Technology

New post for long-time advocate of better digital government

The Volker Alliance has added a prominent federal technologist to its board. She was the deputy U.S. chief technology officer and founded Code For America.

Tom Temin@tteminWFED

March 4, 2024 2:15 pm

12 min read

The Volker Alliance, a premier good-government group, has added a prominent federal technologist to its board. She was the deputy U.S. chief technology officer and founded Code For America, a non-profit that helps government at all levels with digital challenges. For an update, the Federal Drive with Tom Temin spoke with Jennifer Pahlka.

Interview Transcript:

Tom Temin And since starting code for America, which you’re no longer associated with, but it’s kind of got a life of its own. What have you been up to?

Jennifer Pahlka I had the brilliance to step down about six weeks before the shutdown for the pandemic. And my idea was to write a book, which I did. But with the pandemic chaos, I also ended up helping match technologists with governments who needed them during those first couple weeks of the pandemic. And that became something called U.S. Digital Response, which should not be confused with United States Digital Service, which I helped stand up in the White House in 2013-2014. So that was an amazing ride, I’m still on the board of USDR as well. Then went and wrote my book and did some other consulting during the time. And now I’m a senior fellow at both the Niskanen Center and the Federation of American Scientists, and just excited to join the board of the Volcker Alliance, which I’ve just had such high regard for so long.

Join us May 6 at 1 p.m. EST for Federal News Network's Industry Exchange Data where we'll explore how you can employ data to modernize your agency. | Register today!

Tom Temin And with respect to that helping government when the pandemic hit, it must have been somewhat satisfying to see the technology base that was in place already at the federal government that allowed it to quickly pivot, in most cases to everybody teleworking and in some sense not really missing a beat.

Jennifer Pahlka I think there were so many things that government at all levels, did incredibly well during the pandemic that we take for granted, because that’s what we always do with government. If it goes well, we take it for granted, and if it doesn’t, we are quite concerned. Yes, there were the missteps too. So one of the other things I got pulled into in the first summer of the pandemic was a strike team, as they called it. I call it a task force because it sounds less violent for the state of California’s backlog of unemployment insurance claims. And that end up being the first three chapters of my book. And I thought it was it was a great lesson for me to learn, and that I could share through the book about what really underpins the problems of government technology. The technology looks like it’s the problem, but there are much deeper dynamics underlying them that we need to grapple with.

Tom Temin And tell us more about that book, because it sounds like something that the current crew and past crews and future crews ought to read, maybe to help the government keep moving along the digital journey.

Jennifer Pahlka It’s a book that I think a lot of people think is about technology. It has a QR code on the cover in a flag. So it’s clear that we are talking about U.S. government here and some degree of patriotism. But it’s really about sort of what I came to conclude is a driving force of our dysfunction in government technology, which is fundamentally this idea, that policy is this thing over here and the delivery, the implementation of that policy is something separate, that separate people do, and they don’t really talk to the policymakers. And of course, we think of it as sort of a waterfall, a cascade, a linear process from maybe Congress creates the law, gets handed down to agencies and policymakers, etc.. And down at the bottom of this big waterfall, you have the implementers. Well it turns out for many years people have been challenging those assumptions and realizing that those two things should not be thought of as separate. That when we think of them as separate, we are causing ourselves much more pain than we really need to. I’m excited to see so many people start to grapple with these ideas and really put them into practice and stop sort of saying, let’s fix this on the edge here and really go to the core cause.

Tom Temin Yeah. So the implication is that good technology implementation really starts at the collaboration stage between people that make policy and people that have to implement policy.

Jennifer Pahlka Yeah. I’ll give you one example from the book, though there are several. This is a local and state level issue. But when I was at code for America many years ago, we started on the problem of clearing criminal records where legalization of marijuana in many states had meant that somebody who has a past marijuana felony no longer should have it on their record so that they can access things like jobs and housing. But it’s a year long paperwork process that most people can’t persist through. It’s just so much sludge and paperwork and really, there’s no need for it. And so we figured out that if these are just records in a database, you can query the database, find all the people who are eligible for that expungement and clear them in bulk. So that’s automatic expungement. But the problem is some laws are written in such a way that there’s no way to query the database and find all of those people. We had a law here in California called Prop 47 that was written to reclassify burglaries under $950, and I think it was in commercial locations. Well, you cannot query the database. You’d have to go looking in every single person’s file and try to read handwriting of a cop who took that case and see if they noted what kind of camera was stolen, and try to figure out if that was $950. So these laws are simply unable to be automated, and therefore they’re really going to have very little effect on the people. But if you consult with the people who understand the implementation of the law before you write it, you might make different choices in what you actually make expungement so that the law can have a real effect. And that’s a kind of thinking that is rare, but it’s starting to grow. And I’m just really happy to see that we can get implementation kind of all the way up front at the process instead of all the way at the end.

