A scenario for American democracy 20 years from now

Let's do a little dreaming, shall we?

Derek Thompson and Ezra Klein’s new book Abundance is about the need to cut back the bureaucratic barriers to building and inventing things in the US. It starts with a rather dreamy description of how the world might look in 2050—a multitude of clean energy sources, abundant desalinated water, locally sourced food from vertical farms, cheap miracle drugs.

It’s very utopian—and it got me thinking. As I said in my last post, I’ve been trying in Futurepolis to highlight present-day innovations in democracy and governance. What I haven’t done much of is sketch a broader vision of what they could add up to.

So let me take you on a little journey.

It’s 2045. America has gone through a dark period of economic turmoil, social unrest, corruption, and institutional breakdown. Yet the second Trump administration’s dismantling of government, while extensive, wasn’t total. Some agencies are never coming back, but there are still courts and a Congress, and the separation of powers has largely reasserted itself. And alongside it all, there’s been a flowering of civic innovation, as people started to realize that good government matters after all.

Below is a scenario—a design fiction, to use the term of art—for this hypothetical future. It touches on what I think are four key areas of reform we need: state capacity, the government’s ability to execute its policies; service delivery, providing public services effectively and in a timely manner; policymaking, creating laws and regulations that keep pace with technological and social change; and civic engagement, staying abreast of citizens’ needs and giving them a say in governance.

In this post, I’ll just lay out the vision. In subsequent ones, I’ll unpack how it relates to those four areas.

Update (March 5): I’ve revised the final section, on lawmaking, to take account of likely improvements in AI in the next few years.

---

It’s 9am on April 17, 2045, and Donnie Larson is feeling pretty wrung out. He got back to Washington from his district late last night after the Easter recess, and this morning he’s already had two meetings: one with some smooth-talking executives from a robotics company, the second with a curmudgeonly Teamsters vice-president. Though VR telepresence is now almost indistinguishable from meeting in person, his visitors made the pilgrimage to his cramped, lightless office in an obscure corner of the O’Neill House Office Building. Pressing the flesh is still the way to get things done in politics.

The execs wanted handouts to set up a manufacturing facility—mostly staffed by robots, of course—on the outskirts of Duluth. They made reassuring but vague promises about trickle-down effects on the local economy. The grizzled union boss, meanwhile, seemed to take a sadistic pleasure in reminding the fresh-faced Congressman that in each of the last three censuses, Minnesota’s eighth district barely escaped being scrubbed off the map. If jobs keep getting automated and out-of-work residents keep moving away, Donnie could end up without a seat.

It’s a dilemma. Donnie was born on January 20, 2017; his parents named him after Donald Trump. He thinks his namesake did some good things for the country, but he identifies more as a moderate Republican. In particular, he believes government should take a light touch on regulation but throw money at technology development—both in the private sector, to encourage innovation, and in the public sector, to improve the workings of the state. That’s why he angled for seats on the House subcommittees for both technology policy and government modernization. But now, it seems, his enthusiasm for the tech industry is bumping up against his need to keep his job.

Still, he can’t think about that now. There’s barely time to grab coffee with his aides before spending the rest of the day at a session of the tech policy subcommittee.

It’s these brief team meetings that Donnie looks forward to the most. Though there are only two of them, his policy advisers make a formidable pair. Andrew is a wise and sharp-witted career bureaucrat with 25 years of experience in different agencies; Sam is a ludicrously high-energy investment banker in her early thirties. Both are on two-year secondments paid for by the government to get more expertise into Congress.

Unlike the inexperienced 20-somethings who used to staff Congressional offices and scraped by on food stamps, these two get a decent wage—by Washington standards, anyway. They also enjoy the backing of the well-resourced Congressional Research Service, which uses advanced expert systems and unfettered access to federal, state and local government data to produce detailed policy briefs tailored to each district, pretty much eliminating the need for members to have their own research assistants.

Over coffee, Sam and Andrew update Donnie on what’s been happening while he was at home over the recess.

First, there’s the White House bill to formally license non-human psychotherapists. It’s been a bitterly contested issue. The medical profession, of course, opposes putting AI therapists on a par with human ones. Patient advocacy groups are mostly in favor. Health insurers are split, and so too are both Republicans and Democrats, though one of the smaller parties that has emerged since the reintroduction of fusion voting made the issue part of its platform at the last election. Donnie himself is pro-licensing, mainly because he lost an uncle who couldn’t afford mental-health treatment to suicide.

Now, Andrew tells him, the public comment period on the bill is over. There have been more than 23 million comments in the form of text and video. Most are short messages pro- or anti-licensing, but many are long, heartfelt accounts of people’s struggles with mental health while battling the healthcare system. In the past, it would have been impossible to take in even a tiny fraction of them. But the Congressional Research Service’s analysis engines have ingested them all, weeded out the ones created by bots and obvious lobbyists, boiled the rest down to half a dozen main viewpoints, and built avatars to represent each one. The avatars, which any member of the public can talk to in VR, AR or on a screen, can hold lengthy conversations—answering questions, summarizing the arguments for their viewpoint, and drawing on individual citizens’ stories.

