The article discusses the intriguing ability of legal analytics algorithms to draw conclusions about court cases – such as identifying certain behavioral patterns of judges – and the implications that these algorithms bring about.
An article by Henrik Trasberg
In 1949, Lee Loevinger coined the term „jurimetrics“. He considered that while jurisprudence is a philosophical approach where questions of law are subject to speculation, jurimetrics in contrast adopts scientific methods to make relevant conclusions about law. Such approach is often also referred to as „quantitative legal prediction“, „predictive justice“ or “legal analytics” (I will adopt the latter term in this article). They all describe a similar idea: significant predictions about legal outcome can be made by automated analysis of a large amounts of data about judicial decisions.
Today, with the emergence of highly capable AI, legal analytics has started to generate quite a bit of excitement in the US and European legal practices as it has started to mature into a useful tool. Essentially, the legal analytics algorithms are given an extensive dataset consisting outcome of previous cases and certain information about the features of these cases, such as the text of lawsuit or case metadata (name of the judge, the court, subject-matter of the case, etc). The algorithm then learns correlations between the data and the outcome of those cases, based on which it either predicts outcomes of a new case by looking at the features of the case, or identifies patterns about what are the arguments, court precedence or evidence used in previous cases that most significantly correlate with a positive outcome. These correlations can be identified about specific regions, courts, judges or juries. Notably, this effectively enables to draw conclusions about what a particular judge might like or dislike in a litigation process.
The value of legal analytics
We are seeing a sprout of commercial projects such as Lex Machina, Predictice and Premonition, which provide legal counsel or their clients data-based insight into what may be the most advantageous path in proceeding with a lawsuit, by suggesting which lawyers tend to be most successful with a particular judge, which cases the judge tends to cite, which legal arguments it adopts, etc. The output of these algorithms can also be rather complex: based on the historical data about a judge the algorithm can seek to identify what wording to use for arguments with a particular judge, which experts to choose or how to incite the judge to impeach the expert of the opposing side (see, for example, the description of LexisNexis legal analytics tool).
The value proposal of legal analytics is thus evident: it can provide increasingly meaningful input for the litigation parties to make more informed decisions and manage legal processes with better efficiency as it enables to understand tendencies or preferences of judges, helping to tailor legal arguments for a particular judge.
The negative consequences
Yet, there are a number of concerns emerging with the adoption of the legal analytics tools, three in which I would particularly like to outline:
- The most immediate issue concerns the quality of the output. Many algorithms at least partially form spurious (random) correlations, making the value of their output dubious. The restricted extent of the judicial data available also limits the ability of the algorithms. Furthermore, it must be considered that law is subject to continuous development, which any algorithm that uses historic patterns for its output fails to capture. Thus, it is likely that the results of these algorithms may at times direct its user on a wrong path and actually have a detrimental impact on the efficiency and quality on the litigation procedure.
- As the legal analytics tools do become increasingly effective, the unequal access to these tools would enhance the advantage that the wealthier and more powerful litigation parties have over the parties who would not have the possibility to access the legal analytics tools. This undermines the equality and fairness of adjudication.
- Thirdly, legal analytics tools can exploit biases of judges and the disparities of the adjudication system. There is a lot of research (see here, here, here and here) to show that judges’ decision-making process is strongly influenced by its subjective attitudes, ideologies, emotions, heuristics and other biases. Legal analytics tools which recognize personal patterns of judges might partially enable the litigation parties to exploit these biases. The mass strategic exploitation of the tendencies of judges or disparities that exist between different courts might undermine fairness of the adjudication system, given that such practice essentially seeks to tilt case outcome by engaging factors that are not actually rooted in law. This would also erode public confidence in the legal system.
In fact, we have already seen the first significant reaction from the legislators to the adoption of legal analytics – in June 2019, a new French Justice Reform Act entered into force, which effectively bans the use of legal analytics tools for publishing patterns about judges. The punishment for the infringement is, staggeringly, imprisonment for up to five years.
The future of legal analytics in courtroom
Despite the above concerns, the wider perspective should be that the legal analytics tools can increase the transparency in the adjudication process, which enhances predictability of litigation for individuals and helps to identify (and rectify) disparities of the legal system. Therefore, there is real potential for developing a more transparent and harmonized adjudication system that that could be captured with these tools. Transparency and predictability are, after all, one of the key objectives of the rule of law.
In order to mitigate the risks that arise with the amplified exploitation of the biases and disparities of the adjudication system, the progress of legal analytics requires attention and willingness from the judiciary and lawyers to develop the adjudication system in line with the capabilities of the legal analytics tools, likely requiring the legal community to increasingly understand the language of statistics.
Henrik Trasberg is an intellectual property adviser at the Ministry of Justice in Estonia and a legal counsel in technology law at the Hedman Partners Law Firm. He has written his master thesis at Tilburg University on the rule of law’s implications of predictive justice.
Published under licence CC BY-NC-ND.