We write law between inverted commas because the question whether the output of data-driven applications can be and/or should be qualified as law (meaning it would have legal effect) is one of the core research questions.
According to IBM, its platform for legal prediction (ROSS):
“is a digital legal expert that helps you power through your legal research. You ask your questions in plain English, as you would a colleague, and ROSS then reads through the entire body of law and returns a cited answer and topical readings from legislation, case law and secondary sources to get you up-to-speed quickly. In addition, ROSS monitors the law around the clock to notify you of new court decisions that can affect your case” (interestingly, this quote comes from a previous version of its website, for the current version check here - as you will see the market for legal search has been taken over by big law publishers, highlighting the crucial importance of research into the political economy of this domain of research).
The core idea here is that the more lawyers work with such a platform, the better it learns to ‘think like’ a lawyer. This kind of software is capable of analysing unprecedented volumes of legal text that human lawyers could not even dream to process in person, without becoming tired and without succumbing to irrational prejudice. Unless, of course, such prejudice defines the data on which the algorithms have been trained or stems from the choice of algorithms.
Data-driven legal ‘intelligence’ has invited claims that the legal profession is on the verge of a ‘major disruption’, producing a new type of artificial legal meaning, enabling ‘self-driving laws’ potentially moving towards ‘legal singularity’. Others take a more cautious position, or warn against the drawbacks of an ‘incomplete innovation’ causing ‘premature disruption of legal services’, and against pre-emptive environments saturated with ‘algorithmic regulation’. Though niches in legal scholarship have studied the application of legal metrics (jurimetrics) to develop and evaluate legal knowledge and decision systems, the current flux of legal analytics and legal prediction is novel. First, because it is not logic driven but contingent upon statistical inferences; second, it is not necessarily deterministic and its outcomes are not always predictable; third, it can be fuelled by unstructured data instead of being dependent on well organised databases; fourth, it does not merely implement decision trees developed by human lawyers but rather learns how lawyers ‘think’ by detecting patterns in data sets and by improving their (the software’s) performance in an iterative process of training and testing.
This research targets the assumptions and the implications of the ‘translation’ of legal thinking into mathematical patterns, while developing novel conceptual tools to address the relationship between human language and computer operations in the context of artificial legal ‘intelligence. This is pivotal when it comes to the protection of human capabilities that may not be c’ountable in the first place, highlighting that machine inferences cannot be equated with the legal thinking they simulate. The goal, however, is not to reject these technologies (e.g. natural legal language processing or NLLP), but to achieve a more in-depth understanding of how these systems are designed and of what they can and cannot do. On top of that we investigate how the use of data driven legal technologies will affect some of the affordances of text-driven law, notably the checks and balances of the Rule of Law and the kind of legal protection this entails.