Language selection

Search

CSPS Data Demo Week: Regulatory Artificial Intelligence (Video)

Description: The team's platform of demonstration projects allows for experimentation with disruptive technologies to support analysts and regulators in navigating this time-consuming and labour-intensive field. Tune in for a demonstration of three unique prototypes, followed by a Q&A on the implications of this type of work across the public service and how artificial intelligence (AI) and machine learning could play a bigger role in our work moving forward.

CSPS Data Demo Week: Regulatory Artificial Intelligence

Date: April 29, 2021

Duration: 01:00:28

Resolution: 1080p


Transcript

[The animated white Canada School of Public Service logo appears on a purple background. Its pages turn, opening it like a book. A maple leaf appears in the middle of the book that also resembles a flag with curvy lines beneath. Text is beside it reads: Webcast | Webdiffusion.]

[It fades out, replaced two title screens side by side in English and French. At the top, it shows three green maple leaves, each made of different textures. Text on Screen reads:

CSPS Data Demo Week

Regulatory Artificial Intelligence

GC data community

[It fades out, replaced by a Zoom video call. The video window is filled with a Zoom call. Brenda, a woman with long grey hair and a detailed lacy button-down shirt wears a headset. She sits by a large window with a venetian blind.]

Brenda Baxter: Hi, everybody, my name is Brenda Baxter. I'm the director general for Workplace Directorate in the Labour Program at Employment and Social Development Canada, and I am really excited to be the moderator for today's session that's going to showcase some really, really exciting applications for regulatory artificial intelligence led by the Canada School of Public Service and their Innovation and Policy Services Branch. We have a really great event lined up for you today, so I hope you're all looking forward to it.

Before we actually do start the sessions, I just want to acknowledge that I am actually in Ottawa on the traditional unceded territory of the Anishinaabe people, and I recognize that we will all work in different places, we are working in different places across the country, and therefore you work on a different traditional Indigenous territory. I encourage you all to take a moment to reflect on this, and to honour the land, please take a moment to consider the first peoples on the land where you are.

[Three more video panels appear, putting Brenda's in the top left. On the top right sits Neil Bouwer, a man wearing glasses and a short-sleeved checkered button down. He sits in front of a purple Zoom background that features the CSPS logo, the Government of Canada logo and reads "CSPS Data Demo Week" in English and French. On the bottom left sits Mylaine Des Rosiers, a woman with voluminous shoulder length hair and straight across bangs. Behind her is a couch, art on the fall, and a window bench. On the bottom right sits Joseph Kokou, a man with a close haircut and a short salt-and-pepper beard. Two pieces of art sit on the wall behind him.]

Brenda Baxter: So maybe just before we start, I just want to mention that we're going to be taking questions and answers throughout the session and you're going to be using the application called Wooclap, and you can submit your questions through this event by going to www.wooclap, which is w-o-o-c-l-a-p, dot com and use the event code DDW4, in order to submit your questions. First I just want to say that Neil Bouwer invited me to, to help moderate the session today 'cause the Labour Program and myself have been involved in working with the School of Public Service on some of these really exciting regulatory initiatives. The Labour Program is a pretty small regulator, but we think of ourselves as pretty, pretty mighty, and we, you know, we take a lot, on a lot of regulatory work and try to do sort of the innovations that we can with the capacity that we have in the Labour Program. But some of these really exciting initiatives you're going to see today are, are bigger than things that we could take on ourselves in the Labour Program. So we are truly benefiting from the collaboration with the School of Public Service, and I think all regulators across the public service will benefit from these innovations as well and even really enjoy the discussions that, that we have across all the participants, because you share ideas and more ideas come to the forefront, so it gives us an opportunity to look at what else can come forward. How we can build on and advance the ideas and the innovations that are already out there.

We are really, really happy to be part of this, this work, and I hope you will, you will be as excited as we are once you see the demonstrations today. I am going to be turning over to some colleagues that are going to be doing the demonstration, so I'm joined by Neil Bouwer, Mylaine Des Rosiers and Joseph Kokou. And firstly, so looking forward to the presentations they are going to be making and first, I just want to emphasize again to get on that Wooclap application and get ready to submit your questions, because we're really excited to make the demos, but also make sure that you have all the information that you want or you need with regard to these exciting, exciting initiatives. And with that, I will turn it over to Neil, and thanks, Neil, for the invitation.

[Neil's video panel fills the screen.]

Neil Bouwer: Fantastic. Thank you, Brenda. Thank you, everyone, for being here today. I'm going to speak in English like most of the other presenters, too, but there is simultaneous interpretation. Also, we're all happy to take questions in French, either on Wooclap or otherwise. So please, feel free to engage in the language of your choice. I'm going to try to share my screen here, let's see, let's see how this goes.

[Neil shares his screen, and his video panel shrinks over to a small bar on the right-hand side of the screen. On his screenshare, is a CSPS Data Demo Week title page.]

Neil Bouwer: Great. Hopefully you can see me, and I will bring up a little presentation here to start off with. First of all, just to say that this is part of a series of Data Demo Week, so we've been having data demonstrations all week and we've got another one tomorrow. We're really happy to be to be here with you.

[Neil flicks through presentation slides to one that reads "Public Sector Innovation Demonstration Projects." Text underneath reads "What it is: an incubator that explores new technologies with partners through proofs of concepts to help solve horizontal issues using new data analytics techniques. How it works: Demonstration projects provide a unique learning opportunity. PSI works and learns with you by:

  • Providing project and contract capacity
  • Drawing lessons learned
  • Sharing costs and distributing risks across a range of portfolios."

Beside this text is a graphic demonstrating the "project life cycle."]

Neil Bouwer: This, as Brenda said, this project comes to you from a group at the Canada School of Public Service who really works in a consortium of departments and agencies, that includes ESDC, the Labour Program, other parts of ESDC, many other departments and agencies, and this consortium is collaborating on different interesting Pathfinder projects. So this slide sort of shows you the cycle that we use in terms of the ideation and then moving to the conception and then the implementation of some different projects.

[The slide moves forward, showing now only the project lifecycle graphic. It features a curved line punctuated by various icons, labelled "Empathize, Define," and "Ideate." The line flows into a circle at a point labelled "Experiment." Arrows point around the cycle from "Experiment," leading to "Build, Test," and "Learn."]

