October 28, 2020
Ep. #5, Flux with Michael Bridgen of Weaveworks
In episode 5 of The Kubelist Podcast, Marc Campbell is joined by Michael Bridgen of Weaveworks. They unpack Flux, the GitOps operator for Ku...
In episode 6 of Unintended Consequences, Yoz and Kim speak with civic design expert Cyd Harrell. They discuss the elements of scale that government excels at, how technology influences the speed of policy change, and why innovation isn’t always the best goal.
About the Guests
Yoz Grahame: So I've been reading the book.
The thing that really impressed me about it is how practical it is.
And it talks about the high level theory, but it's great for saying--
For finally answering the questions that we hear so often which is, " Oh, that's really interesting, I've always wanted to get involved in government and civic tech, what do I do?"
And now I can just say, "Here. Read this. This is literally the guide."
Cyd Harrell: I am so happy because when I decided to do this, the actual use case that I as a good UX person wrote out for myself was exactly that, right?
We all get that question two or three times a week.
I'd love to get involved, but what can I do?
And I wanted it to exactly be the object that somebody like you with experience could hand to that person and say, here, start here.
Kim Harrison: Can I ask what may sound like a silly question?
Kim: Why would somebody want to go into civic tech?
Because as somebody who came out of school, and I went to school in the Bay Area.
And so when we think about the industry, we're thinking about this flashy startups that are moving fast, we're not thinking about civic tech.
Cyd: I think that's right, yeah.
But people want to go into it often because that great idea of going into tech to make the world a better place. That doesn't always working out so great with the splashy startups, and the marketplaces, and the gig work apps, and all of that. And so as people start to seek more impact with these really cool, valuable skills that they have, civic tech has a direction that people start to turn.
And it is certainly hard, and frustrating, and a long haul, and yet it's incredibly rewarding and fun.
And there are a lot of really fantastic people work within it.
Yoz: What you just said about fantastic people is, I've been very lucky in my career to work with some great people, but I've never worked with as many people dedicated to working together towards a mission, as in government.
Yoz: And that is something I've not found anywhere before or since. It's amazing.
Kim: When I was an undergrad, I actually did work in government.
I worked for the state level government in California.
Not in tech, I worked for an environmental group. I studied science.
And I felt exactly that. I felt like these are people who want to be here because it's hard if you often don't have what you need to get it done.
So there was a lot of creative thinking around, like I have to get to that point and I only have half the things I need, how do I accomplish the rest of it?
Cyd: That is so much of it, right?
Yes, I only have half the things I need.
And if I were trying to do a project of this scale at Google I'd probably have about 10 times as many people.
And yet it's really urgent that this get done.
So how do we make good choices and carefully put our energies in the right place and do this together?
Yoz: So that's actually a good question there is.
What are the elements of scale when it comes to a project?
It's obvious that you're lacking say technology resources or certain budgetary resources, but what are the elements of scale that government excels at?
Cyd: Well, government generally has a mandate to serve all of the constituents of wherever it is.
And even if you're talking about a mid-sized city, that's pretty significant scale.
And when you get to talking about the federal government and its mission to serve 330 or so million people.
That's a kind of scale that even large scale tech companies think is important.
But there's one that I think we completely ignore in the private sector tech industry that I think is really important and shows up a lot in our work in government, which is timescale.
So institutions are things that act on us and on society, not just at scale across all of us, but over time. You can think about there are smaller institutions, like a local restaurant that exists for 50 years and influences the culinary culture. And then everybody in a certain area eats out and has personal milestones out, and all of this. That's a small institution. It influences people at small scale and over time.
But when you think about something like how government works and who is included in the set of constituents that government actually works for on what different types of arenas and needs, and how those things shift, you're really talking about long timescales.
So sometimes I put it that in civic tech, we think about changing things on kind of a technology timeframe a few years, a few months.
Technologies come and go in 18 months, two years.
But we're trying to use some of those same techniques to change institutions that have existed for centuries.
And when you think about the kind of scale that matters, we tend to think about breadth as opposed to depth and length.
Yoz: Yeah, that's fascinating. I think it's amazing to see.
Yeah, I'm used to experience with startups and even larger companies, but still dealing with aspects of an organization where a rule has been in place since before your parents were born is a whole other level of scale.
