Scott Greenberg: This is Cool Science Radio. I’m Scott Greenberg.

Lynn Ware Peek: And I’m Lynn Ware Peek. Today on Cool Science Radio, we explore one of the defining questions of our time: Will artificial intelligence weaken democracy or help save it?

Our guest is Beth Simone Noveck. She’s a professor and director of the Burnes Center for Social Change at Northeastern University and one of the world’s leading voices on technology, governance, and civic innovation.

In her new book, Reboot: AI and the Race to Save Democracy, Noveck argues that democracy’s challenges are not just political. They’re design problems. Drawing on work from the Obama White House to governments around the globe, she shares how AI could help defend elections, improve public services, and make governments more responsive and inclusive if we choose to build those systems for the public good rather than concentrated power.

Beth, welcome to the program.

Beth Simone Noveck: Thank you so much for having me. I’m thrilled to be here.

Lynn Ware Peek: Your book opens with a pretty stark claim: democracy is in retreat and our institutions are more broken now than ever. But rather than blaming technology, you argue AI could actually save democracy.

That feels a little counterintuitive considering a lot of people right now associate AI with misinformation, surveillance, and even manipulation. What’s the most important thing you want our listeners to actually think about before they even open the book?

Beth Simone Noveck: Well, I hope they’ll open the book, of course. But if they do open it, what I hope to point out, and what I hope they’ll learn, is that a lot of what we read about in the headlines, a lot of what we’re hearing about the AI apocalypse — right? The robots are coming to take all of our jobs. They’re going to destroy the planet. They’re going to upend life as we know it.

And surely, we’re living in a time of tremendous and tumultuous change. But these are tools, like the word processors and calculators before them. AI isn’t going to do anything by itself, so to speak. It’s how we use it.

In the same way that you can use nuclear power to heat and cool our communities, or you can use it to wage war, AI is a tool. And it’s how we use it.

What I want people to take away is that we are not passive victims in this story. We can be proactive actors focusing on making the right investments, making the right choices, and thinking hard about how we use these tools to do things in the right way.

I don’t want to discount the seriousness of some of the challenges you’ve mentioned: misinformation, disinformation. Everybody’s heard about deepfakes. But we’re sometimes missing the story about the good things that we can do with these things. That’s what I’m hoping we can get into, and that people are focusing on, and that we can talk about what each of us can also do — not just governments, but also each of us as individuals.

Lynn Ware Peek: Your work history is fascinating. You probably didn’t start your career thinking about AI and democracy. You created one of the first platforms online for democratic deliberation, called Unchat, if I’m saying that correctly.

Beth Simone Noveck: That’s right.

Lynn Ware Peek: And then Peer-to-Patent, which connects scientists with policymakers to improve the patent system. I’m curious: how did those early experiments shape the ideas in this book? And what surprised you the most when they actually worked?

Beth Simone Noveck: You’re right that none of us were thinking about AI a few years ago. Some people were, but now all of us can’t get away from it. It’s all you hear about. It’s all you talk about. It’s really hard not to think about it.

But what got me started on this project was really thinking about democracy. And going back even further, when I was a student, I had the wonderful experience — you know, we got to do a summer abroad somewhere — and I was trying to decide between topics and decided, “Ooh, wouldn’t it be fun to go to Vienna and sit in coffee houses and eat cake and study what happened in Vienna, in Austria and in Germany between the wars?”

Why did this fledgling democracy after World War I devolve into fascism, dictatorship, and Nazism?

So I became interested very, very early on in these questions of what makes some democratic institutions work well and survive stress and challenge, and what makes some institutions break.

Fast forward, and I’m a first-generation child of the web, if you will. The internet is much older, but the web and the commercial internet are really part of my upbringing. I was very interested early on, and that’s how we ended up starting Unchat.

It was creating the first platform for communities to talk among themselves and to have a deliberative conversation about an issue, to run an online class, and be able to get a group together from around the world.

What was fun about that project was it wasn’t just talking online. That was something we all knew about from a very early time. There were chat rooms in those days. It was called Unchat precisely because we were creating a structured way for people to have a conversation.

The cool thing at that time — and this is now back in 1999, almost 30 years ago — was the functionality for moderating the conversation. The controls you would see on your screen would pass from user to user to user. So the person who had control over the microphone would circulate among the users in the room.

By virtue of having the democratic, shared opportunity to participate in running the conversation, you became a better participant in the conversation.

So it was really one of the first pieces of software that was thinking about this intersection between tech and democracy. And that’s really how I got my start.

And now, fast forward, here we are with AI, and the possibilities are limitless.

Scott Greenberg: It’s funny, Beth, that those who are most heavily involved in AI, from every aspect of it, are the ones who are most optimistic about it. And yet, as you say, we can’t, as the public, just sit back and wait for it to happen to us. We need to have an active role.