Neil Bouwer: We've got a portfolio of projects and I won't go through these all today, but I just want to advertise that we are working on projects together around access to information, using AI to make access to information requests easier and more efficient to respond to. We're working with Departments and Statistics Canada on linking data from different sources to make regulatory life and services and program delivery easier and more data informed. We're also working on projects around regulatory impact analysis statements to use those as a source of data.

[Neil clicks, and text boxes on the graphic highlight. A box linked to "Build" reads "rules as code." A box linked to the space between "Test" and "Learn" reads "Regulatory Evaluation Platform (REP)" and a box linked to a question mark between "Learn" and "Experiment" reads "Incorporation by Reference (IBR)."]

Neil Bouwer: But today I'm going to be highlighting three, so I will be speaking to something called Incorporation by Reference, which I'll describe, then our colleagues from Transport who are taking the lead in an area called Regulatory Evaluation Platform, so they will describe that project and then I'll come back and describe a little bit about the Rules as Code project. And we're going to have demonstrations for each of these so that you cannot just hear about it, but actually see, see what's involved.

[The slide changes. It reads "Deep Dive — Incorporation by Reference Search Tool (IBR)." On the right-hand side of the screen, a text box reads "Follow Along!" above a QR code and another box that contains a user login field.]

Neil Bouwer: What brings these projects together is the fact that they all are based on regulations. This is a regulatory AI demonstration, so they all use the regulations as their data. The actual words in the regulation is a source of data for all of these projects. That makes it really interesting, so the data isn't like numbers or some kind of database, it's the actual, it's the law, it's the regulations that are actually used. The first thing I'm going to show you is this tool about incorporation by reference. You can actually scan this, I should point up here, over there.

[Neil points toward the shared screen.]

Neil Bouwer: You can scan that QR code on your phone or on your device, and you can actually go to the demo that I'm going to give you and you can actually play around yourself. There's a login user ID as well as a password here, if you want to go in and just see what I'm talking about while I'm, while I'm doing it. So first of all, let me just tell you what incorporation by reference is. So, in a regulation, sometimes a regulation points to a third-party standard. It might point to something from the Canadian Standards Association or something for you know some other, an ISO standard or something else, and that's really important because it sort of delegates a legal authority to another party that publishes a standard. So, right now, when you want to try to find out incorporations by reference, you have to do it by hand. There are actually paralegals and legal folks who actually read regulations, they circle the part that is an incorporation by reference, they build a list of these incorporations by reference, and then that's used for departments to testify in front of a parliamentary committee or answer questions about incorporation by reference. It takes about 13 minutes to do this, is what was estimated, but there's over 3000 of these in the Government of Canada. So, it takes literally thousands of person hours to make a complete list of incorporations by reference.

So, the tool that we are... that I'm going to show you just now automates this. We've trained an algorithm to search regulations, identify these incorporations by reference, and then automate that process so that it can be updated as the regulations are updated, and basically it can be done sort of instantaneously rather than, rather than by hand. So, if you follow this link, you'll get to this website here.

[A website appears. In the top left-hand corner, a logo identifies it as "SA-IBR." Multiple regulation search engines cover the page.]

Neil Bouwer: This is on Lemay.ai's platform and you can search the different federal regulations, and in particular, you can go right in here by title and you can choose any old regulation. I could... Let's see…

[He types into the main search bar.]

Neil Bouwer: I can choose the Occupational Health and Safety Regulations, is a nice big one, and then the tool basically then pulls in the full text of the regulation.

[A list of regulations appears below the search bar, and Neil selects an option. It takes him to a long policy page with various snippets highlighted. A right-hand sidebar lists the highlighted snippets. He scrolls through.]

Neil Bouwer: So, as I said before, the regulation is the data, so on the left-hand side here, you can see all of the full text of the regulation, so it's not written on the sort of a Canada Gazette type format, but it's pulled into this this format, and then on the right-hand side, you can see where the algorithm is, is picking out different areas that it thinks are incorporations by reference, and it actually highlights those actually in the text. So here's one, here if I click on it, it gives you a little more information, I can go down to the text.

[He selects one.]

Here we go, so this is a reference on towers, antennas and antenna support structures, it's... the regulation is here and then it refers to this, and for sure, this looks like a standard, it looks like Canada Standards Association standard and it's got the title of it here, and then the clue here is it's amended from time to time, which that, which is a really important legal point. The algorithm is guessing, is it incorporation by reference? It's guessed yes, that's correct. Is it static? Is it a document that never changes or is it ambulatory or something which is amended from time to time? Yes, and it's guessed that correctly, so that's, that's awesome. Now, if a human disagreed, they could say so and the algorithm would learn from that, and it would remember the choice. Also, the tool is able to scrape the Canadian Standards Association website because the Canadian Standards Association makes this available and then actually return the document. That functionality is turned off at the moment, but it can actually bring up the document from the Canadian Standards Association, indicate whether it's available in English and French, and also indicate the price that is charged by the source organization for that, if any. And those are important because often parliamentarians or other people ask those questions. They want to know whether the documents incorporated in law are available in English in France and how much they cost. So, this saves thousands of person hours.

[Neil returns to the search results.]

Neil Bouwer: It's a pretty simple tool, as you can see, pretty easy to use, and I invite you to kind of try it out. It's a demo, it's a prototype, so there's a lot more work that needs to be done on it but it's up there now if you want to, if you want to give it a try. Let me me take us back.

[Neil returns to the slideshow. The new slide reads "Deep Dive: Regulatory Evaluation Platform" in English and French.]

Then the next project I'm just going to have this over in a moment to Mylaine and to Joseph on the regulatory evaluation platform, but I just want to say in introducing it for my part, that this is another project that uses those regulations as data itself and it's another one of these Pathfinder projects, which is of interest to a regulator like Transport, but is of interest to many departments as well, so a really cool Pathfinder project. On that note, Mylaine, over to you.

[The call returns to its four video panel format, with Neil using the slideshow as his background. All videos momentarily pixelate. Mylaine's panel takes over the screen.]

Mylaine Des Rosiers: Merci beaucoup, Neil, pour l'introduction. Hello, everybody, my name is Mylaine Des Rosiers, I'm the director of the Office of Regulatory Policy and Innovation over at Transport Canada.