And that kind of dedication to the long-term is amazing.
Cyd: Yeah, and it's always so interesting to talk with long-term public servants who have lived through multiple presidential or gubernatorial administrations, and seen policies shift and figure it out how to pursue missions across, just the comings and goings of mirror administrations--
And how to hold a kind of non-partisan, officially at least, mission orientedness through that to pursue goals that take longer to get there.
I really, really respect that.
Kim: Do you ever have moments where you feel like there's a rule that's in place that we disagree with, but we have to support it with the technology, and I wish I could wiggle the rule a little bit and like I don't know.
Is there ever a mismatch between what you have to do and your own feelings about that, and also the technology supporting it?
Cyd: All the time.
I mean, so first of all, some challenges don't have a technology solution.
Some things really don't.
I was lucky enough to work last year in may and June on this research project about people accessing unemployment benefits in the beginning of the pandemic.
So when employment crashed significantly, all of these state level systems were overwhelmed and you started to see that the technology was falling over, the servers were overloaded, the websites weren't designed to be flexible so that new policies could be implemented quickly.
But also the policies embedded a lot of assumptions that stopped holding true, and maybe were never true in the first place.
And maybe it came from bad discriminatory, often racist places in the first place, like people are trying to scam the unemployment system.
And so the system should be designed to prevent scamming.
It should be kind of hard to get benefits.
Well, that doesn't make any sense in the hot phase of a pandemic, if it ever did.
But you can't change it with technology.
And you could maybe tweak the technology to do a better job if you could get some kind of emergency policy into place.
So we're really not trying to closely examine every request for assistance in this extraordinary once in a lifetime situation.
But you still would need that, the technology on its own you can't just sort of make it easier because the issue is policy.
And part of the aim of that study was to show policymakers what was actually going on so that they could craft policies that would do better.
Kim: Okay, so that's interesting to me, 'cause I feel like we've had a conversation with someone else recently about exactly this.
That often we think about technology as solving the issue when really these are people problems and the technology can maybe help us do it faster or at a larger scale, but we have to still solve the people problem.
Cyd: Right, we have to, and some of those methods that you use to solve people problems aren't ones that technologists tend to be real comfortable with.
They're explicitly political methods.
They're advocacy, they're activism, they're kinds of ways that technology can help with those things.
But in the end it's demonstrating a power and consensus for something new, and getting the people with the power to implement it, to do it.
And then you can use technology to actually make a public interface for it.
As long as you can use technology to demonstrate something like that earlier, but you really won't change much if you don't also work on people issues.
Yoz: But are there ways that technology has brought to the civic tech world and to government?
Or necessarily the technology itself has brought, but ways to do policy evolution faster, better, maybe in hand with implementation that you've seen the possible?
Cyd: I think there's a lot of promise in that.
People in academia experiment with things like expressing policies as code.
There was a cool blog post by our former colleague, Alex Soble, about specifically that.
If the policy is expressed as a set of rules that are code, then when the policy changes you can quickly.
There's definitely a lot of people from the design world who are working to bring design-oriented methods into the tool set that policy makers have.
But I also think we have an obligation to respect some of the tool set that policy makers have.
There are reasons why they have the tools they have and use the tools they have.
The cycles can be slow by that technology timescale that we think about.
What do you mean it's going to take two years to see if it's successful?
We want to know if the sprint was successful for two weeks.
But it doesn't necessarily make sense.
One of the things I think about with this is that the tech industry as a whole has kind of gotten into an optimization stage for certain things like say websites and apps.
And so yeah, if you were working really closely on the famous Google different blues for links, question or what your buttons should look like and AB testing, little bits of microcopy at a broad scale of users.
You can figure out the answers to those kinds of questions pretty quickly.
Those are optimization questions, really foundational questions.
You usually can't get a solid answer to and certainly agreement on a solid answer on those kinds of timeframes.
Yoz: That's really interesting 'cause there's a blog post I've often looked back to, I think it was Mike Bracken from one of the founders of Government Digital Service in the U.K. who talked about kind of co-evolving policy and implementation together.
But it's really valuable point you're making that policy can't move that fast, use or at least the information that you get back from it won't be useful enough.