Going back, I know it was a long time ago when you sat in a coffee house in Vienna, but just last night I finished a World War II book about the agony that was experienced in northern Italy during World War II. And I keep going back to: How did that happen? How did that happen?

Can you tie in that story for us? Maybe men like the Roman Empire and women like World War II — I don’t know what it is, but we all have our historical fascinations. Then bring it forward into what we want to talk about with science and tech today.

Beth Simone Noveck: What happened a lot at the time — and then we’ll bring it forward into what we want to talk about with science and tech today — was you had a set of really institutional actors at that point.

There’s a lot to say about, depending on your view of history, and I’m not getting into the question of evil people like Hitler and the influence of one person. But in addition to the single-leader theory that a lot of people ascribe to, you had a set of institutions, both on the left and the right.

It wasn’t just an issue of which political party. Both sides of the political spectrum had these very, very rigid political structures. It was highly, highly partisan. You were either in my camp or you were my enemy.

A lot sounds similar to today. We live in an age in which more people would agree to marry somebody of a different religion than somebody of a different political party. We live in these very hyperpartisan political times. We have a rising tide of political violence. This is not, again, an issue of left or right. It’s on both sides of the aisle.

And you had something very similar in that period, in terms of this really entrenched camp mentality, where it was like my way or the highway. That kind of fragility, that lack of bipartisanship — as we would call it in our terms today — really contributed to the system breaking down.

Of course, there was also hyperinflation and the economic consequences of World War I, the guilt that we heaped upon them, the blame. There’s lots to say about why that system broke, and it depends on which historians you read.

But I think, going forward today, we’ve set up something similar.

And I would disagree a little bit that there is a great deal of optimism even among those people who are builders of these technologies. In fact, some of the most negative narratives — and we have to ask why — are pushed by those people who created these technologies.

But we have created these two poles, and the media loves these extreme headlines.

Headline one: the world is ending. AI apocalypse. AI is going to come for our jobs. AI is going to take over humanity. It’s going to start doing things for us, and we are going to become slaves to this technology. So you have some version of that.

We have a version of that also with social media and what’s happening to our kids. It’s a very, very negative story about tech.

Or you have this very hyper-positive economic development story, which is about how we in the U.S. have to beat China at the economic race for AI dominance, and it’s going to create jobs, and it’s going to create innovation, etc.

There’s this sort of messy middle that we don’t talk about a lot.

Things happening in Utah, for example, where the state has gone out and built its own tool to help lawyers in the state review contracts, so that a lawyer or contracting officer working for the state of Utah can now very, very rapidly check a contract to make sure it complies with Utah statutes and with the law there. And the state says it saves contract analysts 75% of their time.

That means more contracts getting read, more businesses getting approved for deals, more jobs, and savings of taxpayer dollars.

That’s the kind of almost-in-the-weeds story. Sticking with Utah for a moment, they have rolled out Gemini to employees all across the state. There’s lots happening to use AI, but not what citizens are necessarily seeing, because we’re not replacing humans. When you call to get someone on the phone, you’re still getting a human, thank goodness. But the human answering the phone has a tool that’s helping them get better answers faster.

My students, for example, in a fellowship program we run called AI for Impact, built a tool for the state of Massachusetts to help when you call to get health insurance from your state exchange. The person answering the phone has an AI tool that helps them look across 250 different documents and policies so that you can get an answer faster.

What does that mean? You get an accurate answer. You get a faster answer. You’re not waiting on hold. And it means government works better.

Bringing this back together, the dysfunction we were talking about before with regard to effective governance and flexible, agile governance matters just as much now as it did then.

That’s part of the story that I’m trying to tell: shining some light on these early uses that are really exciting. And if we make the right investments, I think we can have a lot more of that. It will lead to more effective institutions, which in turn may cause rates of trust in government and trust in democracy to increase.

Right now, they couldn’t be lower, especially among young people. We’re talking 7%. Congress, president, doesn’t matter. And again, it doesn’t matter your political party. The rates of trust in our institutions are way too low.

That’s what’s leading to perceived problems and problems with our elections, how our government agencies run, how we talk to one another, how we solve our problems, and the things we think about in this broad bucket called democracy.

Scott Greenberg: It may be sort of simplistic, Beth, to say that AI, if we do it right, will help humans be good at what humans are good at and allow them to shine in these human tasks, and let them use these AI tools to make it more efficient, so that you always get a human when you call.

Hopefully that sort of thing increases trust. But I wanted to make sure we talked about elections, because no matter which side you’re on, you’re always being told that the other side is somehow rigging elections. Elections are really the center of democracy. Can you talk about how AI can help?