[A purple text box in the bottom left corner identifies her:Mylaine Des Rosiers, Transport Canada.]

Mylaine Des Rosiers: I'm joined by my esteemed colleague, Joseph Kokou. He's chief of the CBA, the cost-benefit analysis team in Transport Canada. So today we want to talk to you a little bit about how we're using technology to modernize regulatory functions here at Transport Canada and first offer you a little bit of context with the slide number three of our presentation.

[A screen is shared, taking up the majority of the screen. Mylaine's panel moves to the righthand side of the screen. The presentation slide on the shared screen reads: "Government of Canada Regulatory Modernization Initiatives. Horizontal issues highlighted during the consultations by the Treasury Board Secretariat in 2018 and 2019." Bullet points below read:

  • "Less prescriptive regulations
  • Foster regulatory harmonization
  • Improved stakeholder engagement
  • Alleviate regulatory overlap and consider cumulative burden."

A long link sits below the bullet points.]

Mylaine Des Rosiers: Over the last decade, the stakeholders have been quite vocal about what they want regulators to address in terms of irritants, and to of specific relevance to us today is addressing the red tape, and considering and measuring cumulative burden on the industry as a result of regulation.

[The slide changes. The new slide is titled "Considering & Measuring Cumulative Burden." It reads: "Industry and business stakeholders expressed concerns that:

  • Regulators do not consider the cumulative impact of regulation on competitiveness and economic growth
  • Cumulative burden should be measured by doing a sector-by-sector analysis and engaging industry experts and academics."]

Mylaine Des Rosiers: So, to address this sounds quite easy, however, it is not. From an industry's perspective, the government is one big department developing regulations that adds up to their cost of doing business. And up until now, regulators had no efficient means to comprehend and quantify the cumulative impact of several layers of regulations impacting specific sectors of the industry. Also, there's no standardized taxonomy to quantify this impact, because there's simply too many variables. So, at Transport Canada, we decided to take this challenge on, and we figured it out.

We first need to identify all the regulatory requirements across the federal corpus that would apply to specific sector of the industry, and then we have to calculate the probable impact of these requirements on this specific sector. This requires an economic model, and this is where we need experts to join in with academics to develop those models. Before you can apply those economic models, you need to apply them to the data. Where is the data? So, this is where we faced a big challenge. So next slide.

[The slide changes. It's titled "Challenges Faced by Regulators." A list below it reads:

  • Inconsistent and unclear taxonomy for ABB
  • No standardized methodology to assess cumulative impact of regulations
  • Managing regulatory corpus is resources intensive
  • TC large regulatory footprint]

Mylaine Des Rosiers: Every year, at Transport Canada, we have analysts that have to do a manual count of all the administrative regulatory burden in our regulations in order to meet our obligation as per the Red Tape Reduction Act. In 2017, we had a count of over 30,000. The problem is, it is not only resource intensive because we have a lot of regulations here at Transport Canada, but the taxonomy itself is not clear and as a result it is not applied in a consistent manner. I would give this task to five different analysts, and they would come up with five different numbers. But the main challenge that we face as regulators is a direct result of the large inventory of regulations. Next slide, please.

[The slide changes. The new slide is titled "Current regulatory footprint." Four bubbles on the slide each contain facts. Mylaine reads from the slide.]

Mylaine Des Rosiers: So, here's the picture, the Minister of Transport is responsible for 51 acts and 450 regulations, Transport Canada is therefore contributing to a significant regulatory footprint at the federal level. In 2018, the regulatory corpus counted 3196 regulations at the federal level alone, which amounted to a total of seven million words and the average length of a typical regulation being 2206 words. Now at TC we have 450 of these. Think of the bean counter, in 2020 that number went up. We're now over 3700 federal regulations. So, the challenge here was to find a way to mine that data. And so what we did is develop the regulatory evaluation platform and would like to show you what it looks like. Joseph, the honour is yours.

[The screen share stops, and all four panelists reappear for a moment. A new screenshare starts, showing a web browser open to a webpage. A maple leaf and "TC REP" sit in the top left corner. There is a left-hand menu sidebar, and regulation statistics fill most of the page.]

Mylaine Des Rosiers: So, what you're seeing is when you log in the regulatory evaluation platform, this is where, this is where you land. Joseph, can you take it?

Joseph Kokou: Yeah, yeah. Can you see? Thank you Mylaine, good introduction. Yeah, so we developed that digital tool for you guys. The things I will do here is to work through a little bit on that digital tool. We call that regulatory evaluation platform. So, the first page or the first tab you have here, as you can see, you have Transport Canada. In the future we are planning to add other departments. You also have the number of regulations we already included in that platform. We have more than 400 regulations at Transport Canada, but right now we have only 139 included in that system. Here you can see for just Transport Canada, the number of requirements for the industry is more than 24K. If you add to that, you know, the rest of the department, you guys, you can just imagine the number of requirements we are asking to industry. And we break down the requirements into different aspects. One of them is the Prohibition, which is a little bit different from Restrictions, and we also have Operational Requirement and also, Admin Burden. And the... as Mylaine mentioned, one of our objectives is to try to connect regulatory data to industry, and then we use NAICS code, which is North American Industry Classification System, and then we try to collect the data... we collect through the platform we developed, and we connect that to industry. And you can see for the corpus of Transport Canada regulation we have here, we impact more than 94% of the industry. Let me move to tab two.

[At the sidebar menu, Joseph clicks on "Regulatory Management." A dialogue box with a large red X inside a circle appears along with an error message.]

Joseph Kokou: Oh, something wrong.

[He exits the dialogue box and the image freezes for a moment.]

Joseph Kokou: Can you see that? No?

[He clicks around, and the image pixelates for a few moments. The image resolves, showing the "Regulatory Management" page. Across the top of the page, different selectable filters are sorted by "Department, Regulation, Category, NAICS Code," and "NAICS Code Digits." Below the filters sit more statistics.

Joseph Kokou: Ok, here we go. Now, the things I'm showing to you here, you can see we... as I said, Transport Canada, the number of regulations we have and we have different category here, the categories, the admin Burden, we also have... let me click on that.

[He selects the dropdown menu under "Category" and mouses around a list of categories, all selected already.]

