HData CEO Hudson Hollister recently interviewed Xcential co-founder Bradlee Chang on Xcential’s history, the current (and future) benefits of a Data First approach to lawmaking, and the challenges of modernizing legislative and regulatory processes.
This interview was produced for Data First, Xcential’s monthly newsletter covering the modernization of lawmaking around the world. Click here to see past issues and subscribe to Data First.
HUDSON HOLLISTER: Bradlee Chang, co-founder of Xcential, thanks so much for joining us today.
BRADLEE CHANG: Thank you for having me.
HH: We’re going to spend some time talking about the “legislatures of the future,” in particular when legislatures begin drafting and managing legislation as Data First. What are the changes and modernizations that will make life easier for state legislators, for their staff, and ultimately for their constituencies?
I’m excited to get into that, but I would like to start by asking you to describe a little about Xcential and about your own history.
BC: We created Xcential for just this purpose – to bring meaning to the words that are in law. We saw that the law was done mostly as documents and we’re trying to make these documents into information. It began in 2002, with the legislature in the State of California. We automated all of their bill drafting processes, eventually their codification, and a number of other legislative processes for them.
HH: How did you come to Xcential?
BC: I was working as a technologist inside the development arm of Xerox. Some of the ventures they had were focused on XML and what XML can bring to different applications. We came upon legislation as a good application of XML technologies.
HH: You got that first contract with the State of California Office of Legislative Counsel to transform parts of the legislative process from documents into XML data format. How did that come together?
BC: We presented a proposal for the State of California where the proof of concept involved taking the U.S. Constitution and showing point-in-time analysis. What did the US Constitution look like throughout its history? We could do that by taking the words in the Constitution, in all its formats, and putting it into one data set.
Then, you have a “scroll bar through time,” and can see what the US Constitution looked like at any point. That’s what California wanted. They wanted these documents to be living documents, the law to be living law. When we can look at the law through points in time, we can dive deeper into it, and make connections within it.
HH: So Xcential, as a brand new company, won the contract and began engineering that transformation for the California state legislature. What were a couple of the biggest hurdles and achievements in the first phases of that project?
BC: The first hurdle was to make drafting as data first look like Microsoft Word. And that’s instructive. The legislature wanted to get to this place where they were thinking about information and data, but they knew that the users needed to think about words on paper, and the users knew how to use Microsoft Word. So that was our challenge – to bridge that gap between creating an underlying information model for the law, yet make it easy and accessible for the drafters to create in a familiar form.
HH: You transformed California’s laws from files made of documents using a standardized data model into a data compilation where things like point-in-time analysis are possible. Functionally speaking, what are the things that the California legislators, and the staff and offices that serve them, are able to do today that they couldn’t do before your project began in 2002?
BC: What’s interesting is once we made them comfortable with the idea that they could draft laws in a tool that was familiar, that felt like a word processor, but was creating an information model underneath it, they were willing to go further and asked us if it was possible to improve the [amending] process, because their amending process involved labor-intensive, manual instructions for how to amend bills once amendments were adopted. They had entire crews that had to work overnight to get this done.
If the amendment was adopted at three o’clock in the afternoon, by eight in the morning you had to have all the new versions of all the bills out. They had overnight crews that worked long hours and sometimes couldn’t meet the deadlines.
They were manually executing these amendments which say “on page six, line three do this.” They were literally making all those changes manually, one by one, as editors. So instead of having to do that overnight, we allowed them to pre-draft these changes by using change tracking in our LegisPro editor. We allowed them to change the words and generate the instruction amendments so that when it came time to adopt the instructions, all they had to do was pick up the version that they had edited days before and make it the official version.
HH: Which means that amendments could be automatically and instantaneously executed into the underlying bills.
BC: Right. And not only that, but that the instruction amendments are generated automatically – and with 100% accuracy.
HH: Xcential has now expanded its business worldwide and much farther beyond the state legislature of California – serving the US Congress, foreign governments, foreign legislatures, and has also moved to perform similar transformations on regulation as on legislation. Before we finish talking about that very first project that seeded your growth for the State of California, is there any functionality that you may have envisioned, knowing what was possible with a common data model for the State of California, that hasn’t yet been realized? Perhaps connecting their legislative level to the regulatory level.