Beth Simone Noveck: Sure. And of course, I don’t want to minimize — and it is a good thing — that 49 states now have some version of a law against deepfakes. The idea that you can impersonate someone, whether in the election context or in the sexual context, is something we’ve seen used malevolently.

This is not new with AI. We have had some version of deepfakes — somebody cutting and pasting and photocopying something, or doing that with older technologies. But the robocalls, the impersonating people, the disinformation at scale is obviously something that we worry about. And I don’t want to minimize that at all.

Just in the interest of time, I want to focus on some of the good things for a moment. And that’s what I mean about where we’re focusing our attention.

Because if we’re only talking about a defensive posture, if we only stop with “let’s fight the deepfakes” and we think we’re done, then we’re not doing what we need to do.

Again, it doesn’t matter if you’re a Democrat or Republican or something else. We all should care about the health and well-being of our democracy and believe in the integrity of our democracy: free, fair, and frequent elections, as we typically think of them.

For example, there is software being used — 25 states now use AI-powered software — to conduct what is called entity resolution, to ensure that the people on the voter rolls are not just real people but the right people, and that we are maintaining people’s right to vote.

You move states, and maybe you’re called Scott in one state and, in another state, you’re called — I don’t know, do you have a middle name? Maybe the example works better with John and Jack, which can easily get confused. I don’t know if Scott is short for something. Scottiferous?

Scott Greenberg: My wife says it’s short for a lot of things.

Beth Simone Noveck: You can see how you can get lost between two databases. Whether it’s the right to vote or the right to a school meal for a kid so they don’t go hungry, AI is being used in a lot of places. And we want to make those investments to ensure that we are correctly maintaining those voter rolls.

In Brazil, for example, when you go into the voting booth, they use AI-synthesized voices to read you your ballot so that if you are visually impaired, elderly and can’t see your ballot very well, or blind, you can vote independently and exercise your right to vote.

In India and Kenya, they’re using biometric systems, using AI again, to ensure the identity of voters and ensure integrity in their voting systems.

Then you have lots of examples all over the world of candidates now turning to the simple AI tools that we all have access to, to create campaign posters and multilingual campaign content.

Imagine you live in India, where there are 22 different languages, and you are a less well-resourced candidate. You can now translate all of your materials into 22 languages.

Or you just want to run for office because Lynn has been reading about World War II, she’s worried about the fate of democracy, and she’s now going to run for town council where she lives. How are you going to go up against somebody with a lot more money and power and resources? You can’t afford a fancy consultant. But now maybe with generative AI, you can make a poster, you can make an ad, you can make a commercial. You can compete on a more level playing field.

Those are just some examples.

And then we have lots of work in the disinformation space too. A group like Meedan operates as a nonprofit in 53 different countries. They are using AI, and there’s a lot of research from different universities in this space, to spot and identify misinformation and disinformation so we can flag it sooner, flag it earlier, debunk it, and do the research that we need to do to understand how to fight against disinformation and ensure more effective persuasion.

Lots of great stuff is happening. But I think the greatest threat to elections isn’t that we’re going to have AI that produces a deepfake. We haven’t seen that yet. You’ve had lots of impersonations and fake videos and whatnot, but no evidence that, despite the panic, it’s changed any elections.

It’s that we’re not investing enough and thinking about: How do I use AI to fix the voter rolls? How do I use AI to create simple training for poll workers? How do I use AI to register people who are not registered? How do I use AI to help more people run for office?

That’s what we want to be thinking about: the positive story to complement the negative.

Lynn Ware Peek: If you’re just joining us on Cool Science Radio, we are in a conversation with our guest Beth Simone Noveck. She’s a professor and director of the Burnes Center for Social Change at Northeastern University, and she is one of the world’s leading voices on technology, governance, and civic innovation.

We’re talking about her new book, Reboot: AI and the Race to Save Democracy.

You use, throughout your work, the term democratic AI. A lot of people might hear AI in government and imagine a cold, algorithmic bureaucracy: robots deciding who gets benefits or who goes to jail. How do you distinguish democratic AI from a possible nightmare version? What are the conditions that separate one from the other?

Beth Simone Noveck: Let me state first of all that by democratic here, I’m talking democracy with a small “d.” Not Democratic versus Republican, but “d” for democracy — the thing we all care about, which is elections that are free and fair, government institutions that function, representative institutions that actually speak for us, and our ability to have some amount of participation in the decisions that affect us.

Ultimately, democracy is supposed to be about our own self-governance. How do we move toward a world in which we are more involved, have more say, and have more power over the decisions that impact us?

I deliberately use the term democratic as opposed to, first of all, just AI for good, which some people use, and it is a part of that. But AI for good also includes using AI to detect tumors in healthcare or using AI to fight climate change. There are lots of good things we want to do in the world.

