On April 1, HData CEO Hudson Hollister interviewed Matthew Waddington and Heather Mason, who help manage legislative drafting for the government of Jersey. The Bailiwick of Jersey is a self-governing parliamentary democracy and British Crown Dependency. Jersey is one of the world’s largest offshore finance centers, with a per-capita gross product among the highest. Our colleagues explained their vision for the future of machine-readable and machine-executable legislation to improve the work of Jersey’s legislators and better serve Jersey’s people.
Hudson Hollister:
We’re excited to be joined by Matthew Waddington and Heather Mason of the States Assembly of the island of Jersey. Matthew Waddington is a senior legislative drafter and deputy head of the legislative drafting office. Heather Mason is a legislation editor.
We’re excited to dig into the concept of machine-executable code and the ways in which technology can make legislation and regulation easier for those who produce it and those who comply with it. To begin with, I’d like to start by asking each of you about what led you into legislative drafting.
Matthew Waddington:
I started in law with an interest in community law centers. I’ve had various jobs since working with the law and I fell into government work and then drafting. I first accepted an opportunity in Cyprus, for the British bases there, and when that came to an end I ended up in Jersey. I’ve had a bit of an odd way around to legislative drafting.
Heather Mason:
If you think Matthew’s pathway was odd, it’s not a patch on mine! I spent about 20 years as a science teacher and an examiner for one of the major UK exam boards, writing science examinations for 16 – 18 year olds. I did a master’s in law for fun in my spare time, and when I moved to Jersey 3 years ago, the timing was perfect for a career change into law. I was fortunate enough to join the legislature drafting office as an assistant legal adviser. I initially worked on BREXIT alongside Matthew and when the legislation editor role came about, a new role for Jersey, I was fortunate enough to get it.
Hudson Hollister:
Matthew could you tell our audience a bit about the ways in which the legislative drafting office serves the Jersey States Assembly.
Matthew Waddington:
We draft all the legislation that is made in Jersey, both primary and the secondary legislation, but not the UK legislation that gets extended to us. We don’t currently have much role in the Court rules but otherwise we work on everything. We also draft the amendments by backbenchers [members of the Assembly who do not hold office in the government – in the UK House of Commons they sit behind the “front benches”]. We are in the lucky position of being able to ensure quality control over anything that ends up on the statute book.
Hudson Hollister:
For those who are in jurisdictions that do not use the concept of primary legislation and secondary legislation, could you analogize those to perhaps laws and regulations.
Matthew Waddington:
Yes. Laws and Regulations, really. Jersey is slightly odd. We have our Assembly, which is like a parliament but the Laws (the Acts), the primary legislation, have to go to the UK for Her Majesty in Council to approve them. The States can then make Regulations under those. So it’s slightly odd that we have a parliament making delegated secondary legislation, and then we have ministers who can make Orders under Laws or Regulations.
Hudson Hollister:
We would like to dig into the concept of machine-readability and machine-executability for both primary legislation and secondary legislation. Before we do that, I would like to get a bit more of a sense of the challenges of drafting and amending.
What are some of the hallmarks of the most complicated drafting or amending tasks? What are some of the circumstances where members of the assembly need to make complicated changes and what drives the complexity?
Heather Mason:
My role in amending legislation usually occurs after the amendments have been drafted. As the editor, I go through and check that the amendments fit into the piece of legislation that’s being amended, and also check for consistency, use of plain English, gender neutrality and such things. Matthew would be better placed to tell you about the actual drafting.
Matthew Waddington:
Heather sees everything that goes through. In that sense, she’s been dealing today with backbench amendments to a piece of legislation that’s already been lodged [tabled in the Assembly] where they are reforming the Assembly, changing who gets different kinds of positions, and the backbenchers are coming up with all sorts of changes. Those do get quite complicated and quite difficult to handle.
Heather Mason:
Essentially, what we’ve got at the moment is various members of the Assembly wishing to lodge amendments that perhaps contradict each other or are essentially the same amendment. So we try to work with them behind the scenes so that the amendments achieve what the members wish it to achieve without there being repetition.
Hudson Hollister:
In those circumstances, might you have multiple attempts to amend an underlying or existing piece that might conflict with each other?
Heather Mason:
Yes, exactly.
Matthew Waddington:
We have an advantage over some jurisdictions, in that you can’t amend in debate, you have to lodge an amendment in advance, so that people can think about it, and then they debate it.
