Data Science and Law (and Farewell)

In my last post I want to briefly discuss the experience of Bar-Ilan Law Faculty (where I serve as Dean) in a new joint research initiative with the Bar-Ilan Data Science Institute. This joint project builds on Bar-Ilan strength in data science (DS), especially in the fields of natural language processing (NLP) and network analysis. The project is motivated by idea that the law rich textual and web structure makes it a great medium for analysis using the methods of NLP and network science (see, e.g., my recent paper, Transnational Networked Constitutionalism, co-authored with Ofir Stegmann). We currently have more than 20 on-going research projects using DS methods in various stages.  

Our experience in these joint studies has raised several challenges and questions and I will be happy to hear from others who have been involved in similar projects about their experience. We will also be very happy to cooperate with other institutions.

Probably the most critical issue for the success of such projects concerns the need to move into a team based work. Almost all our projects are based on joint teams that include, DS and law profs, graduate students and supporting stuff. This mode of work brings legal research closer to empirical social sciences and research in the natural sciences. It requires willingness and openness from both the DS and law side.

Another challenge we had to cope with from the start is how to think about the roles of the DS and law profs in such joint project. A naïve way to think about such cooperation is that the law side should be responsible for collecting the data and the DS side should be responsible for analyzing it. We think that this is a mistaken paradigm. A good interdisciplinary DS-law joint-project must involve the two sides across the whole life-cycle of the project. It is important that the DS people will be involved at the data collection phase (which involves critical questions about what data to collect and how to structure it) and in the hypothesis framing phase. It is also necessary for the law side to be involved in the analysis phase (even if the technical analysis will be led by the DS people). This requires ‘each side’ to develop some understanding of the ‘other’ knowledge domain.

Another question concerns the publication and evaluation of the results of such interdisciplinary projects. In most cases the main contribution of the project would be in the legal domain and not in computer science or in mathematics. It will commonly use existing methods to study law-related questions (although law could also trigger innovation in the DS domain). However, a significant work may need to be done in order to adapt and apply such methods to specific research questions and environments. This means that the venues in which such work could be published would probably be in legal journals that accept empirical work. This could create a motivation problem for the DS people. Solving this problem requires university authorities to explicitly support interdisciplinary work and to recognize the contribution of DS people even when the work is published in journals outside the DS domain. Equally law profs should be ready to venture beyond traditional legal publications toward DS journals (where the focus could be on the more technical aspects of a project). Such extension of the publication spectrum is important for the feasibility of such joint-projects.

Let me close by thanking Howard Wasserman and the forum again for having me as a guest this month. Thanks also to all those who responded and commented on my posts.