But I want to be very specific here about our ability, first and foremost, to create more effective institutions that can do things like answer the phone faster, deliver a benefit more easily, and, by the way, faster also translates into reduced cost and lower cost for taxpayers.

It’s also about more effective institutions that are using AI and its profound ability to help us do better research and get better information faster, so that we can research the best solutions to problems — whether it’s climate, healthcare, traffic, or whatever issue we’re facing in our community.

We should have access to the best evidence and the best information about what works. And AI is really supercharging our ability, when we use it well, to do that research faster and better. Think about it as having 100 interns, which generally our under-resourced public institutions and civic institutions can’t afford to do.

It’s about better information and how we make decisions. It’s about more opportunities for us to participate and deliberate.

The web has given us 25 years of talking. Everybody can get online and create a YouTube video, as we know. Anyone can now go put their dance moves on TikTok. What it doesn’t help with is listening. In fact, the more people talk, the harder it is to listen.

So I’m really excited about the ways in which institutions are now turning to AI to help them listen better.

In Bowling Green, Kentucky, for example, they went out and asked the whole community: they wanted to make a 25-year economic plan for the future of the community. What investments should we make? What kind of community do we want to live in?

They went out and asked people. But when 8,000 people answer you, it becomes really, really hard to listen. With the help of AI tools, they can summarize what they heard. They can organize the comments. They can put them in buckets and say, “Ah, here’s what people told us.”

The state of California did something similar after the terrible wildfires last year. They launched a project called Engage California, and tens of thousands of people responded with how we could have done better.

Again, when tens of thousands of people respond, it is really, really difficult to listen. That can take months and months and months. Now, because I can use AI to help summarize what people are saying, not only have they been able to turn those ideas into new policies, but they’re continuing to go out and ask. They just launched a new worker engagement. They just launched another citizen engagement.

There’s a lot of excitement, particularly in this space of what I would call participatory democracy: the opportunities for you and me to actually get involved.

Last thing I’ll say here: I just testified on Capitol Hill, where I was invited by one of the committees to come and answer the question: How could Congress actually better listen?

Right now, what is the way in which you get to participate in what happens on Capitol Hill? There isn’t one. You vote for your congressperson once every two years, and that’s about it.

Even if you called your member of Congress, take a number. You’re lucky if anybody has time to answer the phone. The person answering the phone is probably a college intern, and you’re sitting in a pile with literally 800,000 other phone calls, etc., that have come in.

The conversation on Capitol Hill was genuinely — and this was both Republicans and Democrats — asking: How do we actually do this? How would we use these tools?

We’re right at the beginning of learning how to do this, but I think there are a lot of great examples. Short story long, that’s what I mean by democratic AI: how to make our institutions work better and create better ways for us to participate in them.

Lynn Ware Peek: This is absolutely fascinating, and again, I wish we had so much more time. I’m going to let you wrap up with something I’m very interested in.

I found out that you founded something called InnovateUS, which is now described as the nation’s largest AI training program for government workers. I would love to know what you think is the single most important thing a civil servant needs to understand about AI and how to use it well, and how to use it for the benefit of their constituencies.

Beth Simone Noveck: I think it’s something that we all need, civil servants or not. And by the way, everything on InnovateUS is free, so anyone can use it. It’s used by the state of Utah and its civil servants, but anyone can go there. We do four workshops a week. We have classes that you can take at your own pace.

The most important thing is that we learn what these tools are good at and what they’re not good at. Especially understanding what the dangers are and what the risks are can help us ensure that we are avoiding them, that we are not buying stupid tools that someone tells us will magically predict the future or magically solve all of our problems, but that we are actually understanding what they can and can’t do, and how we each can use them in our own job and in our own work.

A very controversial topic right now is this issue of data centers, for example, and how much energy, water, and electricity we should consume when using these AI tools. It requires a little bit of education. What’s a data center? How is an AI data center different? What does that mean?

I think it’s just really important that we all learn what these tools are and how we can use them — whether it’s as a government worker, how to serve the public, or whether it’s for each of us as individuals, how we can use them to run for office, to organize our neighbors, to go clean up the park, to get involved, to write to our congressperson or to write to our government, how we can use them for good to help our kids better learn how to read.

There are lots of things we can do with these things. We want to be focused on the positive, proactive things we can do to take power and use the tools so that they serve us.

AI on tap, not AI on top.

Lynn Ware Peek: Beth Simone Noveck’s new book is Reboot: AI and the Race to Save Democracy, published by Yale University Press. You can read more of her work at rebootdemocracy.ai.

Beth, thank you so much for joining us today on Cool Science Radio.

Beth Simone Noveck: Thanks for having me. I appreciate it.