Peace Magazine , page . Some rights reserved.
Search for other articles by kgsimons here
Nowadays, political discourse is a shouting match between deaf participants. The idea that we could simply talk our way out of polarization sounds downright quaint. Yet, at Stanford University, amidst the palm trees and the cerebral hum of Silicon Valley, Professor James Fishkin is proving that conversation—structured, informed, and representative—is not just a nice idea. Indeed, carried out properly, such projects may be the only thing that can save democratic decision-making.
This possibility gained plausibility during a Project Save the World forum between Fishkin and Metta Spencer, a sociologist and host, who had sought out Fishkin’s expertise despite her loyalty as an alumna of nearby UC Berkeley (“Even in my day, there was rivalry with Stanford… and Director of the Deliberative Democracy Lab, argues that these snapshots are often illusions. They capture what people think when they haven’t really thought at all—when they are operating on soundbites, misinformation, or sheer inattention. “What we do is very simple and common sensical, but it’s led me on an extensive journey,” Fishkin explained. His method is called Deliberative Polling, and it stands in contrast to the standard town hall or even the increasingly popular citizens’ assemblies seen around the world. Spencer noted that she had wanted to run a citizens’ assembly to discuss AI safety but couldn’t recruit participants by “sortition”— the random selection of citizens by lottery, as juries are chosen. Fishkin was quick to point out that even famous assemblies, like those on mostly by people interested in sports,” she reminisced to her guest).
Project Save the World had already embarked on an experiment related to Fishkin’s work: a small inquiry group discussing the existential risks of Artificial Intelligence. But the participants had realized that to probe the “will of the people,” the dialogue must involve more than just a group of interested volunteers. One needs rigorous sampling and a process of what Fishkin calls “Deliberative Democracy.”
Most of us assume we know what “public opinion” looks like. We see it in the jagged lines of approval ratings and the snapshot percentages of cable news polls. But Fishkin, the Janet M. Peck Chair in International Communication at Stanford and Director of the Deliberative Democracy Lab, argues that these snapshots are often illusions. They capture what people think when they haven’t really thought at all—when they are operating on soundbites, misinformation, or sheer inattention.
“What we do is very simple and common sensical, but it’s led me on an extensive journey,” Fishkin explained. His method is called Deliberative Polling, and it stands in contrast to the standard town hall or even the increasingly popular citizens’ assemblies seen around the world.
Spencer noted that she had wanted to run a citizens’ assembly to discuss AI safety but couldn’t recruit participants by “sortition”— the random selection of citizens by lottery, as juries are chosen. Fishkin was quick to point out that even famous assemblies, like those on climate change in France or Ireland, also sometimes suffer from a fatal flaw: selection bias.
“In the French case, they sent out 350,000 text messages, and they ended up with 150 people who volunteered to spend six months thinking about climate,” Fishkin noted. “I can’t believe that those people were not more interested in climate than typical of the general population.”
If you only gather people who already care about an issue, you aren’t measuring the public will; you are measuring the will of activists. Fishkin’s approach is different. He uses the “gold standard” of survey research (partners like NORC at the University of Chicago) to recruit a truly random, representative sample of the population. These recruits aren’t volunteers; they are the folks who would otherwise be grocery shopping or at home watching football. They reflect the real mixture of public opinion.
So, what happens when you take a random slice of humanity— Republicans, Democrats, the disengaged, the passionate, the semi-literate, and the scholarly—and put them in a room?
The process Fishkin describes is a mix of social science experiment and civic boot camp. He begins with a baseline questionnaire to assess attitudes before any discussion takes place. Crucially, there is also a control group that takes the survey but does not deliberate, ensuring that any changes in opinion can be reasonably attributed to the deliberation process itself.
The participants are then provided with a “briefing document.” This isn’t propaganda. Fishkin emphasizes that these documents are vetted by extensive advisory committees representing sharply contrasting viewpoints. “We’re not asking them to agree with the briefing document,” he told Spencer. “We’re asking them to agree whether the briefing document is balanced and accurate.”