Joseph Kokou: Operational requirement and so on, and we have the list of NAICS code you guys can see here as well. And let's move down a little bit. So, this... So let me change this one.

[Below the statistics boxes, Joesph pressed a button marked "2." He scrolls down the page, revealing a bar graph entitled "Count of Provisions by Regulation." He reads the labels on the graph.]

Joseph Kokou: Here you can see by regulation, we can just focus on their five top regulations and you can see which amongst, you know, the corpus of Transport Canada regulation we have. We can see the regulation that mainly impact on Canadians. So, the first, Canadian Aviation is the biggest one, and we also have lifesaving equipment, and you have motor vehicle safety regulation. And if I move things to somewhere here and you will see the NAICS code.

[He clicks a button marked "3", and a new graph appears, entitled "Count of Provisions by NAICS Code."]

Joseph Kokou: Here the system is showing to you the NAICS code that are mainly, I mean, the NAICS code, I mean, the industry or the sectors that are mainly impacted by that Transport Canada regulation, and you can... Here we show you just the five, the top five NAICS code. you can change that and you have more than five, let's say seven.

[He adjusts a slide above the graph, and the graph refreshes.]

Joseph Kokou: You can see the platform shows you seven NAICS code here, and the other thing we have here is the impact management.

[Joseph clicks "Impact Management" on the page's sidebar. The Impact Management page loads, showing the same set of selectable filters. He searches the "Regulation" filter bar.]

Joseph Kokou: Impact management, let's select one of the regulations here, let's see commercial... Yeah, commercial vehicle driver's hours services. When you select that... And then you can go through the regulation and let's see, select Section 1.1.

[In the "Sections" filter, Joseph selects "1.1" from a dropdown menu. Details regarding section 1.1 appear in a small table below the filter bar.]

Joseph Kokou: When you select that section, you can see that under that section, the operational requirements are the one of the requirements we are putting on industry.

[He navigates back to the "Regulatory Management" page.]

Joseph Kokou: And then we can come back to the previous tab and so on. So, it's a just two-minute presentation, I want, that's the thing, I just want to tell you. [Indistinct] I don't know, I never seen this number, or this huge number of regulatory data stored somewhere. And in the future, we will, you know, work to store more than just to Transport Canada regulatory data, but more... the other Federal Department of Regulatory Data. So Mylaine, the floor is yours.

[The screenshare stops. The screen returns to the four video panels, then Mylaine's panel fills the screen as she speaks.]

Mylaine Des Rosiers: Thank you so very much, Joseph. So, this is to show you what we wanted to do. So, this system allowed us to do three things. First of all, it enables us regulators to timely collect and analyze large amounts of regulatory data without having to go through regulations, by regulations or legislation, by legislation. The system will mine across all federal regulations eventually, and the system allows the identification of all the administrative burden in all regulations and automates the count, first of all.

Second of all, the system identifies all the regulatory requirements that are applicable to specific industry by sectors using NAICS code. So, it links regulatory requirements with a North American industry classification system number. This gets to be automated throughout the regulatory inventory. And the third outcome out of this platform is that it offers us an interactive web-based interface, and it had to be easy to navigate, for me, and it had to offer efficient visualization of search results.

[A screen share pops up, filling most of the screen with Mylaine's presentation. Her video panel sits on the screen's right-hand side. The current slide is entitled "Objectives and Outcomes." It reads "Objective — Develop a solution that would mine the large amount of regulations and generate regulatory data needed for analysis. Outcomes — Enable regulators to timely collect and analyze large amounts of regulatory data:

  • Automate the ABB count
  • Automate the identification of regulatory requirements applicable to specific industry sectors using NAICs codes
  • Interactive web-based interface to visualize search results
  • Economic formulas to quantify cumulative impact of regulations on industry sectors."]

Mylaine Des Rosiers: This to us was a crucial component, which is in fact half of the equation to be able to quantify the cumulative impact of our regulation on specific sector of the industry, because now we would have the data that we would then apply the economic model that would allow us to assess the accumulation of regulations on the specific sector of the industry. Transport Canada is also developing those economic models under the leadership of Joseph Kokou.

So, of course this is only a first phase, we started this phase in January, we just finished in March. So, under three months that's what we were able to produce. The next phase will be to refine, expand and enhance. We want to refine because we want to refine the next results to be as precise as possible. We want to expand because we want to include the not only all of TC 450 regulations, we want to include all the federal regulations because that's how you calculate the cumulative impact on industry, not just Transport Canada regulations. We also want to tag those regulations with NAICS code and other codes such as UN number, and national occupational code system. We also want to include RIASs in there because there's a lot of information in the RIAS that would be necessary for our economists to do this assessment. We want to include legislation mapped out with departments and naval regulations. We also want to expand the search with IBR, because when you think about a sound management of regulatory inventory, it's not just legislation or regulations. Like Neil said, it's also all those technical standards that we are referring in our regulation, and the industry needs to comply to. And unfortunately, up till now, somebody had to count manually all those IBR and 3700 regulations. So, we need to stop this insanity and automate this count.

We also wanted to include those definitions, because if for a sound and accountable management of regulatory inventory, we also have to be mindful of how many times we define the same word differently throughout the regulation. That's something that we're thinking about, too. We want to enhance as well, the user experience. We want to have a French version of the interface, for example, and we want to expand the search function. So, to conclude, in Transport Canada and me personally, we believe in developing solutions that are user-centric, and that benefit the entire regulatory community. Having been the regulator for over 20 years, I can tell you that I'm quite happy with what we've been able to produce.

[The screen share ends, and Mylaine's video panel fills the screen briefly before the other panels rejoin her.]

Mylaine Des Rosiers: Because to me it's going to end 20 years of misery going through regulations to find information that we're looking for to conduct sound, evidence-based assessment and analysis. Thank you very much for your time. Your feedback is important. Our contact information is at the very end of our deck. Neil, thank you again for the invitation to present to you today. Merci beaucoup, merci Joseph. Merci.

Neil Bouwer: Fantastic, great presentation guys. We have one more thing to show you before we turn it back to Brenda for, for the Q&A session on Wooclap. So, let me just see if I can pull things up here.

