A Kentucky Town Experimented With AI. The Results Were Stunning

Gizmodo

4 days ago

A county in Kentucky conducted a month-long 'town hall' with nearly 8,000 residents in attendance earlier this year, thanks to artificial intelligence technology.
Bowling Green, Kentucky's third largest city and a part of Warren County, is facing a huge population spike by 2050. To scale the city in preparation for this, county officials wanted to incorporate the community's input.
Community outreach is tough business: town halls, while employed widely, don't tend to gather a huge crowd, and when people do come, it's a self-selecting pool of people with strong negative opinions only and not representative of the town at large.
On the other hand, gathering the opinion of a larger portion of the city via online surveys would result in a dataset so massive that officials and volunteers would have a hard time combing through and making sense out of it.
Instead, county officials in Bowling Green had AI do that part. And participation was massive: in a roughly month-long online survey, about 10% of Bowling Green residents voiced their opinions on the policy changes they wanted to see in their city. The results were then synthesized by an AI tool and made into a policy report, which is still visible for the public to see on the website.
'If I have a town hall meeting on these topics, 23 people show up,' Warren County judge executive Doug Gorman told PBS News Hour in an interview published this week. 'And what we just conducted was the largest town hall in America.'
The county got the help of a local strategy firm to launch a website in February where residents could submit anonymous ideas. For the survey they used Pol.is, an open-source online polling platform used around the world for civic engagement, and to particularly great success in Taiwan.
The prompt was open-ended, just asking participants what they wanted to see in their community over the next 25 years. They could then continue to participate further by voting on other answers.
Over the course of the 33 days that the website was accepting answers, nearly 8,000 residents weighed in more than a million times, and shared roughly 4,000 unique ideas calling for new museums, the expansion of pedestrian infrastructure, green spaces and more.
The answers were then compiled into a report using Sensemaker, an AI tool by Google's tech incubator Jigsaw that analyzes large sets of online conversations, categorizes what's said into overarching topics, and analyzes agreement and disagreement to create actionable insights.
At the end, Sensemaker found 2,370 ideas that at least 80% of the respondents could agree on. Some of the most agreed upon ideas included increasing the amount of healthcare specialists in the city so that residents don't have to rely on services an hour away in Nashville, repurposing empty retail spaces and adding more restaurants to the north side of the city.
The online survey was able to reach people that the county could not have reached otherwise like the politically disengaged or those who could not find the time from work to attend town halls.
The format was also better at reaching immigrants by offering the survey in multiple languages and then automatically translating answers. That was welcomed by people like Daniel Tarnagda, an immigrant from Burkina Faso and a local non-profit founder who leads a soccer team of under-18 immigrants who struggle to speak English.
'I knew that people want to be part of something. But if you don't ask, you don't know,' Tarnagda told PBS.
The volunteers for the project are now compiling the ideas from the report to make concrete policy recommendations to county leadership by the end of the year. According to a survey that Jigsaw conducted with local leaders, AI saved the county an average of 28 work days.
The Bowling Green experiment was Sensemaker's first large-scale proof of concept, Jigsaw wrote in a blog article from earlier this year.
One of the most striking things they found out in Bowling Green was that when the ideas were anonymous and stripped of political identity, the constituents found that they agreed on a lot.
'When most of us don't participate, then the people who do are usually the ones that have the strongest opinions, maybe the least well-informed, angriest, and then you start to have a caricatured idea of what the other side thinks and believes. So one of the most consequential things we could do with AI is to figure out how to help us stay in the conversation together,' Jigsaw CEO Yasmin Green told PBS.
Jigsaw announced this week that they are now partnering with the Napolitan Institute, a research and public polling organization founded by famous pollster Scott Rasmussen, to compile information on how Americans from every congressional district view the founding ideals of America, the state of the country now, and where it's going. Unlike the Bowling Green experiment, the aim is not policy but to get an understanding of where the nation stands.
There are still concerns inherently tied to this experimentation with AI in local governance. Although the website for Bowling Green's survey explicitly notes that 'no personal information was captured, and no demographic data was stored,' that does not necessarily mean any future applications of this elsewhere would follow suit.
Artificial intelligence is the subject of privacy concerns due to their vulnerability to data breaches, which would become a problem if people were to get doxxed for their political beliefs they submitted in confidence.
AI also has an issue with the bias of its creators getting baked into its algorithm. Just last month, researchers found that Elon Musk's Grok chatbot would consult with Musk's own –rather controversial– opinions before answering sensitive questions. If an AI were to generate neutral policy suggestions, a flaw as such would be insurmountable.
But if these concerns are adequately addressed, then AI could have the potential to completely revolutionize civic engagement. And it could show a pathway for moving past political polarization and towards enacting tangible change, similar to the way AI created a space for a divided community in Bowling Green to find its common ground.