Can You Predict Litigation Outcomes Using a Judge’s Personal Information?

Pre/Dicta is a litigation analytics company that seeks to provide lawyers and clients with a steer on case outcomes, such as motions to dismiss, by examining personal details of judges in the US. This includes: the judge’s net worth, education, work experience, and political affiliations, among other things, in order to find likely biases toward certain issues and parties in a case, or their biases toward or against certain lawyers involved in the litigation.

They have started with a ‘motion to dismiss tool’, but added they plan to provide more tools based on the same approach in the future.

‘Pre/Dicta helps top litigators understand and apply judicial behaviour in ways no human ever could. We are confident that our motion to dismiss prediction tool, and the others to come, will be indispensable parts of any top litigator’s overall litigation strategy,’ the company said.

Companies such as Gavelytics, which is closing down (see last night’s story), have tried to help predict outcomes by looking at how US judges act in court, such as allowing or denying certain motions. But, this is going far beyond that, and moving into ‘considering biographical information of all federal judges’.

Leveraging those insights, Pre/Dicta ‘evaluates how case-specific data points affect a judge. For example, does the size or prestige of the firms involved influence a judge? Are they more sympathetic to an individual plaintiff? Do they favour publicly traded companies over privately held ones?’ they explained.

They added that they have ‘tested and proven predictions against more than 1,500 federal judges over 10 years of federal cases’.

The company concludes that with this data they can suggest whether a certain judge will find whether ‘attorneys are more credible if they also attended Harvard Law like the judge did’, for example.

Now, before getting into the pros and cons of this approach, Artificial Lawyer asked Pre/Dicta CEO, Dan Rabinowitz, some more questions.

How does this work?

We are focused on data that has not been previously accounted for – judicial biographical information and how that influences judges’ decisions. Until now, the focus has been on historic data related to a decision (i.e., how often is a motion granted.) But that is backward-looking and not intended to predict future behaviour. Our focus goes beyond looking at the historic decision – we turn our lens on the judge.

What is the effect of particular firms or litigants on the judge? If the judge is in debt, is he or she more or less plaintiff-friendly? Personality-driven questions such as these cannot be answered by traditional research or the current platforms, although they will ultimately decide the case. That is where Pre/Dicta comes in.

We collected that behavioural data and use it in our algorithms to answer the question of how your judge will rule in your case. Our approach is simple: just provide the case number and our algorithms do the work of analyzing past decisions and then making a verifiable prediction.

Do the results get any human ‘reality checking’?

We objectively assess our accuracy based on an analysis of decades of cases and how our algorithms performed against their results. We have achieved a very high degree of accuracy – 86% – giving us confidence in our predictions. Of course, our models and analysis are the product of the dedicated work of our data scientists, software engineers and analysts. But now that we have built our models, they operate independently. 

Also, how do you get enough data to prove a point? e.g. that a judge favours lawyers from Harvard? Wouldn’t you need a lot of data to do this?

Yes, we identified and collected large datasets of many different types and do not rely upon any one variable. There is no one element that determines outcomes. Instead, to accurately make predictions, we cast a wide net and incorporate multiple dimensions about the judge, lawyers, and litigants. Ingesting all that information allows us to avoid the trap of inappropriately weighing judicial characteristics. 

Which courts does it cover and where do you get your source court data from?

Pre/Dicta covers civil cases in all federal district courts. We use a combination of publicly available data sources and ones that we have culled and created ourselves. Judicial decisions reside in PACER and some biographical information is collected by the Federal Judiciary. But other data, especially biographical, require us to collect, ingest, and process from a number of disparate sources. At times we have a researcher combing through hardcover judicial publications and others a sophisticated scraping program used to access digital media and other sites.

The Pros and Cons

First, it has to be said that such an approach would go down like a lead balloon in the UK and also across many other countries where judges aim to be seen as apolitical and where their personal life, e.g. their net wealth, is certainly not meant to be considered in any way at all.

Second, this is a product for the US market, and things are very different there. For starters, some judges are political appointments, and some openly flaunt their political affiliations.

So, we have to see this in its cultural context. Next, is this a good idea, or rather, a workable idea?

One could argue that the more you know about the people involved in a case – especially a judge who is making key decisions – then the better. One could also argue that this system is just doing what American trial lawyers do already by instinct every week, i.e. consider the personality of a judge and then see how that will play with their case and their clients.

Will it work? They say it works very well. But would you base your whole case’s chance of success on this? That would seem to be a very risky strategy, just as it would be for that level of confidence in any kind of predictive system for something as complex and sometimes random as a court case.

But, can it help to contribute some valuable intelligence to the overall case assessment, or parts of it, such as a dismissal? Yes, that seems fair, as long as you see this as just one slice of data in a mass of other very complex data, rather than some kind of guarantee.

Finally, do we want to live in a world where lawyers and judges have their personal lives examined like this? This site would say: no, but maybe that’s a UK-centric point of view that reflects a legal culture where a lawyer’s or judge’s personal life and political views, or wealth, are their own business. But, maybe elsewhere this approach would not raise any issues at all. What do you think?