In the much-anticipated New York Democratic primary in June, centrist George Latimer defeated incumbent Jamaal Bowman by a landslide margin of 58.7% to 41.3%.
Ahead of the vote, two 19-year-old college dropouts from Manhattan conducted a poll that predicted the outcome by within 371 votes. Their secret? Instead of surveying one person, they asked thousands of AI chatbots which candidate they supported.
Welcome to the future of polling, according to Cam Fink and Ned Ko, co-founders of a seven-person company called Aaru. They say they’ve figured out the secret to accurately predicting election outcomes, which have come under increasing criticism since most polls failed to predict Donald Trump’s victory in 2016. The answer: ignore the humans whose behavior they’re trying to capture.
For election outcomes, Aaru uses census data to replicate voting districts, creating AI agents that are essentially programmed to think like their replica voters. Each agent is given hundreds of personality traits, from aspirations to family relationships. The agents frequently surf the internet, gathering information to mimic the media intake of their human counterparts, which can cause the agents to change their voting preferences.
For example, when Donald Trump was shot in an assassination attempt, many Aalu agents quickly changed their votes to support the former president, but after more information about the gunman emerged in the hours following the attack, many of the agents changed their votes back.
A poll typically collects responses from around 5,000 AI respondents and takes anywhere from 30 seconds to a minute and a half to conduct. Aaru charges less than one-tenth the price of human surveys.
“By the time of the next general election, traditional polls won’t exist,” Fink said in an interview with Semaphore. “Using real people creates a big problem: you never know if someone is telling the truth.”
He said the company has been hired to conduct polls for Fortune 500 companies, political campaigns, think tanks and political action committees, including one California campaign that relies primarily on Aaru for its polling.
During one investigation, the company noticed that one of its AI agents said he would vote for Mickey Mouse in the next presidential election. Fearing that one of the bots had gone out of control, the Aaru team investigated. They found that the bot was accountable. “The agent’s response was, ‘I don’t like Kamala or Trump. I’m going to fill out my ballot and vote for Mickey Mouse,'” Fink said.
