Doing Law by way of Computation?
This project is focused on two types of computational law: data-driven and code-driven
Artificial legal intelligence has come to a point where the prediction of the outcome of cases, ‘argumentation mining’ and the collection as well as the analysis of relevant legal texts can be performed by machines to an extent previously unthought-of.
These machines get better as their algorithms are ‘trained by’ legal experts in a process of ‘machine learning’, mostly based on data-driven natural language processing ‘trained on’ legal text. Many predict that such data-driven applications have the potential of outperforming the lawyers at some of their core tasks.
Katz, Ii, and Blackman developed software to predict the outcome of Supreme Court case law, outperforming legal experts, Aletras et al. built an application that predicted the outcome of judgments of the European Court of Human Rights with a similar suggestion of proficiency in tasks previously reserved for human intelligence.
Whereas artificial legal intelligence has a history, exemplified by the institutionalisation of jurimetrics in law schools in the 1980s, the current advance of ‘legal tech’ is novel and seems less prone to the discontents of old-school artificial intelligence.
The latter was based on an algorithmic understanding of law, celebrating logic as the sole ingredient for proper legal argumentation.
However, as legal philosopher and Supreme Court Justice Oliver Wendell Holmes noted, the life of the law is not logic but experience. Machine learning, which determines the current wave of artificial intelligence, is built on data-driven machine experience. In line with Holmes’ understanding of law, legal intelligence based on machine experience (data-driven law) may be far more successful in terms of predicting the content of positive law than its logic-driven predecessor.
Code driven law
Despite the failure of logic-driven legal expert systems, the quest for self-executing regulation or contracts has found a new impetus with blockchain applications that supposedly automate compliance with legal obligations that are hardcoded into the ‘ledger’ (a distributed archive of records that contain instructions that determine the behaviour of digitized assets).
Here we have a breed of logic-based ‘law’ that aims to rule out non-compliance altogether, or to automate the legal effect of non-compliance (e.g. compensation). The idea is that the blockchain enables a new type of ‘trustless’ computing, with less or no need for law’s conventional trusted third parties: the state and the courts.
Insofar as behavioural rules and assets can be translated into code, blockchain applications have the potential to self-execute contractual and regulatory obligations (‘smart contracting’ and ‘smart regulation’).
As the underlying technology depends heavily on cryptography, Wright and De Filippi speak of a new ‘lex cryptographia’, the most extreme example of an algorithmic law that is entirely code-driven and self executing. Code-driven law (blockchain applications) and data-driven law (machine learning applications) are investigated here under the heading of computational law.