A Few Problems with All Attitudinal Prediction Models and Assisted Decision Making

December 1st, 2011

Someone else may have written about these issues, but they occurred to me the other day.

All attitudinal models look at the ideology of a judge to determine how he/she will vote. Some projects use the political party of the appointing president as a measure of ideology. Others use the judge’s voting history in cases (tagged broadly and imprecisely as conservative or libertarian) to determine ideology. But there is one big problem.

But all of these models start with the presumption that you know the identity of the judge. On the Supreme Court, we know the identity of the 9 Justices well in advance, but for all predictions on the inferior courts, the parties usually do not know the identity of the judge until a relatively late stage. In some courts of appeals you do not know the identity of the judges on the panel until a few weeks prior to oral arguments. And in reality, all of these models are usually applied retrospectively (after the case has actually been decided), so the late identification of the judges seems irrelevant.

It would seem that from a practical perspective, such models are purely academic. That is, such models would be of little use to a practicing attorney. In the case of a court of appeals, by the time the identity of the panel is released, the case is already two or three years old. The parties have spent countless hours in researching, discovery, briefing, and arguing at the district court and the court of appeals. By the time the identity of the panel is released, the case is full steam ahead. A prediction at that point will likely be of little use. Perhaps a party will settle on the eve of oral arguments, but due to the sunk cost (fallacy) parties will likely go for it, and see how the juges decide.

The most important time to offer a prediction of how a case will turn out is at the beginning–before the complaint is even filed, and well before the identity of the judge is known. Or, it would be useful to know the likely outcome of a case before an appeal is filed–at a significant cost.

Here, all attitudinal models fail. A super-crunching algorithm that can abstract and forecast likely outcomes of case in a given jurisdiction based on different draws of judges would be the only way to remedy this deficiency, and provide practicing attorneys with something of value at the early stages and key junctures of litigation.

This is something I hope to offer. I like the term Assisted Decision Making. We would not be delegating the decisionmaking to computers, but simply relying on technology to aid in our decisions. Do you like that term?