Video: Better government in 10 minutes

A new scenario for the citizens of the future.

There’s a saying: “The government is like the toilet; people just want it to work.” Ideas about reinventing democracy often entail giving citizens all sorts of new ways to participate beyond voting. But how many people really want to go to town hall meetings, sit on panels, send letters to politicians, fill out surveys, or vote on ballot propositions they know nothing about? It’s time-consuming and often seems pointless.

So: could we create ways for people to have a much bigger say in governance without having to spend time on it, and also see the results? I think we could.

The video above is of a talk I gave last week at a gathering organized by PACE, a “philanthropic laboratory” for foundations that fund democracy work. Much like this post I wrote back in February, it imagines the US democracy of the future, but this time from the point of view of a citizen.

It’s a morning in the life of “Sandra,” a new mother who gets her news, files her taxes, and performs no fewer than three acts of civic duty, all in the space of 10 minutes and without leaving her breakfast table, by interacting with her AI assistant, “Harry.” (If you can’t be bothered with the video, here are the slides.)

This scenario is actually pretty feasible. It weaves together a bunch of innovations in government service delivery, legislative processes, media, digital public spaces, and civic participation. Here’s the thing: almost all of them already exist to some degree somewhere in the world. And while the AI applications in it are futuristic, they’re not far off being technically doable—almost certainly well before 2040.

For those who want a bit more detail, I’ve listed these innovations below.

Data privacy laws

Harry has access to pretty much all of Sandra’s digital traces, including her email, so it can personalize her news. Amazon’s Alexa+ already has this access, though it can’t yet give you a tailored news roundup or interact the way Harry does. But what’s it doing with your data? Right now, for example, it can send all voice recordings to Amazon for analysis. And it has to send some data to third-party services it connects to, like Uber, Google, or OpenTable, but what happens to that data afterwards? A strong data privacy law would give users some peace of mind.

Personalized news

Google already has a version of Harry’s personalized news: a five minute “Daily Listen” based on your web searches and other preferences. It’s not interactive—you can’t get it to go deeper on a topic, for instance. But from a technical point of view, that can’t be far off.

Feedback on law via news item

OK, so this is key to my scenario: the idea that civic participation can be worked into someone’s daily routine instead of being a thing they have to go out and do. How plausible is it? The AI needs to

1. recognize that this news matters to Sandra; we already know that’s easy.

2. be able to determine that the news pertains to a bill that has an open comment period. This could be a matter of querying a Congressional API for bills mentioned in the news.

3. be able to ask Sandra for feedback and process it. Pretty much within the capability of a current LLM.

4. have a means of sending that feedback, appropriately privacy-protected. It wouldn’t take much—mostly bureaucracy—to set up an interface to Congressional systems to make this possible.

So, some protocols, maybe moderate improvements over the LLMs we have today, but hardly science fiction.

Interactive AI avatars

I wrote about Anno Takahiro, the Japanese political candidate who created an AI avatar that people could ask to explain his policies. It was pretty basic, but with the AI video tools now being rolled out, realistic, human-like characters are getting easy to create. Train them on the data from thousands of public comments and they could quite easily hold a convincing conversation that sums up what’s in those comments. They might not be as reliable at accurately reproducing individual anecdotes, but perhaps that can be improved.

Automated tax filing

There were audible sounds of discomfort in the room when I proposed that the IRS might have access to all your financial records, including your bank account, and just calculate your taxes for you. But Estonia has been doing exactly this for more than two decades. And dozens of other countries have at least a partial version. There’s a deep cultural thing here around Americans’ (mis)trust in government that merits its own post.

Automated benefits

See above.

AI summary of community message boards

There are lots of different community message platforms—Facebook groups, Nextdoor, Everyblock, and so on. In today’s atomized society, they could theoretically be great tools for community cohesion, bridging across political divides, and so on. In practice, they’re a hot mess that contributes to the atomization—vast amounts of content all silo’ed off in different platforms and subgroups. An AI assistant that could just scour them all and quickly tell you the mood, especially on issues you care about, could make you feel a lot more connected to your neighbors.

Liquid democracy

This is a nice idea in theory: When you have a vote on issues you don’t understand, delegate it to someone you trust who does. In practice it’s been tried in very few cases, none of them real-world politics (most of them, seemingly, crypto-related), and has problems, most notably the potential for people to game the system. So it would take work to make it robust in the setting I described.

Citizen panel/jury/assembly

This is by far the most developed innovation in my scenario. Hundreds, if not thousands, of these things have been done in various forms and levels of complexity all around the world. Here’s my write-up of a weekend-long convening about gun control in Tennessee.

Outcomes-based legislation and legislative review

Both the Biden and Trump administrations have made a big show of negotiating down prices for certain prescription drugs. But what if instead of making prices the target, you made health outcomes the target? In other words, what if success didn’t mean that more people got access to drugs but more people got better? Radical, right?

Some versions of outcomes-based legislation already exist in the US. The Family First Prevention Services Act is an effort to change the way the child welfare system deals with unstable families. Instead of looking for reasons to put kids in foster care, it prioritizes keeping families together and directs funding towards evidence-based preventive services.

Now pair outcomes-based legislation with the idea that the law has to be reviewed every so often to see if it’s working, and revised if it isn’t. Really radical. I don’t actually know if anywhere does this. In a place like the US it would require a huge overhaul of current processes, and investment in legislative capacity to assess the impacts of laws. But we can’t keep on spending years making and passing laws and then letting them sit unchanged for decades. The world is just moving too quickly.