Is that kind of what you're saying?
Cyd: Sure, a large degree.
I also do think a bunch of these tools are useful, but it's one of the biggest pieces of tech arrogance, the idea that we just kind of need to take over policy with our current methods.
Yoz: We can be arrogant occasionally, can't we?
Cyd: We can. And I will let you go about my user research skills and abilities too.
But I think there were some really thoughtful work being done on how to do it, and it's being done closely with policy people and it's taking time to figure this out.
And I also think that's probably okay.
Kim: There is a quote, either on the page where you're talking about your book or in the book where you say, "Innovation is a flawed framework for change."
Cyd: Yes, I do think that. I don't know if you want specific about that?
Yoz: We got a bunch of questions about that.
Cyd: So first of all, newer isn't always better.
Just because something is newer, or more current, or more up-to-date, or more high tech doesn't necessarily mean that it better serves its purpose, especially when we're talking about a public purpose.
I always think about the examples with ballots, and how paper and ink is an enormously powerful technology in that arena.
When we think about the idea that we should have paper ballots in a lot of instances, although of course they have some accessibility problems themselves.
But in many cases, paper ballots are one of the most secure things.
Printing is a really durable technology.
They actually use really advanced printing technology to print ballots, right?
It's customizing the ballot for the exact block on which you live and the racism which you're qualified to vote.
And barcoding, and so that it can be tracked and all of that.
So it's this kind of advanced versions of this really old technology, but would you call it innovative?
No. And often, I don't know if you have, but I haven't worked on anything cutting edge in government that was like technically cutting edge. Have you?
Yoz: I don't think so.
I remember there were loads of things that we would have to do that, especially one of the worst when it comes from stakeholders, is stakeholders who are not technical but see something technical and go, ooh, let's do that.
The most common one was let's have a mobile app.
And no, you don't want a mobile app. You want something that everybody can use.
Cyd: Now, it's, let's have a chatbot.
Let's have a chatbot. I've heard that they can have machine learning chatbots.
Yoz: Yes, machine learning.
Yoz: Can we spot what frauds this tend to look like?
What's that thing where you measure people's heads to decide if they're criminal or not?
Cyd: Yeah, that's sound like it would be subject to bias at all or nevermind.
What partners I work with at the court are really enthusiastic about chatbots.
And they have heard from vendors that these chatbots we'll be able to learn to answer people's legal questions.
Yoz: Oh, God. And that's really the thing about trying to encode policy as code.
Is a very optimistic, idealistic, and frankly unrealistic in many cases.
Cyd: Right, and that goes along with sort of like, and then we're going to use advanced technology to implement it and everything is going to be great.
I think about there's this group, U.S. Digital Response, that spun up to help governments during a pandemic if they were struggling with technology.
And a lot of what they ended up doing was helping people implement Zoom or calendaring appointments for things, or old fairly solid technology used in ways that hadn't been used before.
So that's innovation. But the idea that we should put innovation out there as a goal I'm very uncomfortable with.
Yoz: Oh, completely. The stuff that I worked on that was more leading edge.
Firstly, cloud.gov working with the Cloud Foundry people.
And the cloud.gov team soared away to make provision of technical internet infrastructure much easier.
And so jumped on that and has helped the Cloud Foundry effort considerably.
But the other one which I think is much more relevant is the U.S. Web Design Standards. which is--
Yoz: Oh, well, do you want to. 'Cause you talk about it in the book.
Cyd: I was going to say we're actually building on those.
I work at the courts, we're building a judicial design system based on U.S. standards.
And it wasn't the first design system.
Wasn't the first civic design system, but the government's ability to make it available at scale and to support a team that can sustain it suddenly puts this capability in the hands of if they wanted every government organization in the country that they didn't have before.
Yoz: It's an amazing effort in a few ways, part of which is because despite the name it is not something that is mandated.
It's something that you can choose to use.
And because it's a choice people put a lot of effort into making it work, but also it's an open source collaborative evolution.
There are contributors from all over, both the public and private sphere.
And it's continually improving and as a way to use modern web components or standards technologies to make best of breed website.
Cyd: Right, and one of the things I think that ties back to the theme of this conversation is that the scale that enables, the thing that takes off the plate that might be difficult for many a government agency, which is the dealing with or making of modern web components and allowing them then to deal with the central problems that they are trying to solve at scale.