[Neil's video panel fills the frame. He shrinks his own image down to the bottom left-hand corner. He sets a slide deck as his background. The current slide reads "Deep Dive — Rules as Code I: Vacation Pay." A text box beside it reads "Follow Along!" and a QR code sits below it.]

Neil Bouwer: So firstly, I'm going to be talking about a "Rules as Code," I'll explain what it means and I'm also going to show you something, but if you want to jump ahead, you can scan in this QR code and just go right to the website and you'll see the, the thing that I'm about to demo for you. First, though, I do need to explain sort of what Rules as Code is.

[The slide changes. The next one is entitled "About Rules as Code." It reads:

"Rules as code is a process of drafting rules in code, language understood by machines. The process has the potential to help us make better rules. Today, government creates rules in both official languages using a variety of instruments, including legislation, regulation, policy, etc. Writing code in addition to drafting in both official languages could:

  • Improve the quality of rules by testing the effects of rules in real time and identifying gaps between a rule's intent and how it is implemented
  • Support more consistent interpretation rules and greater accessibility of the law to Canadians
  • Reduce cost of compliance
  • Offer new possibilities to improve service delivery using automation and other techniques.

If government organizations published their rules as code, these services could connect directly to this authoritative source, eliminating the need to encode rules into the software."]

Neil Bouwer: So, Rules as Code is basically the idea that you take regulations and you actually publish them, not just in English, not just in French, but you actually included in a coded form, a machine-readable form. You take the actual logic of the regulations, and you code it. So why would you do this? Well, you do this because it would reduce the cost of compliance, because there are many regulated parties that would love to be able to read regulations in a machine-readable way, because maybe they've got software that they want to be compliant, or their compliance systems want- they want to use for regulatory compliance. But also, it makes the rules more easily interpretable, it makes them clear, it removes any ambiguities, which is really useful. Also, it helps government to simulate the impacts of regulatory changes. Because if we code the rules and have a micro simulation engine, we can actually tinker with the rules and then simulate the outcome. So, there's all kinds of interesting ways that this might be applicable.

[The slide changes, and Neil moves his image out of the way to the bottom right-hand corner. The new slide is entitled "Example of Rules as Code." It shows two graphics. The first graphic is labelled "Today" and shows the Income Tax Act. An arrow points from the act to a picture of a laptop labelled "Electronic Tax Filing company." An arrow points from it to another laptop labelled "Interface for Canadians. The second graphic is label "Future: Rules as code." It shows the interface for Canadians directly connecting to "Source code for Income Tax Act," via by wifi symbols labelled "Direct connection with source code and official interpretation."]

Neil Bouwer: So, in fact, many of us already sort of do a kind of a Rules as Code. For those of us that fill out your taxes in a software, that tax software has basically taken the Income Tax Act and all of the regulations and turned it into code so that when you're filling in the fields with your income taxes, that software makes sure that what you're filling in meets the regulatory requirements, and also where there's got to be arithmetic, you know, the pluses and minuses and the multiplications. It sort of does those automatically because they've been written in code. Imagine that the government actually published its own regulations, and requirements, and laws in a code form, so that it wasn't up to a company to figure out how to do that in a software, but actually the government itself published rules in a code form. The project I'm about to show you...

[The slide changes.]

Neil Bouwer: Whoops, that's jumping ahead a little bit.

[Neil switches back to the previous slide and moves to an interface titled "Rules as Code Discovery Project. It shows fillable fields, but the text labelling them is distorted.]

Neil Bouwer: The project I am going to show you right now is... comes out of the Labour Program at ESDC, and they took a piece of their regulatory requirements and encoded it. That was really useful because it was a discussion between regulators and coders to say "Well, what exactly do you mean when you talk about your regulatory requirements?" So, in this case, it is looking- Brenda could probably explain this better than me, but it's basically the part of the regulation that looks at your entitlement to vacation leave.

[Neil fills out the fields.]

Neil Bouwer: It kind of depends when you started working, let's say that... Well, let's just say that it was in July of 2016. Let's say you make $50,000, then this will basically calculate the amount of vacation time that you're entitled to this year based on the number of years in the year that have transpired. You can also... so the regulations themselves spell this out in words, and those words need to be interpreted and each variable has to have its own definition in law but instead of having a lawyer read it for you, you can have this, you can have it coded and so the software can easily pick it up. And then if, you know, if the person was terminated, say, on April 1st, tough day for them. Let's say they also took some, some medical leave, let's say that it was, you know, well, more than 17 weeks.

[A blue box below the fields entitled "Total Employment Days" generates statistics regarding the worker.]

Neil Bouwer: Then you can see the new calculation comes out with an amount payable based on the vacation leave entitlement. You can play around with the system, and you can put in different variables, and if you were to go back to the original regulation, you'll see that basically what we've done here is just coded those regulations in and created code. It basically helps you simulate and if you were to say, well, what happened if the person wasn't terminated, how would that, how would that impact on the compensation? Or if they did not take medical leave, or if they had a different salary in the year in question, so then you can see that the prototype here basically calculates that for them.

It's easy to see from this demo, I hope, how that would be used for something like vacation leave or leave entitlements, and also how it could be used to help service individuals. So, instead of somebody phoning a department and asking about the legal interpretation, if they actually had the code and a solution like this, then they could come to a website and they could stimulate their own enquiries and get the answers that they wanted. But also, they could actually build this into their own payroll systems and into their own computer systems with definitive and authoritative code from the Government of Canada. So that's the idea there.

[Neil switches from the interface back to the presentation. As he speaks, he moves to the next slide entitled "Rules as Code II: Motor Vehicle OT." It reads "The project participants worked through an intensive 3-week sprint to convert the Motor Vehicle Operators Hours of Work regulation into code, create a public API, a Policy Difference Engine (Micro Simulator), and collect lessons learned and user experience along the way." Below the text, a timeline shows the team's progress.]

Neil Bouwer: We are working on a next phase of this project. ESDC really is a leader in this area and the next one we're working on is around overtime provisions for motor vehicle operators. We've actually just finished a sprint where we have turned the motor vehicle overtime regulations into pseudocode. We've created a decision tree around that to help folks understand better. It required unpacking the regulation, talking through all of the ambiguities and all of the different circumstances into it, and then the idea is that then we can publish that code and those regulatory requirements as code.

