We write law between inverted commas because one of the questions core to our research is whether code-driven rules can be and/or should be understood as law.
Whereas data-driven legal intelligence aims to support legal advice and legal decision-making by means of description and prediction, code-driven law is an attempt either to translate or to directly write legislation into computer code. This can be done by seeking to write legal norms into blockchain technology, creating a cryptographic law that implies automated execution via a distributed ledger. Our focus, however, is not on self-execution but on what has been termed Rules as Code, including systems like RuleSpeak, Catala, and Akoma Ntoso.
The aim of code-driven law may be
- to foresee potential incompatibilities between applicable legal norms within or outwith the same legal domain,
- to help reduce the complexity of the legal universe by designing unambiguous legal rules that avoid conflicts with other relevant legal rules, and/or
- to support the automation of legal decision making in public administration, based on legislation that is expressed in dedicated programming languages.
Those involved in designing ‘rules as code’ often emphasise that only dedicated legal domains or subdomains lend themself to this type of computability, turning legislation into an unambiguous set of rules, enabling the construction of simple and complex decision trees, which in turn afford automated calculation of legal decisions. The prime example is tax law, which (1) should avoid uncertainty, to preempt attempts to ‘game’ the system, (2) seems inherently open to logic and calculation though often highly complex arithmetic, and (3) requires huge numbers of decisions to be made that should be taken speedily while at the same time honouring the principle of equality before the law.
Though we do not focus on blockchain applications, we do investigate the issue of code-driven compliance on the side of, for instance, companies that use software to integrate compliance with e.g. data protection into their business process. Here, the project links to another research project, ALTEP-DP, which is focused on the automation of data protection, investigating the difference between ‘legal by design’ and ‘legal protection by design’. The first indicates an attempt to fix legal problems by way of technical solutions, the second takes a broader scope, aiming to integrate checks and balances at the level of a computational architecture, with the aim of providing the kind of legal protection required under the rule of law.
Again, though we do not focus on blockchain applications, we do investigate the issue of code-driven enforcement by public administrations, for instance, the tax office, student grant provision or social security agencies. Once regulators embrace the idea of automated enforcement, they will need to articulate legal norms in terms that easily translate into computer code. In domains such as the regulation of financial markets this may be a welcome development, depending on how issues of jurisdiction are resolved that are core to the complexity of ‘fintech’ regulation. We should expect, however, a further integration of code-driven regulation into mainstream legal practice, forming a hybrid with text-driven enactment of legal norms. As De Filippi and Hassan (2016) argue, this may effectively ‘turn code into law’, thus effectively changing the meaning of law-as-we-know-it.
Automated compliance or enforcement based on ‘softwired’ legislation encounters three pivotal issues:
- the decision trees that follow the code will implement a specific interpretation of the relevant legal norms, which have been agreed upon by human beings speaking their natural language. This implies that issues of interpretation are inherent in natural language’s ‘open texture’ are now decided by those who write the code and the developers of compliance or enforcement software
- those subject to these kinds of ADM systems may have trouble contesting decisions that affect them, especially if it is unclear how the logic of the system relates to the legal justification of the relevant decisions. This becomes particularly relevant in the case of hybrid systems, where the decision tree is based on the outcome of machine learning applications
- a further point surfaces when those subject to these decisions cannot even resist them, turning law into administration or discipline, notably when contestation (appeal or complaints to a supervisor) is costly, takes a long time or is limited to predetermined classes of objection.
This research digs into the assumptions and implications of inscribing or transcribing legal norms into computer code, while developing novel conceptual tools to uncover the interactions and transformations that result from such in- or transcriptions.
The goal is:
- to achieve a more in-depth understanding of how code-driven law affects the affordances of text-driven law,
- how it will interact with data-driven law or fit into hybrid AI, and
- what implications this may have for the kind of legal protection that is key to the Rule of Law.