- What is CRLP about?
- Why do this?
- How can we do that?
- What is the CRLP doing?
- Who else is working on this?
- What could it be used for?
What is CRLP about?
Finding a way for drafters to produce computer-readable versions of the logical structure of our drafts of legislation.
Why do this?
So that computers can use those versions to guide humans through the legislation, but also to check the drafts for inconsistencies or unexpected effects.
How can we do that?
By finding an easy-to-use way for legislative drafters to mark up the logical structure of draft legislation while they are drafting it.
- It needs to reflect just the way the drafter is structuring the legislation, without adding more.
- It needs to be as easy as the ways in which drafters currently mark up the paragraphing structure of their drafts (using Word styles, or XML editors).
What is the CRLP doing?
- Our project runs in 2023, 2024 & the first half of 2025, and builds on the work we had already started in LDO to apply this to the way we draft legislation for Jersey.
- We are currently embarking on work with the Centre for Computational Law (Singapore Management University) to see how we can advance our ideas on drafting tools, reading tools, and AI assistance for a logical reasoner, all using their L4 language.
- We have been working out how “If-Then” structures fit in our legislation, like in the illustration above for an imaginary rule on feeding animals.
- We have also produced a mock-up of a tool for legislative drafters to help them draft coherent logical structures and automatically mark them up to be computer-readable.
- Other streams of the project include -
- producing non-technical guidance and training material for legislative drafters on what we have learnt about drafting itself,
- working with Digital Jersey to see whether Artificial Intelligence can mark these “If-Then” structures, and
- digitising standard provisions for a Jersey version of “Common Legislative Solutions”.
For more see the list of streams in “Our work”, and the documents for each stream posted on OSF.
Who else is working on this?
- Globally there is the “Rules as Code” movement, which started in 2018 in New Zealand, and has since spread to Australia, Canada (and video), Singapore, UK, France and elsewhere - see reports by the Observatory of Public Sector Innovation of the OECD “Cracking the code” (2020) and “New Techniques for Building and Using Legal Encodings in the Drafting Room” (2024).
- Here in Jersey the Financial Services Commission is working on “RegTech”. JFSC did have plans to “enable the digitalisation of our regulatory content … delivering machine readable rules”, but is now working with Digital Jersey on use of AI to read and answer questions on JFSC’s guidance and codes of practice.
- Some tech firms are working on Artificial Intelligence (generative AI, or LLMs, like ChatGPT) to make chatbots read legislation and answer questions about it. But LLMs are not able to reason as such, so they will remain unreliable for tasks that involve applying legislation to given fact scenarios, particularly where that involves applying definitions (as opposed to looking up an answer that is already on the face of the legislation). So we are working on using AI to help with reliable logic-based systems.
What could it be used for?
Here is an example of what programmers can do once they have the logical structure of the legislation set out so that a computer can read it –
AustLII (who are the equivalent of JLIB, who publish our legislation) have produced DataLex, which is a free program that you can try out on the web.
It runs a “consultation”, asking you questions based on the wording of the legislation.
- It shows the facts it has taken from you so far, the conclusions it has reached so far, and links to the relevant parts of the legislation.
- At each question you can ask it why it is asking, or check what will happen if you give a particular answer, or tell it to forget some or all of the facts you have given it.
- At the end it gives you a report that tells you its conclusion about how the legislation applies to your facts, but also sets out exactly how it came to that conclusion.
We are working with AustLII on applying DataLex, yScript & yLegis to our example laws.
In the longer term we also expect this work to help with creating new tools for legislative drafters to check our work, like those available to programmers.