[The slide changes. The new slide is entitled "Explainer Video — Rules as Code: Concept & Benefits" in both English and French, and features two large QR codes.]

Neil Bouwer: So, if you're still wondering, well, what is this code? The team put together a great video on to explain Rules as Code. So once again, if you want to, here's a couple of QR codes you can go to. It'll take you to a YouTube video where there's an explanation of what Rules as Code is and sort of the project that we used in this case. Maybe before I leave, I should maybe just do a bit of an advertisement for a Rules as Code event that's happening tomorrow, actually.

[Neil switches from his presentation to scrolling through event pages on the Government of Canada website.]

Neil Bouwer: If you're interested in Rules as Code, you want to hear about the latest and greatest on Rules as Code, you can go to the Canada School of Public Service website, it's tomorrow at 1:30 and you can register for it there. There is also another data event tomorrow on financial risk with Mindbridge at 11:00 a.m. If you're interested, we can fill your day up. And my last commercial, Brenda I promise, before I turn it over to you, we have on May 3rd a series on Trust and public, and the public service, and it's actually with Edelman that puts out the Edelman Trust Index, which could be really interesting to folks. So, you can check all of those out on this Canada School of Public Service website. So, with that, Brenda, that's what we wanted to show today and happy to take questions.

[The screen Brenda's panel fills the screen.]

Brenda Baxter: Great. Well, thank you, Neil, Mylaine, and Joseph. Great presentations, and I've got a whole bunch of thoughts going on in my head about how to use these innovations and how to even... What the possibilities are or opportunities that come out of these, to even expand them. So great, great work and I think others as are... looked to be as excited as I am about these and have lots of questions. Just a reminder to folks to put your questions into Wooclap.com. Pass code is DDW4, so Wooclap is w-o-o-c-l-a-p dot com. And so I will turn over to some of the questions. First one maybe to Joseph and Mylaine, is about how Transport Canada stakeholders have responded to the regulatory evaluation platform. Have you shared this talk to industry, academics and other groups about the pilot you put in place?

[All four panelists return to the screen. Each panelist's panel fills the screen as they speak.]

Mylaine Des Rosiers: So, the answer is we have not yet, because we started the first phase in January and barely finished the end of March, and we first wanted to see if we could pull it off, frankly. And also we would like to have a more fulsome platform before we do share, because I suspect that it will generate a lot of interest, and the first question that we're going to be asked is if the industry will be able to access this platform. Because for industry stakeholders, one of the most crucial thing that they would need is to actually have a quick and efficient access to the information on what kind of regulatory requirements they need to comply to. Right? So, we want to be able to be in a position to say that we're addressing this as well. But it is my hope that it will be well received.

[A purple text box in the bottom left corner identifies him:Joseph Kokou, Transport Canada.]

Joseph Kokou: Yeah, I just want to add the fact that behind the scenes are working with some academic to develop that community impact model. So, I think for that model we need mainly the mathematical models and then we need to work with those people who are more versed on how to develop, how to quantify the economic impact of the regulation on Canadians.

[All four panelists return to the screen.]

Brenda Baxter: Great, well, thanks so much, and I know that I'm sure industry will be really excited, and especially when it's expanded to include more regulations than just Transport Canada, to really see that cumulative burden on a particular sector. So, thank you for that, and thanks for the question. Neil, there was a question about whether registration was needed to use the incorporation by reference tool, and I think you mentioned the user ID and pass code, so anything else to add?

Neil Bouwer: Well, first of all, I just want to say on the academic point that there is a... Université de Montréal has academics that are working on the ingestion of the RIAS statement and so that they're pretty well versed in this area, the University of Ottawa as well has some researchers that have been working on the regs as a data set. The regulations themselves are accessible by Justice Canada, who sort of got this ball rolling a few years ago. Pretty much anyone can take a look at this as a data source if they're interested. In terms of the- so the incorporation by reference tool really is a prototype, we do want to improve it and put it into production at some point. And so right now, you can log in with those credentials, so it should be open to you to log in. That won't be the case forever, because we're doing it really just for this Data Demo event today, so we'll probably turn off that account in the coming days. If you want to try it out, I would, I would try it now and if you're having issues with that, maybe in the Wooclap, just leave your email address and ask that somebody get back to you and we can have somebody email you if you're having technical challenges.

Brenda Baxter: Great, and so thanks for that, Neil. Now, moving on, there is another question about would the system, and I'm presuming this is a system, the demo that you provided Mylaine with both the regulatory evaluation platform. So, they're asking, would the system allow for automatic updates to the regulations after there is an amendment?

[Mylaine's panel fills the screen.]

Mylaine Des Rosiers: That's a great question. And, yes, we plan to address this in subsequent phases, depending on funding of course. But this is some... it's an important feature actually, that this platform needs. Because if you want to calculate the cumulative impact, you need to have some sort of baseline and the regulation is amended continuously, so it's something that the system would have to do. Absolutely.

[All four panelists return to the screen.]

Neil Bouwer: Yeah, and just on that, I know the team right now is updating it every two weeks based on the information from Justice. So, that might change just as the prototype evolves, but right now I think it's every two weeks, which is pretty good for regulatory changes.

Brenda Baxter: Great and so Neil, we have a question on Rules as Code. Someone's asking how a department can get started on Rules as Code? Is there a standing offer to access the developers that can create something like the ESDC product?

[Neil's panel fills the screen.]

Neil Bouwer: Interesting question. Well, it's a very new area, I would say the first thing that you should do is probably attend tomorrow's event, if you're interested in Rules as Code, and check out the videos that I mentioned. All of those include contact information for the folks that are working on the project who would be best placed to guide you in that. Right now, the Rules as Code project that I mentioned was done with a, with a group called HabitatSeven, and I know that the current project is something that the team is building on its own, so there isn't yet any kind of standing offer like that. Although this may fit the definition of AI and of course, there is a procurement vehicle for AI projects which could be accessed.

[All four panelists return to the screen.]

