Predicting US Supreme Court Justice Votes

New submitter Pierre Bezukhov writes "Researchers Roger Guimera and Marta Sales-Pardo of Spain set out to ask whether one of the nine Supreme Court justices could be plucked from the bench and replaced with an algorithm that does not take into account the law or the case at issue, but does take into account the other justices' votes and the court's record. These researchers say their computational models, using methods developed to analyze complex social networks, are just as accurate in predicting a justice's decision as forecasts from legal experts. 'We find that Supreme Court justices are significantly more predictable than one would expect from "ideally independent" justices in "ideal courts,"' that is, free agents independently evaluating cases on their merits, free of ideology, the study said."

Comments

Re:A really good judge

[icebike]

I wish a law clerk had written the last sentence of the summary.
It might have had a chance of being intelligible.

I don't get it...

[adamchou]

Algorithms like this have to be modeled after the historical decisions that the justices decided upon. So of course they accurately "predict" the historical decisions. So how do they know how accurate these things are for future decisions? I couldn't RTFA because the damn article isn't loading on my crappy government Internet connection.

Re:I don't get it...

[PhrostyMcByte]

Typically when doing this kind of statistical analysis, one uses half the data for training and half the data for accuracy tests.

I haven't RTFA though, so I don't know what they've done.

Re:I don't get it...

[Jane Q. Public]

"So of course they accurately "predict" the historical decisions."

In tests of this nature, the way such algorithms are tested is by giving them a subset of historical data (ideally chosen randomly), then seeing if the program can predict outcomes from historical situations that were not included in the "learning" data.

So your objection has little merit.

Courts are supposed to be predictable

[Hentes]

The whole point of the legal system is that courts don1t just hand out verdicts randomly, but according to the laws available for everyone. Without that, people wouldn't know how to act legally. In fact, ideally one should be able to predict what a courts decision would be in every situation. unfortunately, the legal systems of the world are too complicated/contradictory for that.

Re:Courts are supposed to be predictable

[Attila Dimedici]

Just look at the Bush v. Gore decision where they decided that one man one vote meant that enough votes were left untallied to effectively hand the election to Bush because they were concerned that too much attention was being paid to those particular ballots.

That is a complete misrepresentation of the Supreme Court decision. The U.S. Supreme Court ruled that the Equal Protection clause was violated by the recount called for by the Florida Supreme Court and that there was not enough time to conduct a recount that met the Equal Protection clause. A significant factor in their decision was that the end result of any Constitutional decision by them would be the same as the result of their ruling. If the Supreme Court had not ruled as it had, it is probable that the Florida legislature would have appointed a group of Electors who would have voted for Bush (according to the U.S. Constitution, each state legislature decides how that state's Electors are chosen). If that had happened it is quite possible that the end result would have been that there were no Electors from the state of Florida (there are several other scenarios in which this would have happened). In which case, the election would have been decided by the U.S. House of Representatives, who would have chosen Bush. Finally, the fact of the matter is that several groups that consistently favor the Democratic Party counted the ballots in Florida later and found that under no reasonable standard would Gore have won the state.

This is informative how?

[Hartree]

I used to work with a group that simulated the folding of proteins.

You'd take an assortment of protein sequences and train a neural net on how they folded. Then you try to use that to predict the folding of another different protein that wasn't in the set you trained it on.

But, in this case, they don't try to predict the behavior of an independent case, they use it to predict the behavior of one of the 8 items (justices) they trained the simulation on. That's fine as an exercise in simulation, but using it to reach conclusions on intent and bias is a real reach. I suspect the journalists hyped that part of it a lot more than the researchers themselves.

That's underwhelming enough as is. But what do you use as a measure of how "independent" a judge is?

Assuming no relationship between decisions is ludicrous. On many items that aren't terribly controversial, Ginsburg and Scalia, for example, would rule similarly just because they are trained judges with a background in US law.

Similarly, you wouldn't be surprised if Krugman and Friedman agreed on the proper answer to a question from an Econ 101 textbook, regardless that they would differ massively on more complex issues.

Add to that, the Supreme Court doesn't get the expected and routine "no-brainer" type decisions. It's where the ones with thorny legal interpretation and constitutional issues end up.

I'd be really surprised if you didn't have a correlation between how one particular justice votes and how the rest of the justices vote.

Obvious for the last 100 years

[AK Marc]

I liked my time in law school. Nothing pissed off a law-worshiper more than pointing out that the Supreme Court was a means to code unreasoned opinion into law, as the decisions use law to justify opinions, the opposite of what the courts assert (where they say they come to their opinions through examining the law, rather than force their personal opinions into law). The legal experts have been able to predict not only the direction in which they vote, but also the reasons they would give. But it's interesting to learn that an algorithm is sufficient, with no analyzation of the facts and law necessary.

Re:Fantastic

[Jane Q. Public]

That is interesting, but from your usage, it appears that you misunderstand what being an "activist" judge means. An "activist" judge is one who attempts to create law in the courtroom, as opposed to evaluating existing law. It is not a "left vs. right", "liberal vs. conservative" concept.

Re:Fantastic

[Bacon Bits]

Uh, that isn't wrong. You're wrong.

The legislative branch can and has overridden regulations, rulings, and the constitution. It does that by passing new laws. In the case of the US Constitution, new amendments must be passed. The most obvious example is the 21st Amendment to the US Constitution, but . You can also consider the example of the judicial interpretation of Eminent Domain in Kelo v. City of New London. Several states have since passed amendments to State Constitutions (Michigan, for example) which restrict or bar the state's ability to use eminent domain.

The judicial branch interprets the law, which includes removal of laws which are proscribed by other laws. It does not get a choice about what the law is, only how to interpret it and how each law interacts with the others or how a law applies in a given case.

The executive branch executes an enforces existing laws, and is allowed to issue rules and regulations that enable executing the laws. It does not get a choice about what the law is, only how to go about executing it.

The legislative branch exists solely to create new laws. It alone determines what the law is. It alone is given the power to amend the Constitution (Article 5). Neither the judicial nor the executive branch is given that power. Indeed, the executive branch doesn't even get veto power for constitutional amendments. Additionally, the legislature alone is given the power of impeachment (Article I), which may be used to remove any civil servant from office, including a President or Supreme Court Justice.

Congress has the ultimate trump card. The problem is that it's legislation by committee, meaning they spend all their time talking and very little actually doing anything. This, I think, is simultaneously the greatest and worst idea the founding fathers had.