Tom Temin We were speaking with Jennifer Pahlka. She’s former U.S. Chief Technology Officer and now author of Recoding America. And that example of where the law doesn’t fit, really, with what the policy actually is to say in this case, it’s okay to expunge these records. This is something I was talking to Senator Warner about not long ago, with respect to artificial intelligence and the many ways that Congress could act on that by simply doing little block and tackling legislation to make things in sync in the way you’ve just described for that particular function of expungement in another domain, which gets to the big issue, that Congress can’t even do those little things anymore when everyone agrees this is the policy, all we need to do is change clause A, subchapter one, paragraph B of this law, and it’ll all match up. They don’t even do that. That must be frustrating.

Jennifer Pahlka Well I think we’re going to have opportunities for the kinds of legislation that I’m interested in, because this stuff is nonpartisan. There’s no culture war at risk here. We’re really just talking about stuff that benefits everyone. And I think there will be windows for it. But no, I can’t fix congressional dysfunction.

Read more: Technology

Tom Temin Anyone that could would be a genius and would be showered with a crown of America. Let me ask you this. Since you were in government, golly, a decade ago, that’s kind of light years in a technology sense. And now we have a company called Nvidia, been around for ages. People used to be happy to get one of their cards in your computer because it could really rev up the gaming. Now they are worth a couple of trillion dollars thanks to artificial intelligence, which has come on the scene in a big way. What are your thoughts about good ways for government to bring this in to the whole digital effort? Really, that’s never ending.

Jennifer Pahlka Yeah. I used to work in the video game business. I ran the Game Developers Conference for eight years, and so I knew Nvidia back in the day when it was gaming that drove their business. Artificial intelligence is fascinating to me, and I think I came to it a little bit late because I had been fatigued by the hype around blockchain. But when I started really paying attention to it, I started to see that this is a profound opportunity to bring our government forward. And I’ll tell you one thing that brought me along. I was visiting the Department of Labor in New Jersey about the spring before I did my book, and ChatGPT had just come out. Now, you had this great team at the [Department of Labor (DOL)] and also the new Jersey Office of Innovation, wonderful folks that they were working to make unemployment insurance just better every day. It’s a fantastic strategy. They’re not going for some big procurement, they’re just bringing on people who know what they’re doing and really fixing it week by week. One of the things that this designer was doing was rewriting the letters that people get, their emails that are letters you get in the mail about your claim about adjudication. And they’re so hard to understand. They’re written in legalese, and it’s one of the reasons people don’t reply. And then you get a longer backlog. Well, she had been rewriting them for a sort of eighth grade, ninth grade level and then going to the policy team and saying, is this still correct? And then if it is, let’s bold the call to action, put it in big letters, then we read the letter and then people start to interact with unemployment insurance a lot easier.

Jennifer Pahlka Well, ChatGPT had just come out and she was just feeding those letters into ChatGPT with the prompt, rewrite this so I can understand it better. Now. She still went to the policy team and checked it with them. This is not giving over decision making to AI. It’s an AI as an assist to somebody. And I asked her, so is it really helping you? She said, I think we’re getting through these letters about 4 or 5 times as fast as we used to. It’s not an automatic process, it’s an assist. And I thought, that is a fantastic use of AI. We should not be getting things like that. We should just let people do this. Now there are things that are going to need some review. And of course, the AI executive order has called out some of those things. I think we need to be really careful that the guards that are put in place through things like the AI EO for really risky applications don’t get applied to these really low risk, high value applications. And the use of them to make letter simpler is just one example. There are dozens, hundreds of others where we really want to enable. We want to put our foot on the gas, not on the brakes, I think in, in those areas. The one area that I think we do need to have caution about, that is not the one that most people talk about, like we don’t want to give over decision making to AI. I think that that is much less important than people think, and there’s so many other applications. But we have such complex rules and regulations like unemployment insurance. We’ve been adding rules and regs for 90 years. We never take them away. So you’ve got thousands and thousands of pages of regs that cover this pretty simple program. I think a lot of people are excited about AI’s ability to sort of manage that complexity. And I want people instead to be excited about AI’s ability to help us simplify that complexity, not just like, ok, we’ll be able to get through this, but here are some proposed simplifications that state legislatures and Congress ought to really take seriously, so that if you’re trying to do unemployment insurance in the next downturn, you have 100 pages of regs to deal with, not 9,000.

Tom Temin And I want to bridge to another use of AI and the generative AI, which is to make computer programs. And as someone whose endeavors usually have the word coding in them, one technologist said the other day, well, thanks to the generative AI, English is the new programing language. What’s your thought on that?

Jennifer Pahlka I think that is part of a trend that’s been happening for a long time. Technology is sort of the ability to make things that other people can use has been democratized for a long time. AI is for the next phase in it. But go back to that time I talked about right when the shut down sent us all home and all these governments were reaching out to volunteers through USDR. It happened to be a time also when no code, low code tools were becoming available. And about a half of the requests that we got from state and local governments at the time. Can you stand up a form for emergency rental assistance, for example. Those were actually pretty easily filled, not by some fancy tech team, but by a couple of volunteers teaching local government officials themselves how to use something like Airtable. It’s just that much more powerful than Excel. And you can really actually run a benefits program. That’s not too advanced off of it. And I think that is part generally of how we’re going to make all of these things that used to be very specialized, really possible for anybody to do.