Yoz: It's great to be able to have one fewer question to answer.
Cyd: Yeah, there are so many.
Yoz: So let's talk about the many projects you've been involved in or I've heard about.
The things that we're most interested in as the name of our podcast implies is unintended consequences.
And I've heard different stories from different places, but the larger the scale the more likely you will see unintended consequences happen.
And so government work is one of the best places to see it.
Are there any particular instances that really taught you anything or were just interesting?
Cyd: So I think, one of the things that the government does, right, is try to ameliorate the unintended consequences of private sector action.
But sometimes I think it ends up with the same thing.
I'm very intrigued and scared, and maybe a little optimistic about this whole question of vaccine passports that's going around right now.
This is one of these things that must be at a society wide scale, possibly a worldwide scale if it's going to work at all.
It seemed right possibilities for unintended consequences, particularly on fairness and discrimination, and how long would we need such a thing.
And so that open question really puts it in a puzzling category for me, because if it's sort of, well, we're actually going to pretty much get the pandemic closed out by late this year.
Are we going to spin up that level of infrastructure for six months?
And how much does it add to our safety to have this thing?
And in particular, I've been talking with people who are designing these things and reporters for reporting on it.
How much do we want to prevent fakes? Like what is the actual risk factor of a particular fake? And so how much burden do we put into this idea of being able to present a vaccine proof for credential, and where do we ask for it to be presented? And it feels like it would have the unintended consequence, one of these cyberpunk dystopias where everyone has to present health criteria to go in.
And whole swaths of people are locked out of the system for days or years because of a loss of some piece of paper or some piece of bits.
I hope it doesn't.
Like I said, I've talked to some people and I think there are some smart people working on it, but there's sort of been an assumption that, well, this will be a technological thing that people will present on their smartphones.
Which access-wise we know that's not everybody has these tools.
We know that some people struggled with the vaccine registration websites, right?
Which were the last wide-scale.
So for rollout of a government tech interface that we've seen more in the last few months than in a long time.
I think this will put a smile on your face.
Canonically right, every procurement role furthers the problems that is intended to solve, right?
Yoz: Yeah. When putting together the questions we didn't actually think about procurement much.
Which is we can still stay away from the topic, generally, because I think it's worth a whole episode on its own.
But the one thing I will say is a quote from our colleague, Sasha McGee, who always came up with a kind of timeline for when people get involved in civic tech.
That's like yeah, year one I've just arrived why is everybody talking about procurement?
To year three or fours, oh my God, we've got to solve procurement. It's the most important thing.
Kim: Right, always. Wait, tell me more.
Because, okay, I was an intern.
I pretty much got to hang out 'cause I was with this environmental group.
So I got to hang out in streams and water bodies collecting data about like are public outdoors clean and safe?
Aided by Kelly PA. Like keep these spaces clean.
I wasn't always in the office. So what is this procurement?
Cyd: A lot of them are there to prevent sort of grifting, and double-dealing, and giving contracts to friends of the administration or profiteering people selling poor quality goods or services to the government when the government really needs those goods or services.
They end up in a just absolutely complex, not of rules that you have to follow to buy anything.
And a lot of them talking timescale.
And a lot of these rules were developed for when you're going to buy 2 million pencils for the government's accountants or you're going to buy fleets of trucks for a city maintenance department, yeah.
Because there's sort of this assumption that the way that you get the right pencils is you specify exactly.
We want them to be 10 inches long, and we want them to be painted yellow, and every single one has to have an eraser and we can have this percentage of flaws.
And for trucks we need them to have this kind of gearshift and these kinds of reinforced windows. I don't know what.
Whatever you would specify for trucks, but it's possible.
And then you say, well, we want you to build us a website.
And if you do this you know the teams usually learn something while building a website, or a database, or et cetera.
And if you take the approach of trying to specify every last little piece of it upfront you're probably not going to end up at the end with the thing that actually does what you hoped.
But that's the only way to get a contract through a lot of the time.
Yoz: This is why you end up with websites released in 2020 that probably announced that they work on Netscape Navigator 4.
Kim: Oh my God.
Cyd: Or have closing hours.