First trialed nearly a decade ago, the avatars are one of the things that persuaded Congress to build the CRS back up after it was gutted under Trump. Lawmakers love talking to the avatars: it’s like talking to millions of voters at once, except they never get angry. And voters apparently feel heard: 23 million comments isn’t an unusual number these days. Sam and Andrew recommend Donnie spend some time with the avatars. They think it could give him some arguments to bolster his pro-licensing position.

Next on the agenda: the final report of the Minnesota citizens’ assembly on sex education in schools has come in. This is something Donnie instigated two years ago, when he was a state representative. Under a state law passed in 2037, any Minnesota legislator can propose putting a question to a small, cross-partisan group of citizens chosen at random. If the proposal wins a simple majority vote, the state government must pay to hold the citizens’ assembly, and legislators must then hold formal hearings on whatever it recommends.

The assemblies have proven popular as a way to break political deadlock on hot-button issues; since the law was passed, there’s been one almost every year. They typically spend several weekends gathering information, listening to experts and lobbyists, and deliberating. Now the assembly Donnie proposed has come up with a draft curriculum for giving kids age-appropriate sex education that seems to satisfy both liberals and conservatives—a consensus the adversarial power dynamics of politics had failed to produce. Andrew, who worked as a Congressional liaison in the Education Department back before it was abolished, thinks the curriculum could fly in other states too. Donnie would take the credit. That’ll be good for his profile within the party.

Next up: Minnesota has finally launched its new benefits system. People will now get unemployment insurance, disability, and other benefits automatically without needing to apply. Building the system took nearly eight years—the state still has a long way to go in fixing its creaky technology procurement process—but nonetheless, this makes it only the second state in the country to offer automatic benefits.

Donnie originally opposed this idea. He didn’t think benefits should be easy to get. But he’s been swayed by data from European countries where this has been the norm for more than a decade. No application process means no way to commit fraud and no need for claims assessors, so the system saves money, and nobody falls through the cracks. Though he had nothing to do with the initiative, it’s an opportunity for him to get on some of the influencer channels and get some publicity for his state—and himself.

Finally, the RATE (Robot Autonomy for Technological Excellence) Act is up for its triennial effectiveness review. Twenty years ago, such reviews would have been an impossible burden for Congress, which could barely deal with writing legislation, let alone revising it. But under the second Trump administration, two factors led to a drastic shift.

The first was a one-two punch that emasculated federal agencies. Seven months before Trump came into office, the Supreme Court had overturned the so-called Chevron deference doctrine, which had given those agencies a lot of leeway in how to interpret the laws passed by Congress. The 2024 court ruling curbed that leeway. Combined with the Trump administration’s willy-nilly slashing of the federal bureaucracy, agencies were left practically incapable of turning laws into regulation.

That meant it was now on Congress to draft clearer and more explicit laws. For an already dysfunctional legislature this was a near impossible task. But this began to change with the advent of the second factor. Within a few short years, AI systems had grown powerful and accurate enough to turn even a few bullet points outlining a law’s intent into fully-fledged legislative language. Not only that, they could scour the entire body of existing laws, regulations, and executive orders to look for potential conflicts, and suggest amendments or repeals as necessary.

That made the drafting of laws much faster, more precise, and less vulnerable to unintended consequences. Political horse-trading was still the bulk of how legislators spent their time, but now it was based on a much clearer understanding of a new law’s likely effects.

Of course, that understanding still wasn’t perfect. Laws didn’t always do what they were meant to. Which was where the triennial reviews came in.

The RATE Act was supposed to make the US robot industry more competitive with China’s by relaxing restrictions on Level 5 autonomy. But three years later there are still practically no Level 5 robots on the market. Reviewing the act now falls to the tech policy subcommittee Donnie sits on.

Sam has been digging around in industry data and talking to some contacts via her investment banking colleagues, and she reckons she knows why the act has failed. A little-noticed provision in the tax code inadvertently gives robot manufacturers an incentive to delay rollouts of new models. Donnie, she notes, could make an early name for himself on the subcommittee by pointing out the problem and proposing a change.

That gives him an idea. He could go back to those execs who wanted to build a robot factory in Duluth, and offer a quid pro quo. If the company will make some firmer guarantees to ensure its factory benefits the local economy, and helps him lobby for the change in the tax code, Donnie will get them some juicy subsidies. Hopefully that’ll create enough jobs in his district to counter the losses from automation, as well as scoring him kudos on the subcommittee.

He knocks back the rest of the coffee and stands up with a grin. This is going to be a helluva day.

---

Related: Jen Pahlka reviews Marc Dunkelman’s book Why Nothing Works.