Here, we point to international research initiatives we have identified in the course of this research. We identify them for policy makers for two reasons.
A group of academics with experience in “law as code” research pre-dating the present rules as code renaissance have advocated for a regulatory sandbox to be established in Australia.150 There is little detail at this stage about what that would entail, but we imagine it to be similar to the multi-stakeholder incubator we describe in our recommendations.
In oral comments, Prof Casanovas summarised as follows in a manner similarly framed to our own recommendations:151
In short, we need a bold government framework for collaborative public private rule coding innovation. In this space there are things we can do, things we cannot do and things we should not do. The framework, by direction [inaudible]forward for the next steps. The framework should also allow more challenging projects to be undertaken in order to expand our rule coding knowledge. The more challenging questions relate to coding of complex legislation. Pilot projects in this space require, for example, the collaboration of members of parliament and parliamentary council, and government departments and agencies too. Ideally, a regulatory sandbox should be created to support rule coding projects, evaluate the results and eventually support implementation. Having a framework for innovative collaborative pilot projects planned, and support for resources to implement the plan, will help position the government as a global leader in digital law and regulation, strengthen Australia’s law-tech industry and facilitate legal compliance by companies, government agencies and citizens. But much more is at play. By getting it right we will make law more accessible to citizens and enhance the transparency and accountability of our legal system as a whole. In short, it will foster trust. On the other hand, delaying or getting it wrong can do damage.
Australia has had significant experience of the negative consequences that can be caused by automated decision-making systems through the “robodebt saga”.152
The Australian Research Council has funded a Centre of Excellence around Automated Decision-making and Society. It “brings together universities, industry, government and the community to support the development of responsible, ethical and inclusive automated decision-making.”153 The centre has 79 total members and funding of approximately $70 million AUD from the Australian Research Council and Australian University and Industry Partner Funding. It has an extensive number of partners in industry, academia and civil society, both domestically and internationally.154
Prof Mireille Hildebrandt is a leading scholar in the area of law, philosophy and technology systems. She leads a research project funded from 2019-2024 by the European Research Council (ERC) under the HORIZON2020 Excellence of Science program. Dr Laurence Diver, another leading scholar on the topic of “Digisprudence” is also a member of the COHUBICOL team.
The COHUBICOL project is split into two streams. A data-driven law stream focuses on data-driven computing technologies such as machine learning. Another stream is “code-driven law”, which focuses on the representation of law in code, and the use of code to implement law. The existence of this code-driven law stream is an important indicator of the volume of scholarship that already exists and is also still being written.
The project’s orientation and situation within the EU will generate interesting conclusions about the way that “code-driven law” systems interact with article 22 of the General Data Protection Regulation (“GDPR”), which imposes controls on automated decision-making systems.
Policy-makers wishing to implement coded models of the law would benefit from engaging closely with work associated with the COHUBICOL project.
Singapore Management University has received a $15m research grant for a program investigating the use and development of computational law.155
Two prominent voices in the rules as code space with significant practical and research experience in law and computer science (Meng Wong and Jason Morris) are conducting research within the Singapore programme. Work is being done to develop machine executable languages that can closely reflect natural language drafting (Legalese). Work is also being done to develop tools that lawyers and others can use to create coded representations of legal instruments and interpretations (Blawx).
Some of the work being done by this program focuses primarily on private law applications, for example in contract law or the development of legal advisory tools. This is a point of contrast from better rules and rules as code approaches which sit primarily in the domain of public law, governing the relationship between State and citizen, including the passing of legislative instruments.
Work produced by the Singapore programme is likely to be highly influential and instructive for New Zealand.
Computational law has been a central focus for Stanford University’s CodeX programme for some years. The program is based in the Stanford Law School.156
At CodeX, researchers, lawyers, entrepreneurs and technologists work side-by-side to advance the frontier of legal technology, bringing new levels of legal efficiency, transparency, and access to legal systems around the world. CodeX‘s emphasis is on the research and development of computational law — the branch of legal informatics concerned with the automation and mechanization of legal analysis.
Rules as code and better rules advocates would benefit from close scrutiny of the Centre’s work over the years.
During the course of this project, and prompted in part by the OECD primer, we became aware of a method attributed to Dr Mariette Lokin in Denmark. We spoke with Dr Lokin to assess the extent to which a better rules approach mirrors her own approach and concluded there was striking similarity.
Denmark has an “Agency for Digitalisation”, which is responsible for ensuring that legislation is suitable for implementation in digital systems. This is a core benefit attributed by advocates to the better rules process.
The Dutch Ministry of Interior commissioned a report by Hooghiemstra & Partners on “the supervision of the use of algorithms by the government”.157 The report identifies Lokin’s method as one way of ensuring that algorithms giving effect to legislation conform with the provisions of that legislation and recommends it is investigated further.
Legislative process: the transformation of laws in computer programs so that they can be executed automatically is now concealed from the view of the House of Representatives or the City Council. This can be improved by trying out the Lokin model in which legislative texts are written in a very structured and precise manner to make the programming more easy. The discretion on how to interpret the law, will then (again) be part of the democratic checks and balances (Lokin 2018).
