Artificial Intelligence Predicts Case Decisions

There are several reports stating that a company managed to use their Artificial Intelligence bot to predict Human Rights cases in the European Court of Justice with up to 79% accuracy.

This percentage may seem high but with 79% accuracy is still quite far from the prediction success rate of a highly skilled and experienced lawyers.

Although one may argue that the speed of learning by an AI bot is faster than an average human being (I am still sceptical with this claim), it is highly unlikely that AI bot can succeed beyond what a highly skilled and experienced lawyer can do.

First of all, it is not uncommon for Judges and Courts to change their minds on a certain issue of law and we have seen in many cases where this had been done.

One classic case is the case of R v R [1991] UKHL 12 where the House of Lords determined that under English criminal law it is possible for a husband to rape his wife. In 1990, the defendant, R , had been convicted of attempting to rape his wife. He appealed the conviction on the grounds of a purported marital rape exemption under common law. R claimed that it was not legally possible for a husband to rape his wife, as the wife had given irrevocable consent to sexual intercourse with her husband through the contract of marriage, which she could not subsequently withdraw. Both the Court of Appeal and subsequently the House of Lords upheld the rape conviction, declaring that a marital rape exemption did not exist in English law.

Now if the AI bot have been fed with all the cases prior to this case, which probably made up to more number of cases than the ones after this case, then the AI bot may have a different type of statistics. So to speak – garbage in, garbage out.

The second reason is there is just not enough data on published court cases as compared to other types of data such as search data. There are only so many judgments being delivered by a single judge in his or her lifetime. And if the AI bot is trying to predict outcomes of decisions by younger judges, it will probably be even more difficult. It is impossible for a Judge to have delivered a million or even 100,000 cases in his or her lifetime. So it will be an uphill task trying to increase the accuracy rate of the AI bot in predicting court decisions.

The third reason is court’s decisions are not all based on logical conclusions. There are a lot of human elements in it – emotional elements as well. It is very different from playing chess where a logical AI bot can beat a human being player.

In conclusion, I am of the view that it is very difficult for an AI bot to beat a highly skilled and experienced lawyer’s prediction of a Court’s decision.