An AI tool and deliberating ballot measures
Californians vote on many ballot measures, but we rarely participate in significant public discussions about the measures’ contents and impacts.
This isn’t simply a result of apathy or poor civic education. Rather, it’s an example of “rational ignorance,” a term coined by economist Anthony Downs in his 1957 book, “An Economic Theory of Democracy.” The term defines this democratic reality: Since you have just one vote out of millions, your vote doesn’t much matter. So, it’s rational to not devote precious time to reaching well-considered decisions about how you vote.
And when too many of us remain rationally ignorant, our election results don’t match the public interest.
This year, however, Stanford political scientists are seeking to counter our rational ignorance with an advanced tool: a digital Deliberative Poll.
The poll would give hundreds of Californians the opportunity to seriously deliberate over certain ballot measures. One set of likely candidates for deliberation are three competing constitutional amendments involving the voting requirements for taxes.
Here’s how digital Deliberative Polls work. Stanford’s Deliberative Democracy Lab assembles a representative sample of the California electorate. Participants are paid for their time (and reimbursed for child or elder care obligations). Their internet speeds are increased if necessary.
First, the platform, developed in collaboration with Stanford’s Crowdsourced Democracy Team, polls participants on the measures to establish a baseline. Then, some members of the group deliberate. (The rest are in a control group that doesn’t participate in the deliberation).
Next, the platform, which is AI-assisted with no human operator randomly divides the sample into small groups of 10 that engage in video-based dialogue over the pros and cons of the ballot measures and decide on key questions to ask panels of experts representing different points of view. The participants then ask their questions in plenary sessions by video with all the groups present. The small group deliberations and plenary sessions alternate throughout the process.
The AI tries to facilitate an equal discussion. The tool nudges reluctant participants to speak up. The platform can intervene if people become uncivil.
At the deliberation’s end, participants (and the control group) are polled again on the measures. The before-and-after difference between the survey results are shared with the public, to demonstrate the impact of deliberation on participants’ views.
“It is a social science experiment and a form of public education,” said Stanford’s James Fishkin, who leads the polls and the Deliberative Democracy Lab. “It overcomes ‘rational ignorance’ because each person, instead of one voice in millions, has one voice in a small group of 10 engaging in meaningful dialogue.”
Fishkin originated the concept of the Deliberative Poll as an in-person event, in 1988, and has deployed it on issues ranging from Korean reunification to civil service reform in Brazil.
Recently, Fishkin and his team conducted a series of Deliberative Polls known as “America in One Room” that got Americans of different views to deliberate with one another on issues from energy to immigration. Those Deliberative Polls showed that such conversations can still produce common ground, even amid nationwide polarization.
Fishkin and Siu say California’s ballot proposition system badly needs deliberation. Some measures get little public notice. Campaign donations from special interests determine which measures get on the ballot and get our attention. Measures have grown increasingly long and complex, defying voters’ attempts to understand them.
The momentum to use Deliberative Polling in California dates to the attempted 2021 recall of Gov. Gavin Newsom, which increased public concerns about flaws in California’s direct democracy.
In response, Secretary of State Shirley Weber asked the bipartisan team of former Gov. Jerry Brown and former Chief Justice Ronald George to recommend reforms of initiative, referendum and recall. With assistance from Nathan Gardels of the nonpartisan Think Long Committee for California and the Berggruen Institute, their report lamented the absence of a public, institutional platform for informed deliberation on ballot measures. A 2022 Public Policy Institute of California survey found 77% support among likely voters for an independent citizens commission to study ballot initiatives.
Today, Think Long and the Stanford team are circulating a proposal to funders to apply Deliberative Polling to the measures this November. (Full disclosure: I’m a fellow in Berggruen’s Renovating Democracy program this year.)
Fishkin and Siu say Deliberative Polling can be especially effective when multiple measures address similar issues — as with the competing taxation amendments. They recently held a well-received Deliberative Poll around four different proposals to the Finnish parliament (which showed that the automated platform produced similar results to deliberations with human moderators.
Fishkin said that campaigns and stakeholder groups on opposing sides of measures participate in the deliberations because they want to have their best case heard. Participants in processes like Deliberative Polls also become more engaged and better informed citizens.
The rest of us can become less ignorant from seeing the results of their deliberations.