BC: There’s one legislature [in California], but there are dozens of regulatory agencies that have to implement the law and there isn’t yet a connection with it. Regulatory processes are very fragmented – and that hasn’t changed yet, but it’s an opportunity.
HH: I want to shift gears a little bit to the notion of data standardization. Xcential harps on this essential focus, on the need to adopt and then apply a standardized data model to legislative materials. How do you explain [the Data First] concept to people who are perhaps legislative experts but might never have heard of XML?
BC: One thing is consistency – consistency in the language and consistency in the documents to make its meaning a clear one. In fact, coming back to the State of California, they came to us and said, “for this amending process the language we use varies, and it doesn’t need to. The variances in the language we use make it harder to execute and harder to understand what we mean. What are we supposed to do with this?” So we proposed standard language and standard grammar behind the language and computerized processes behind that and a standard data format which allowed them to make their amending language consistent. Anytime they wanted to do a specific type of thing, the language would always be the same. Everyone knew what it meant. It improved their crafting processes and improved the clarity of their bills. In some ways the customers already know they need to get there, but it’s in terms of consistency of their documents, not in technical terms. But technology can help that.
HH: It wouldn’t have been possible to adopt a standard data format without standardizing the mandatory language, and similarly, it wouldn’t have been possible to create a data model accommodating all the different possible permutations of legal language if the lawyers didn’t want to change.
BC: Yes. Actually, we had to do that when we started working with the U.S. Code [and working to develop the USLM (U.S. Legislative Markup) standard]. The U.S. Code is an existing body of language and law — we’re not going to change the words, but we have to apply consistent meaning to the different ways things are said. Part of data standardization is an underlying representation, kept in a computer form, which normalizes different ways of saying the same thing with language.
HH: For instance, mandatory text. Suppose a jurisdiction has many different ways of expressing an amendment within legislation. How would that connect through to one meaning, if it indeed is the same meaning, within the data model?
BC: Well, I’m glad you use the word “model.” One thing we have to do early on in these cycles is model the information. Find the underlying meaning of the words. If two things are said, do they mean the same thing or do they mean different things? If they mean the same thing we have to come up with a data model that can cover those two sets of language, but has the same meaning underneath.
So we need to understand the way people say things and what they mean. Once we understand the various amendments they want to represent, we can create data models for that. One of the nice things is we have done enough of these jurisdictions that we understand the commonality of these models in, for instance, implementing language – what is the action for the amendment and what is the thing you are affecting, understanding how to represent these various semantic instructions in law in a common format.
HH: We’ve discussed data modeling, but your customers are probably more interested in the functionality that you can help them build and deploy [once a Data First approach is adopted], rather than in the delightful abstractions of a data model, is that right?
BC: That’s right. They don’t care about the technology. They want to get their jobs done.
HH: Let’s talk about the ways in which the [Data First] transformation that you have engineered in data modeling helps legislators get their jobs done better and for what purpose. Earlier, I asked you to identify three different categories of functionality that might characterize state legislatures of the future. We’ve got three C’s for our audience: Contextual drafting, Comparisons, and Coordination.
[First,] what is contextual drafting and how does the adoption of a standard data model facilitate contextual drafting?
BC: Contextual drafting is the ability to describe what result you want and have the system build up the language that creates that effect. So it can be generating the bills that you want to introduce by going to the law that exists and then making changes that you would like to see to the law. This is like redlining. You take a section of law, make changes in various places, and then generate the bill. That improves the drafting process both in terms of productivity of creating these bills and also in clarity of what you’re intending to do and making sure that the bill does it.
HH: Do you think that legislators and their support offices understand that contextual drafting is possible? Do you find that contextual drafting is still very new in the legislative world?
BC: I think people feel like it’s new and scary. It’s sort of putting trust in the computer and in technology that’s just been in the hands of people — albeit knowledgeable people. I think there’s some encouragement in the successes we’ve had in the amendments to bills. We use the same process of making the change directly to the bill and generating instruction amendments. It has been successful in terms of productivity and creating consistent documents. I think people could make the leap to say it’s a similar process.
HH: Let’s talk about our second category: comparisons. What kind of comparisons? Obviously, comparison is a big deal for the legislative process. When a legislator proposes an amendment to a bill or when a bill itself changes an existing law, it’s very important to understand exactly how that change works. These processes are frustrated by practices such as moving a section to a different place within a bill and the process of creating comparisons and understanding them takes up a lot of a legislator’s time and their staff’s time.
How does the application of data standards, a Data First approach, transform comparisons?
BC: You’re right. It’s hard to do because the legislation changes in its life cycle. It changes in ways that make it difficult to understand how it’s changed. Legislators need to know that and it’s been a frustration that the process is usually very manual. They have a staff member take two stacks of paper and some pens and figure out what the differences are or ask their lead counsel to have one of their staffers generate a report. Those have always been done manually because the differences can be hard to find. Especially if you use a standard word processing approach to create the documents that you want to compare. In Word, for example, you ask for the differences, and you’ll have the whole bill – one was stricken and the whole bill was inserted – you won’t know what the changes are at any meaningful level.
By applying a Data First model to it, we can understand the identity of the provisions that are in these laws, be they bills or existing law. Understanding the identity of them can help us. And these comparison programs know what happens if they move, if they change their read number, if they have different headings, and of course, if the content changes. But all of those things can happen simultaneously. If we have an information model that understands its identity, its location, its number, it’s heading, and its content all as separate pieces of information, we can track those differences and know that the section it just moved didn’t change. So it’s the same section, we can compare them and say, don’t worry, this building change for this section of the bill didn’t change, it just moved down because another section came before it, or it didn’t change. Or, you know, it moved and you didn’t think it changed, but subtly it did.
HH: Can you identify a comparison project that is now public that exemplifies what’s made possible by comparisons in a Data First approach to legislation?
BC: A few years ago, the U.S. House enacted a rule called the Posey Rule which required these comparisons, because Mr. Posey and the rest of the House thought it was important to be able to see these comparisons. So in the US House we implemented a project to do these types of comparisons. That’s been quite popular, quite successful, and the rules recently adopted by the House expanded that concept and asked the House to move forward with comparisons in general.
HH: Let’s take our third and final area of functionality: coordination. Coordination between jurisdictions and coordination across levels. What kind of coordination is made possible by a Data First approach to legislation that wasn’t possible before?
BC: Coordination, I think, is one of the most important things. At the fundamental level, it’s coordinating the existing law with a [proposed] bill. We talked about having identities and references so that when a bill references a specific section of law or provision of law, it is known exactly what it’s connecting to.
[Moreover, a Data First approach can coordinate] law with corresponding regulations that implement that law. As a regulation is crafted, it can reference the law. A draft [regulation] could easily find the law that’s driving this new regulation, or find existing law which provides constraints on [this new] regulation. [This is] coordinating within a jurisdiction.
We would also like to see coordination [of related legislative and regulatory content] between jurisdictions: between states, [even] between countries, between a country and its constituent states.
HH: Do you think that it will someday be possible for a legislature to require an agency to draft regulations, and for the law doing so to automatically generate such regulations, or automatically generate a portion of them, or is that going too far?
BS: I think [a law] could generate the framework of the regulation..
But, but I think, by necessity, the regulation is more detailed and has more contextual knowledge of that [subject] area; [think about the complexities of] environment or healthcare.
The law [can] set forth some requirements, the base. Then through Data First drafting, the regulatory agency could provide specific [details].
HH: I’ve got one final question for you. What’s one thing that you would like legislators to know about a Data First approach to legislation? That most of them today might not know.
BC: That’s a big question.
[Legislators should know] that it’s not as far away as they might think. All of this is quite possible. All the pieces are being put into place.
[The Data First approach] is being proven in legislatures, from the U.S. [Congress] to other places, [including] states and other international jurisdictions.
This is all quite possible.
HH: And possible in this new year, 2021.
BC: Especially as we have a new year.