Hudson Hollister:
Yes, there are certainly some jurisdictions around the world, in which the amendment can be written on the proverbial paper napkin and as long as it gets a vote and the voters approve it then the amendment must somehow be integrated into the existing.
Let’s describe the concept of machine-readable amendment or machine-readable primary or secondary legislation. Matthew, you’ve become a leader worldwide in envisioning the application of technology to the law. What are some of the threshold problems that machine-readability or machine-driven amending might be able to help us to solve?
Matthew Waddington:
So there’s two things there: machine-driven amending and actual machine-driven legislation. Machine-driven amending, I think, we are a lot further down the road already with XML and Heather might be able to talk more about that. The idea of “rules as code,” certainly isn’t my idea but something I’ve piggybacked on, is to try to capture some of the meaning of the legislation. What we’re trying to bring to it as a drafting office is a drafter’s perspective.
Sometimes the tech folk run away with the idea that you can encode the whole of a piece of legislation. To some degree we’re trying to rein it back because what we get is lawyers, academic lawyers in particular, objecting to say, “You techies are just not understanding what the law is or how the law works. It may look like a set of algorithms but it isn’t.”
We are arguing for a way of looking at it in which all we’re trying to capture is what we ourselves do as drafters in a text. We say “if” this “then” that. We are careful with “and,” “or,” and “not.” We say you “must” do this or you “must not” do that.
Hudson Hollister:
Matthew I’m sorry to break in. It seems as though you may be describing the more advanced concept of machine-executability and I’m curious about the basic one first, machine-driven amending. You said we’re much further along when it comes to machine-driven amending.
Matthew Waddington:
We’re much further along when it comes to machine-driven amending but “we” in the sense of the world, not us. The traditional Westminster-style approach to amending is very detailed – “for these words there are substituted those words”. Great long lists which someone looking at it would find very difficult to comprehend, including the policy officials who are telling us what they want. It’s the tradition with Westminster debating because the idea is to limit the debates just to particular tweaks.
There are various ways in which people will come up with solutions like, “Okay, just rewrite the law as it would be if these changes went through. Let’s compare the two documents, the existing law and the changed law, and let’s get the computer to compare it and generate the fiddly text of the instructions for the change.”
We’ve been looking at whether we could do that but the more that we look at it the more we wonder whether it’s worth bothering. Because, in a sense, it might be easier just to present the assembly, or a minister, with the new text and a marked up version that shows what’s changed.
Hudson Hollister:
That sounds quite simple. I want to ask Heather about the same concept. Heather, what do you view as the primary advantage of machine-enabled amending?
Heather Mason:
The main advantage, as far as I’m concerned, is it would save a lot of work for me and my editorial team. Because at the moment, we do all the consolidating. We consolidate the amendments into the existing legislation and if any of the assembly members or ministers request a marked up copy to show the amendments, that is something we do manually, which is obviously quite time consuming. The idea of it being able to be generated automatically would certainly save us quite a lot of work.
Matthew Waddington:
Drafters are quite solitary people. We don’t necessarily talk to each other about what we do. But this kind of thing gets us to talk to each other and some of us are producing amended versions each time we do a draft. That’s the only way for us to get our heads around what the overall effect is and for us to be confident that the policy officer has understood what the end result of it will be.
But this duplicates the work Heather’s team are doing. We realized that we are wasting effort and that’s partly what makes me wonder whether we skip to redraft the thing and we get the tech to help show what’s changed. That way we focus the assembly’s mind on the new text and the differences, rather than on “in sub-paragraph (b) for the word ‘this’ substitute the word ‘that’ ”, which they struggle with.
Hudson Hollister:
Let’s move from this more simple concept of machine-enabled amending or machine-readable drafting and amending to the more advanced concept of machine-consumable legislation.
It’s somewhat difficult to get our terms straight because much of the legislative world has been discussing these different concepts and perhaps not separating them properly. Matthew, I’d like to ask you to take the first stab. What is machine-consumable legislation?
Matthew Waddington:
It’s a slippery concept. It’s not necessarily machine-executable legislation, for starters. It’s something that the machine can read some of the meaning in. Currently around the world we are producing legislation where the machine can understand, “this is sub-paragraph (a) of paragraph (3) of section 6 of the so-and-so Act.” We’re quite good at that. In that sense the machine is reading from metadata.
