an algorithm to predict any specific justice’s vote on a case. They used a huge amount of data, inputting 16 individual features into every vote – things like who the justice was, the issue and the court of origin. For the years from 1816 until 2015, they created a machine-learning algorithm. The ‘random forest,’ as it’s called, looked at all the years prior and evaluated cases and their outcomes. After using the past years’ decisions as guidance, the algorithm predicted the outcomes of cases for its specific year. After the predictions the actual decisions were input, allowing the AI to update it’s base strategy. For the period in question, the algorithm was able to predict outcomes more than 70% of the time, and the individual justices’ decisions almost 72% of the time. AI blew the experts out of the water.

Obviously there will likely be some business implications for this sort of tech. For example, those seeking to take a case to the highest court could decide whether it was worth it or not, thereby saving the substantial costs involved with attorneys of this caliber. And even more interestingly, lawyers could use the algorithm to help them decide how to best argue the case, based on the judges who are currently sitting on the court. Finally, the stock market should have a hay day with this sort of information, as they would be able to buy or short stocks based on the likelihood of an affirmative decision. Regardless of how you slice it, information is power, and the team involved has promised to continue inputting information including the entire oral argument for each case, thus providing even more data for the AI system.

Read more about predicting human behavior here.

LEAVE A REPLY

Please enter your comment!
Please enter your name here