The Lokin method is described in a PhD Thesis from Lokin. Only the abstract is available in English. The core problem as described by Lokin is more or less the same as identified by the Better Rules discovery report (2018). Lokin examines five applications for digitising legislation and states:
Striking in these five cases is that knowledge modeling is not yet based on a direct analysis and interpretation of the legislation. Where it is based on analysis of legal sources, no tools for unambiguous interpretation of the legislation are available. Furthermore, law drafters are rarely directly involved in the conversion of legislation into IT-applications. The approach proposed in this thesis could fill these gaps. The underlying assumption is that agile execution of legislation also requires agile legislation and that the role of the legislator as the administrator of the legal system nowadays expands to ‘technical’ system management. This leads to the central research question of this thesis: To what extent and in what way can digital execution of legislation by national government agencies be taken into account in the legislative process?
The approach advocated by Lokin, like better rules, focuses on the use of knowledge assets that are used across various systems, one of which may be
The essence of this approach is that knowledge from legislation (rules, data, process steps), required for automated decision making, is no longer ‘locked up’ in the system, but is shaped into knowledge models outside the system, forming the basis for modular IT services to execute processes. In doing so government agencies aim to increase the agility of their IT systems and thus of the implementation of legislation. In recent decades, a great deal of research has been carried out to find ways of sup porting knowledge-based working, for example by displaying knowledge in a formalized way, supporting the conversion to automatically executable specifications. However, a method for clarifying the meaning of the legislation on which the specifications are based is still lacking. This thesis elaborates an approach for this; it focuses on the creation of legislation and is based on three pillars:– clarifying the meaning of legislation;– improving insight into the legal rules relevant for the task performance by government agencies;– a different way of cooperation between actors in the legislative process.
Lokin effectively advocates for the use of multidisciplinary teams using agile approaches to the policy process, including the use of specialists from business process analysis, developers and legislative drafters work together.
Therefore, the next question is how to bridge the gap between language and technology, by establishing a different way of working among the various actors in the administrative-political legislative process. In answering this question, a distinction has been made between the ministerial and political phase of preparation of legislation. With regard to the ministerial phase of preparing legislation, an analysis has been made of several agile working methods in systems and software development, such as Scrum, Lean and DevOps. The main characteristics of these methods are that projects are carried out in an iterative manner (in small subprojects with a short turnaround time) and in multidisciplinary teams (in which representatives from the business processes, IT developers and managers sit together). Elements from these working methods have been combined in an approach that is referred to as LegOps, a composition of legislation and operations. This approach aims to achieve agile legislation by viewing the legislative process as part of the chain of policy making, legislation and execution and by using an iterative and multidisciplinary way of working.
Lokin also considers the long-term implications of increased digitalisation of the law:
For the long-term, a more fundamental reflection on the relationship between government and parliament is required. For a properly functioning parliamentary democracy, not only the relationship between citizens and representative bodies is key, but also the relationship between (co-)legislative and executive power. The latter determines the legitimacy of government: the extent to which the execution of legislation leads to legitimate and just decisions and has the intended effects. In this regard, research could be conducted into extending the LegOps approach to the parliamentary phase of the legislation process, integrating this phase in the chain of policy making, legislation and execution. Of course the research should keep an open mind for the values and guarantees that are (and must remain) anchored in the legislative process.
Lokin also raises a question about the role of the judiciary, but again, the thesis is an example of the merits of international academic cooperation so that teams can learn from each other and avoid making the same mistakes.
Another research theme concerns the impact of the application of the language model and linked data for the judiciary: to what extent can or will a judge be bound by the meaning given to legal provisions by means of annotations of the legislator? The aforementioned themes deserve a place in a broader discussion on the consequences of technological developments for legislation, governance and the judiciary, enabling them all to fulfil their role as system administrator adequately in the long-term.
A submission was made to the Australian Parliament’s Select Committee on Financial Technology and Regulatory Technology: <https://www.aph.gov.au/Parliamentary_Business/Committees/Senate/Financial_Technology_and_Regulatory_Technology>. ↩︎
Hansard: Select Committee on Financial technology and Regulatory Technology, Senate of the Commonwealth of Australia, Thursday 11 February 2021 at p 30: submission by Prof Pompeu Casanovas, Dr Mark Burdon, Prof Louis de Koker, Dr Guido Governatori, Dr Anna Huggins. ↩︎
Whiteford, P “Robodebt was a policy fiasco with a human cost we have yet to fully appreciate” 16 November 2020: <theconversation.com>. ↩︎
Prof Valerie Frissen, Dr Marlies van Eck, Thijs Drouen “Research report on supervising governmental use of algorithms” (Hooghiemstra & Partners, 2 January 2020): < https://hooghiemstra-en-partners.nl/wp-content/uploads/2020/01/Hooghiemstra-Partners-rapport-Supervising-Governmental-Use-of-Algos.pdf>. ↩︎