The ideal judge is completely rational and free of consistent bias, making their judicial decisions completely unpredictable. But the decisions of Supreme Court justices in real life may actually be quite predictable, according to a new study published in the Nov. 9 issue of the online journal PLoS ONE.

For the study, Roger Guimera and Marta Sales-Pardo created a computational model to investigate whether it would be possible to predict a justice's vote based on the other justice's votes in the same case.

"Here we have studied to what extent can one predict the vote of a justice from the votes of other justices in the same case (and the track record of the court)," they explained in their study. "Our approach thus focuses in the stable correlations between justice behaviors, which, we argue, reflect consistent attitudes towards the law."

The study was based on 150 cases of each of the courts from the first Warren Court in 1953 to the last Rehnquist Court in 2004.

Guimera and Sales-Pardo found that their model was more accurate than predictions made by legal experts or algorithms that take the content of the cases into account.

"We have found that justices are significantly more predictable than one would expect from an ideal situation in which justice decisions are uncorrelated," they said.

These results are "something to keep in mind when discussing how legal decisions are made", Guimera and Sales-Pardo added.

The study also found that justice predictability has been consistently and significantly lower during Democratic presidencies than Republican, and that justice predictability has decreased over time.

Illustration: Flickr user DonkeyHotey.