Once the participants have the facts, the magic happens. Whether face-to-face or, increasingly, via a sophisticated online platform, the participants break into small groups of ten. They don’t just shout over one another. The discussions are moderated—sometimes by humans, sometimes by AI-assisted technology—to ensure civility and equal speaking time.
“It controls the queue for discussion. Each person can only speak 45 seconds at a time….It actually creates very equal participation,” Fishkin explained regarding their online moderator.
These small groups identify the key questions they want answered. They then convene in a plenary session where panels of competing experts answer those specific questions. Then, they go back to their small groups to weigh the trade-offs. At the end of the weekend, they take the questionnaire again.
The results of Fishkin’s 160 projects conducted around the world are nothing short of startling. When people are given a safe space to discuss complex issues with “diverse others,” and when they are armed with accurate information, their views change. More importantly, the toxic fog of polarization lifts. Fishkin calls this moderated discussion with diverse strangers a “magic elixir that brings public opinion to life.”
He highlighted data from “America in One Room,” a major deliberative polling experiment. The media narrative tells us that Republicans and Democrats are living in different realities, intractably divided. Fishkin’s data suggests that this division is often a product of ignorance and isolation. “The most polarized opinions are the ones that depolarize,” Fishkin revealed.
He offered two striking examples from their work:
1. Immigration: Initially, 80% of Republicans in the sample wanted to deport all undocumented immigrants to their home countries. After a weekend of deliberation, hearing arguments, and meeting actual human beings affected by policy, that number dropped to 40%. A 40-point shift!
2. Social Programs: Democrats were not immune to reality checks either. On proposals for the most expensive social programs, such as certain forms of Universal Basic Income, support among Democrats dropped by nearly 40 points as they grappled with the economic tradeoffs.
“On both sides, the mean positions of the two parties moved substantially closer together,” Fishkin said. “This idea that we are inextricably divided is an illusion.”
The Host, whose then-ongoing AI inquiry group was a smaller, scrappier attempt at this kind of dialogue, was visibly impressed by the scale of Fishkin’s operation. She asked about the concept of a “Day of Deliberation”—a proposal Fishkin and legal scholar Bruce Ackerman had suggested years ago for a national holiday dedicated to civic discussion.“Fishkin’s data suggests that this division is often a product of ignorance and isolation. “The most polarized opinions are the ones that depolarize,”
While a physical national holiday remains expensive and logistically difficult, technology is bridging the gap. Fishkin is now working with major European governments and various U.S. states to deploy his organization’s online platform. This allows for “mass deliberation” without the prohibitive costs of flying hundreds of people to a single hotel.
“We think that it’s eminently practical to do this on our platform,” Fishkin stated. The platform uses AI not to replace human judgment, but to facilitate it—providing realtime transcripts, nudging quiet participants to speak, and using Large Language Models (LLMs) to digest the arguments being made.
The implications of Fishkin’s work extend far beyond academic curiosity. In a world where “fake news” and algorithmic silos drive us apart, Deliberative Polling offers a structural fix. It serves four distinct functions:
1. Truth-telling: It reveals what the public would think if they actually had the time, occasion, and resources to think.
2. Policy Impact: It gives governments the political cover to make difficult, nuanced decisions that the public has actually vetted.
3. Civic Education: It informs the participants deeply.
4. Depolarization: It proves that we can still talk to each other.
As the conversation wound down, the contrast between the current political climate and Fishkin’s laboratory results felt stark. Outside, the world is shouting. Inside the Deliberative Democracy Lab, people are weighing trade-offs. They are listening.
Project Save the World’s Inquiry on AI risks, while smaller, shares the same aspiration. It is an attempt to introduce friction into the slippery slope of hot takes—to slow down, read the briefing, and ask the hard questions.
“We have no priors as to what direction they’re supposed to change their opinions toward,” Fishkin reminded Spencer. The goal isn’t to manufacture consent, but to manufacture thought.
In the end, the rivalry between Berkeley and Stanford that Spencer had mentioned jokingly seems trivial compared to the rivalry between democracy and dysfunction. And in that contest, James Fishkin has provided a blueprint for how democracy might just win.