Brenda Baxter: Great, thanks, and I just want to say that we really enjoyed working on the Rules as Code project. And the second part we're working on, we're doing in conjunction with our stakeholders. So, we're working with an organization called the Canadian Trucking Alliance to look at the regulations we have around overtime calculation, because we get a lot of complaints around that. And having the regulations written as code, as Neal said, would allow potentially certain employers to write this into their HR systems or even to have an application developed that some of the truck drivers would be able to use in their phone to do a little bit of calculation around what the overtime would be that they're allowed. So really exciting and I'm sure generating a lot of excitement about other departments wanting to do this thing, so this is fantastic. One question we've received is a more general question about, this presentation today is showcasing some of the regulatory artificial intelligence work, and I guess the question is really around how are we defining artificial intelligence? 'Cause I think people may have a particular perspective on what artificial intelligence is and how it actually works. The question is, how are these considered artificial intelligence? And maybe I'll turn that to you, Neil.

[Neil's panel fills the screen.]

Neil Bouwer: Sure. Yeah, for sure the term artificial intelligence covers a lot in people's imagination. The proper word for most projects is really augmented intelligence because these tools are not meant to be automated decision making or to replace the judgement of people. All the three cases that we've shown here, and frankly most of the projects that I've seen are actually augmented intelligence. They're assisting a human to make a decision in the same way that a calculator, or an Excel spreadsheet, or anything else would assist a human in making a decision. So first of all, we should properly call them augmented intelligence. The other thing is, of course, it covers a wide band, so at the simple end of things, we have robotic process automation, and macros that just do the tasks you tell them to do, then you have data analytics or, you know, statistical analysis kind of in the middle, and then you have proper machine learning and competitive learning algorithms, and the more advanced AI that you hear talked about as well. And so, we're kind of in the middle here in the sense that these, like the different algorithms are using enhanced search, they're using text analytics, but they are not yet necessarily learning as they go. In the case of incorporation by reference, we did have training data sets, so the algorithm learned and then as it's corrected, it will continue to learn, so that is machine learning properly understood. But in some of these other areas, we are basically just applying enhanced search features and text analytics, and then it's not really machine learning in the same sense, but it still fits the definition of artificial intelligence. I hope, Brenda, that's clear.

[All four panelists return to the screen.]

Brenda Baxter: It's clear to me and I think you may have sort of touched on a response to another question that came in that is: is the primary AI functionality to calculate the regulation metadata and will this be expanded? I think you touched on a bit of it, but if you want to expand on that answer.

Neil Bouwer: Sure, and I know Mylaine and Joseph will want to add to this because, you know, the quality of the data are really important. When Justice released the data, they did some tagging of different parts of the data, which is great, but there's way more we would love to do. As Mylaine said, it would be great to be able to tag the regulatory requirements, the administrative requirements, the regulatory burden. Some of that can be done by algorithm, which is great, but some of that, if it was done by hand, would also be great. So actually, one of the projects I didn't talk about is called Metadata, and there's a number of departments under the leadership of ECCC and we're actually having human beings tag the regulations with this metadata, which will help us to do even more analysis. So that's a way that we're using humans to enhance the quality of the data in that case.

Brenda Baxter: So, I'll turn it over to Mylaine and Joseph to see if they want to add to that, but there is an additional question that came in with regard to the regulatory evaluation platform. The question is, will it be able to measure the "what if" or proposed changes and their impact on regulatory burden? So right now, you're looking at the current regulations and what the burden is, but if you wanted to make a change to that regulation, would the, with the system you put in place allow you to sort of test what that burden would be?

Mylaine Des Rosiers: Yes, that's the intent, and I'll pass the mic over to Joseph because he'll be able to explain in a more precise and expert way than me.

[Neil's panel fills the screen.]

Joseph Kokou: Actually, we decide to use a kind of step-by-step method. It's a complex field and our first goal is to build the foundation, and then we are trying to build something that in the future you can just enter the system and then check how much that will cost to the industry when I'm developing X, Y, Z regulations. So, that's actually, that's our main goal.

[Mylaine's panel fills the screen.]

Mylaine Des Rosiers: Yeah, so the answer is yes, because once we develop the formulas and the models and we do have the data at our disposal, we will be able to do that type of analysis because we'll have a baseline, we'll have something to compare it with. We'll even be able to go back in time and see, maybe a couple of years before, what was the previous version of the regulation and compare it to the amendment that we're proposing now to see what would be the difference on the... In terms of impact on the industry. That's something a little bit of, a bit more complex, it could take a bit more time, but it's certainly the intent, absolutely.

[All four panelists return to the screen.]

Brenda Baxter: Well, in it and I'm sure that in the future when this is a little bit more sophisticated and you're able to look at all the regulations across departments, to be able to look at a regulatory change in one department and see what the cumulative burden would be, but also how it may impact other, the regulations in other departments would be fantastic, and I know our stakeholders were really excited about that. One other question- oh sorry, go ahead Joseph.

Joseph Kokou: I think talking about the future, we know that not only the federal government is publishing regulation and so on. We have provincial and territorial government, so our plan is to see how we can take into account those regulations as well, and then to see the big picture of the costs of the burden that we are putting on industry.

Brenda Baxter: Great, really exciting. And again, all the opportunities and possibilities coming out of these initiatives are amazing. There is another question, and I think it, I opened it up to any of you. It's with regard to the platforms or approaches that are presented, and it's looking not just at external, so those that we regulate outside of the public service, but saying how can we apply these to the Treasury Board policies and guidelines? Has anyone considered looking at TB administrative policies, which are also based on rules and requirements?

[Neil's panel fills the screen.]

Neil Bouwer: Well, it's a really good question, because really what we're talking about, even though we say these are regulatory, really any requirement can be thought of as in the same way. So, we don't have a pilot project in this area yet, but I think it would be really interesting to look at internal regulations as well as guidance that departments and agencies put out, because it's not always in the regulations themselves, often there's manuals or templates and other things that might be useful.

Brenda Baxter: Great.

[All four panelists return to the screen.]

Mylaine Des Rosiers: And if I, if I may build on that...

Brenda Baxter: Yeah.

[Mylaine's panel fills the screen. A purple text box in the bottom left corner identifies her:Mylaine Des Rosiers, Transport Canada.]

Mylaine Des Rosiers: I think this is a fantastic question because we tend to want to focus on the costs on the external stakeholders. But if you'll notice, in the 3700 federal regulations that we have out there, there are several of these requirements that impose a burden on the ministers, which is us really. And so, one of the things that we wanted to address in the subsequent phases is to identify those requirements so that we can have a proper idea of the added pressure on the departmental resources when we develop future regulatory amendments, whenever the regulation says "the minister shall" that's, that's not imposed on the industry, it's imposed on departmental resources. So, we should have a mean to assess [indistinct, audio cuts out], so we're looking into that as well.