Cyd: Sorry, this website is offline between 6:00 pm and 9:00 am, Eastern.
Yoz: Yes, I'd forgotten about those. God. Those are real. Those really do exist.
Kim: I want to see this.
Cyd: There are several government websites that still operate this way or have batch processing.
That means they go down for a few hours a day.
Or one of the other things that happens is I have partners that I work with who can't look at their code for their website that somebody built for them because the contract was set up so that the code is owned by the vendor.
And then the vendor can charge for all kinds of.
And they're not allowed to look at it because that would break the terms of the contract.
Yoz: I'm glad we're getting to this, just because I know that there are a bunch of listeners who only think about government and technology when they see yet another procurement disaster, right?
The standard newspaper story of how some system thought some big department cost $2 billion that took 10 years and it's broken on delivery.
Cyd: I knew that I was truly a government person when I heard that the government had spent $40 million on the canonical vaccine registration system.
And it kind of worked. And I was like, why are you complaining about that?
Cyd: It is bad in a way with all these horrible stories.
And we saw colleagues of ours like, well, I can build a better website for $50 in a weekend.
Yoz: Yeah, it's amazing how they don't factor in the time, usually.
Nor kind of they don't seem to have run it past any actual users or collected any bug reports.
Kim: Well, when your users are this broad you are like tasks, you service all of them.
It's a lot different than I just made a thing where we can, now I won't name names, but like latest apps that are out there that people think are great and fun, but it's like, who does it serve?
Is it for a very small slice of the population or is it for everybody?
Cyd: Right, who does it serve?
And that's a huge concern with this vaccine passport or question, or is it a concern with the vaccine registrations.
It's really been an issue throughout.
I mean, I have been of course, professionally fascinated by everything that's gone on with COVID data and COVID government interfaces.
And then how governments have shifted other government responsibilities when everyone had to stay home.
It's been an astounding exposure of what the gaps are, and who isn't being served, and where the government has the capability to do it or not.
And I think one of the largest scale civic tech projects ever, I think I'm allowed to call it this, even though it was done by journalists, is the COVID tracking project where effectively journalists from the Atlantic got together with some other people and ended up having a group of about 300-plus volunteers.
They were the primary source of data on the COVID pandemic throughout 2020 in the United States, because the government wasn't doing what it should have been doing.
And they took that on and kind of patched it. And I thought they did a tremendous job.
And they've been incredibly transparent about working, I think, now to archive everything from that project to show how this happened, and how they did it, and let historians see it going forward.
But that may be the biggest volunteer patch I've ever seen. Scale-wise you guys, I don't know.
Yoz: That sounds like, I mean, I think there were a bunch of things where it calls into question, like what is government responsibility and what isn't.
So you have giant volunteer movements such as OpenStreetMap is one of the big common ones.
And then you go, well, hang on, is that actually the government's job to do mapping?
And that is one of those really hazy areas, especially like I come from Britain and the ordinance survey there sells its data commercially.
And so the question is, what is public, what's private?
Cyd: Right, what value do you add? But this was very clear.
Yoz: And something else amazing, just remind me about the tracking and vaccine registration in particular.
When I signed up for my vaccine shot, it was because FEMA was doing a run through Oakland.
And they used for the sign up a survey site that looked nothing to do with health at all.
It was some 10-year-old funded or survey startups thing.
And part of that made me really worried because I wasn't sure how much I could trust it.
And on the other hand, there's part of me hugely admiring that they were using existing mature, tried-and-tested technology really cheaply.
Cyd: Yes, exactly. There were a bunch of them that were on Eventbrite in various places.
And people were like, oh my God, is Eventbrite secure for your health information?
And not much of it is health information, unless you had to say, do you have a preexisting condition?
But in most places you just had to check a box yes or no, instead of anything specific.
And that was the sort of adaptability that I think we in civic tech have said we want government to show, we want government to take pieces that exist and glue them together, and make things that work for the public.
The problem was when it didn't work for the public, then people really weren't happy with it. It's fair.
Yoz: Yeah. It's certainly such a better alternative than reinventing the wheel that happens so much in so many government projects.
Yoz: Especially so much if it's driven by vendors who are incentivized.
Obviously not by public good, but by profit.