Mireille Hildebrandt’s forthcoming article is a companion piece to her Chorley Lecture of 2015.1 In the earlier piece, she highlights the relationship between the ‘deep structure of modern law’ and the printing press and written text – building on this a case concerning constitutional democracy and transparency, both in the world of print and the world of electronic data. In this new paper, the emphasis is on law as computation – as compared with law as information in the earlier lecture.
Machine learning is often discussed as an opportunity for legal practice and adjudication, but what will that mean? Hildebrandt highlights how machine learning in the context of law is primarily a simulation of human reasoning found in written legal text; one needs to identify how law is associated with ‘meaningful information’ rather than information simpliciter. Key concerns with applying machine learning in law include the catch-22 of deskilled lawyers becoming unable to verify a machine’s output, and various ways in which such systems can be opaque.
Hildebrandt hopes that we can ‘speak law to the power of statistics’ and argues that machine learning and related practices and technologies ‘may contribute to better informed legal reasoning – if done well’. There is an interesting and healthy scepticism about the funding of current efforts and what this might mean for the consequences of what may be reported as innovation. Much of this relates, of course, to the driving factors around innovation in the legal profession and the changing ‘law firm’. The work therefore also sits within the body of literature now interrogating algorithmic governance (e.g. Kathy O’Neal’s Weapons of Math Destruction, Frank Pasquale’s The Black Box Society, and, more recently, the question of whether data protection law might provide a remedy for such concerns in Lilian Edwards and Michael Veale’s Slave to the Algorithm? Why a ‘Right to Explanation’ is Probably Not The Remedy You are Looking For).
Provocatively, Hildebrandt wonders whether the result of a certain type of interdisciplinary engagement is that law is simply treated as one kind of regulation e.g. in the mind of the law-and-economics scholar; this is contrasted with a (perhaps deliberately idealised) lawyer as the ‘dignified steward of individual justice and societal order’. Her response, which may resonate with many legal scholars, is to draw upon Neil MacCormick’s presentation of law as an ‘argumentative’ discipline (MacCormick did, as Hildebrandt does, engage with speech act theory as a means of understanding legal reasoning). The challenge, then, is to identify the way(s) to test and contest emerging forms of decision making, and to ensure that the relevant people are equipped with the skills and/or the nous to ask searching questions and to scrutinise the systems that we are rapidly putting in place.
This draft paper will appear in a much-anticipated issue of the University of Toronto Law Journal. Already, the Canadian journal has contributed to a debate around the legal singularity (of interest even if you think that the legal singularity is about as likely as The Singularity itself), in a special issue on artificial intelligence, big data and the law; the forthcoming issue, based around a March 2017 symposium, includes further contributions on democratic oversight and the future of legal education. Indeed, that question of how future lawyers will be trained is something that Hildebrandt ruminates upon in her article and struck a chord with this reviewer (currently working in a legal system where the training of solicitors is about to undergo significant change. If the next generation of lawyers and legal researchers is to be able to take on the socially important challenges outlined by Hildebrandt (especially in countering the arms race between those with the requisite resources and motivations), we may need to think a bit harder about the shape of the law school.
- Published as Mireille Hildebrandt, Law as Information in the Era of Data-Driven Agency, 79 The Modern L. Rev. 1 (2016).