[All four panelists return to the screen.]

Brenda Baxter: That's great. Thanks, Mylaine, for that, and so next question is one around the potential to code privacy regulations from requirements to clarify privacy policy for individuals. So, really interesting idea there. Any thoughts on that? Neil, maybe?

[Neil's panel fills the screen.]

Neil Bouwer: Well, just to say that I think we would welcome the, the idea of other regulators or other regulatory areas getting involved in Rules as Code. I think one of the insights that we found for Rules as Code, obviously the more prescriptive your requirement is, or your regulation is, the more amenable it is to Rules as Code, and so that's one thing that I wanted to say. The other thing I wanted to say is also think about where your pain points are. If you're spending all your time answering people's questions about regulatory requirements, where those regulatory requirements are clear, or at least it's clear to you, not very clear in the regs, that also is a good area because it will save us time. Or like incorporation by reference, if you're spending time combing through regulations to do something, or if you're combing through regulations on administrative burden accounts or other things, think about those pain points. And those are the things we can automate, and by doing this Pathfinder projects, we can maybe automate those things that are pain points for sort of like what Mylaine was describing.

[All four panelists return to the screen.]

Brenda Baxter: Wow, and there's another really good point that's been raised about looking at the Rules as Code and saying we just had budget 2021 being presented, which proposes lots of legislative and potentially regulatory changes, and so the question is: is anyone sort of combing through that to figure out which one's potential regulations could be written and in English, French and in code? And that would actually help with the understanding of the regulatory changes and the application of the regulatory changes.

Neil Bouwer: Yeah, what a great idea for the budget and even, you know, how many of us go through the budget searching for certain terms, or searching for certain dollar amounts attached to certain programs? If that were released in a machine-readable form, you would actually have the relationships between when a number is cited under a certain year, for a certain program as proposed spending, or maybe it's a report on actual spending. To release that in a machine-readable way would help lots of people. They wouldn't have to read 700 pages to find what they're looking for and they could also do advanced analysis. The same could be said for a budget implementation bill, or a regulatory modernization act that follows the budget or anything else. So, really cool idea.

[Joseph's panel briefly leaves the call and rejoins.]

Brenda Baxter: Yeah, and I'm just noting from a time perspective, we have five minutes left, so we have to hold a few minutes to wrap up, but we might have an opportunity for one more question. And so, just wondering, is the intended purpose of the regulatory evaluation platform to replace aspects of the regulatory analysis or just provide a starting point for analysts to reference? Mylaine or Joseph, do you want to take that?

[Mylaine's panel fills the screen.]

Mylaine Des Rosiers: OK, so I have... I always have a strong opinion about that question because to me, technology will never replace sound analysis. It's there to assist. We're not there yet where a system will be sophisticated enough to generate the type of analysis that we need to conduct those type of assessments, because it's more than economic impact. It's more than calculating the cost of industry, it's more than that. It's assessing the impact and the benefits as well, of some regulation. So, I'm not sure we're there yet where we're going to replace the analyst's assessment with technology. However, this platform will save us enormous time in generating and automating search through a vast amount of data that is unstructured. So, I would say it is a fantastic starting point, yes. Joseph, did you want to...?

[Joseph's panel fills the screen. A purple text box in the bottom left corner identifies him:Joseph Kokou, Transport Canada.]

Joseph Kokou: Yeah, actually, Mylaine and I, we had that discussion in the past, how you can replace some of the activities, for example, my team is doing now, building triages, when it's come to amend regulation. Most of the time when you are developing the cost-benefit analysis, you have to, you know, do things from scratch. So, maybe in the future, because we have this system, we can store some of the previous amendment of regulation like RIASs and so on. When it's come to amend that you take that previous, you know, amendment and then you make the changes you have to make to your regulation to have it published. So, I think, yes, we are not... The case we are doing now, as in Mylaine says, is to have somewhere we store data, and then from there we can build on that.

[All four panelists return to the screen.]

Brenda Baxter: Thanks, Joseph. And so, we have two more minutes left, I'll be quick on the wrap-up, but there weren't any questions about the incorporation by reference, and Neil I'll give you 30 seconds to see what to... talk about what the next steps are with regard to this tool.

[Neil's panel fills the screen.]

Neil Bouwer: Thanks, Brenda. Yeah, we would like to bring this into production so that it's a tool that analysts can use, and then as Mylaine and Joseph alluded to, we would also like to bring that data into things like the regulatory evaluation platform as well, so that functionality can be brought in to other platforms, so that's really exciting. And then we're going to bring other data sets in, as they both mentioned, like the RIAS and others, so this is a growing kind of ecosystem and we hope to have platforms available to folks to use, ultimately to make our jobs easier as we regulate and deliver regulatory services.

[Brenda's panel fills the screen. A purple text box in the bottom left corner identifies her:Brenda Baxter, Employment and Social Development Canada.]

Brenda Baxter: Well So, I just want to thank all of you, Neil, Mylaine and Joseph, for your presentations. And thank you to all those that supported the setting up of this event today, and all those participants. There's been tons of questions, great discussion, and I hope, everyone, you enjoyed the event and there is an electronic evaluation form, so I'd encourage you to fill that out. And as Neil mentioned, the School has some additional events tomorrow. The Financial Risk Discovery with Mindbridge at 11 o'clock, GC Data community meetup at 1 o'clock, the event is available on GCconnex, not the CSPS website, and Leveraging Rules as Code tomorrow at 1:30. So I think that sort of wraps it up. Really exciting, looking forward to the continued work on these tools Neil, and how they can get expanded. So, thanks to everybody and thanks for all the participants.

[All four panelists return to the screen. Mylaine and Brenda smile. Neil gives a little wave and Joseph nods. The Zoom call fades out. The animated white Canada School of Public Service logo appears on a purple background. Its pages turn, closing it like a book. A maple leaf appears in the middle of the book that also resembles a flag with curvy lines beneath. The government of Canada Wordmark appears: the word "Canada" with a small Canadian flag waving over the final "a." The screen fades to black.]


Date modified: