QuickTake:

A software engineer decided to see if he could create AI simulations of a process in which a jury of citizens assess ballot measures. The results are in your voters‘ guide.

For decades, Clay Shentrup has been trying to find ways to make elections better.

The Portland-based software engineer has co-founded a nonprofit organization called the Center for Election Science and has helped push through election reforms in Fargo, North Dakota, and St. Louis, Missouri.

But he kept running up against a fundamental issue:

“People just aren’t informed about both the issues and how the candidates match those issues,” he said. “What can you do to solve this?”

So he was intrigued to learn about an experiment, the so-called Citizens’ Initiative Reviews.

Here’s the basic idea: Pull together a jury of two dozen or so randomly selected residents, chosen in part to match Oregon demographics. Bring them together for a week or so and have them learn from experts about the pros and cons of a ballot measure. Let jurors hash it out and come to a recommendation on the measure. Publish the jury’s final report in state or local voters’ pamphlets.

A handful of these reviews have taken place in Oregon (the state was the first to use the process, in 2010), but there’s a catch: It can be expensive to pull together two dozen people for a week or so and funding has not been consistent.

This is the point in the story where it’s important to remember that Shentrup is a software engineer. Because here’s where he took this: He started thinking about whether you could use artificial intelligence to simulate a Citizens’ Initiative Review. He asked himself: “Could we devise a hypothetical scenario where people with different backgrounds evaluate these issues and kind of simulate what a real jury would do?”

“It’s really more like a demonstration of the idea, since I don’t have the money to fund a real jury,” he said. 

So, even if Shentrup couldn’t pay for a real jury, he could fire up the AI tool Claude, ask it to assemble a demographically accurate “jury,” feed it as much information as he could find about a particular issue, add background information about how the Citizens’ Initiative Review works and see what emerged.

He picked a state issue, a county issue and a city issue for this initial experiment — and then submitted the final decisions generated by AI to voters’ pamphlets.

You can see all three results in the Lane County Voters’ Pamphlet:

  • Shentrup’s 23-member AI jury voted 16-7 to recommend that voters pass the city of Eugene ballot measure that would allow city department heads to live outside the city limits. The argument summary published in the Voters’ Pamphlet lists points where the “jurors” agreed and also laid out arguments for and against the measure.
  • The jurors voted 13-10 against the Lane County Watersheds Bill of Rights, “not because they opposed clean water, but because they concluded this law cannot achieve it,” according to Shentrup’s summary in the Voters’ Pamphlet. Again, the summary lists arguments both for and against the measure.
  • In the closest of the three simulated Citizens’ Initiative Reviews that Shentrup conducted, jurors voted 12-11 in favor of repealing proposed increases in gas taxes and other fees to pay for transportation projects. Even though polling suggests voters will resoundingly reject the increases, the jury’s narrow call “shows this is a harder choice than either side admits,” Shentrup wrote.
The argument created by an AI-simulated citizen jury regarding the Lane County Watersheds Bill of Rights, as submitted by Clay Shentrup, is seen at the top of the first column on the righthand page of the spread shown here. Credit: Lookout Eugene-Springfield

‘Conduit of information’

Capturing that sort of nuance — and allowing people to be open to new information and to change their minds in light of that new information — is exactly what a Citizens’ Initiative Review strives to accomplish, Shentrup said.

“It’s actually kind of a conduit of information, right?” he said. “It helps you break out of the left-right thinking, where one side is saying, ‘vote against the other side’ and the other side is saying ‘vote for it.’ … So we want to show you that there are multiple sides of this issue and have it be unbiased.”

He said he tried to model that kind of experience with the simulated reviews and added that the finished summaries are meant as another proof that the Citizens’ Initiative Review can work.

“I think that by exposing people to this idea that there are views across a spectrum … helps trigger that empathetic wiring that helps look at other perspectives with an open mind,” he said.

And he believes citizens’ juries can be boons to democracy in areas way beyond voters’ pamphlets. For example, he points to Henry County, Georgia, where the county’s ethics board includes four members appointed by a grand jury. Why not convene citizens’ juries to choose, say, four members for a city council and have voters elect the remaining positions?

To that end, Shentrup has filed a proposed initiative petition that would establish citizen juries to select justices whenever a vacancy occurs on the Oregon Supreme Court. (Now, the governor gets to choose a replacement, who must then be retained by voters at the next election.)

There’s not much chance the idea will show up on a ballot any time soon, because it needs 156,231 signatures on petitions. But anything that gets people talking about ways to improve democracy could be a step forward, Shentrup said.

“We’re just trying to find ways to get legs on some of these more innovative ideas.”

Mike McInally is a Pacific Northwest journalist with four decades of experience in Oregon and Montana, including stints as editor of the Corvallis Gazette-Times and the Albany Democrat-Herald.