We would like to move it on so the machine can read that “this is a criminal offense,” or “this is a definition.” That doesn’t seem like rocket science to me, in that we could put in some metadata saying that’s a criminal offense or that’s a definition. Those are sort of the baby steps. The next step is to say, “these are all the occasions where that defined term is used.”
Every time we use a term, even if it isn’t defined, it should mean the same thing. That’s one of the rules of how we draft that isn’t necessarily obvious to the reader. That isn’t how English normally works. You can use English words to have all sorts of different meanings. You can use words in all sorts of different ways. We as drafters have our own rules, so if we use a different word, if we use “feline” here and “cat” there, that is meant to flag a difference in meaning. If we use “cat” here and “cat” there, that’s meant to flag the same meaning. We’d like to see if by using markup, or whatever, we could flag that to the computer – that the same word means the same thing.
It’s not a question of using artificial intelligence or natural language processing to go back over old legislation where the drafter may have made a mistake and may have used the same word twice with a different meaning. It’s more to get us to think “I am now using ‘cat’ here, I know I’ve used it somewhere else, I better make sure I mean the same thing.”
Hudson Hollister:
Heather, I’d like to ask you if you’ve got anything to add. As we define these new concepts that have never before been implemented in legislative practice, how do you separate the concept of machine-enabled amending or machine-enabled drafting on the one hand and machine-consumable legislation on the other?
Heather Mason:
With machine-enabled drafting we almost want something that’s going to test the logic of the draft as we go through. If we can mark up that x is a certain type of thing and y is another type of thing, the computer is actually checking the logic for us as it flows through the draft so that each part relates to what’s come before it. If it’s a piece of amending legislation that’s being drafted, we also want to ensure that it properly relates to the instrument being amended. If you’re making an Order under a Law, for example, the terms that have been defined in the Law don’t need to be defined again in the Order, so you’d want to flag those sort of things up to you automatically to save human checking.
We’ve got quite a stringent human checking process whereby every piece of legislation is checked by three different people for different things, and it would be quite nice to have some of that checking being automated.
Hudson Hollister:
Yes. How about some of the more science fiction ideas?
Heather Mason:
I think that whole, “this is all going to be done by robots thing” just puts people off. I don’t think it’s particularly helpful. People see it as, “the robots are coming to take our jobs,” but it’s more that these are tools to help humans do our jobs more effectively.
Hudson Hollister:
I want to get into some of the benefits of the concept of machine-consumable legislation. Matthew, can you envision any circumstance where compliance with the law might become easier as a result of machine-consumable legislation?
Matthew Waddington:
Yes. I think, as Heather says, people let their imaginations run away with them into RoboCop. You don’t have to have that as the outcome to think this is worth doing. And equally, you don’t have to assume that is the outcome, throw your hands up in horror, and run a mile from it.
When you think about what businesses are doing, across the world, every bank is running software for its procedures, and is trying to comply with the law. In order to do that, the banks hire software people, throw the law at them, throw the bank’s policies at them, and say, “put all this together and make a program that will enable a frontline worker to process such-and-such a case.” Those software people are looking at the law and thinking, “well, I suppose that means you’ve got to do this.” What would make this easier is if there was a government-published, machine-readable version of legislative rules that made the use of definitions and repeated words clear and made the “if” this “then” that language clear.
If that machine-readable version, which may or may not be the definitive version, if it goes to court, for example, the court decides what it means and courts can decide all sorts of things. But at least we publish the English text, we don’t take it to a judge and ask the judge to say what it means, and then publish that. We’ve published the English text of the legislation, with that we could publish that text with some markup language that meant that people were all starting from the same place when they try to produce a version which they will use for compliance.
To me, it’s much more interesting to think about the person who isn’t trying to automate a procedure, but the person who is just trying to figure out, “What am I supposed to do? What does this law mean?”
If we could make that easier, perhaps an app on your phone, that when you put in your human assessment of the facts it works out what the result is, using the machine-readable rules, and you as a human then interpret that result.
Heather Mason:
The whole point of this is to make it easier for people to understand the law and to present it in a way that is more familiar. Most people are going to be more comfortable with using an app on their phone to find out what they can and can’t do than having to go and read several long pieces of potentially complex legislation.
Hudson Hollister:
We can all work towards that future together. I do appreciate your willingness to discuss this with me and we look forward to working together to bring about that future.
