The other day Axios ran a piece that cited “findings” that a majority of people trusted their doctors and nurses. Turns out, those “findings” were completely fabricated by a company called Aaru – using AI (causing Axios to issue an editor’s note and ‘clarification’)Aaru uses something they call “silicon sampling,” where large language models (the AI) can emulate humans at a fraction of the cost and time required for traditional polling, the NY Times reports.
Silicon sampling isn’t polling. It is the outright fabrication of public opinion by machines – and major news outlets and research firms are now publishing those fabrications as legitimate findings.
This is not an isolated slip. The technology is being embraced by some of the biggest names in media, polling, and corporate research. Gallup has partnered with the startup Simile to create thousands of AI-generated “digital twins” that stand in for real people. Ipsos is working with Stanford to pioneer synthetic data for public opinion studies. CVS, whose venture arm invested in Simile, is already using these fabricated insights to shape customer strategy. And outlets like Axios are treating the output as news.
The entire point of polling has always been authenticity – capturing what actual humans actually think (after oversampling your preferred party to make it look like as if people like Hillary Clinton).
That process is imperfect and messy. Let’s say a pollster wants to learn how many people in the United States are in favor of a certain policy measure, but the pollster ends up with a survey that includes 80 percent Republicans and only 20 percent Democrats. The pollster may think that in reality the country is closer to a 50-50 split, so the results are rebalanced to reflect that perceived reality. This means that the percentages you read as the results of polling are the output of the model, not numbers from the actual survey data.
The problem is that every model is designed with its own biases, because pollsters disagree about which variables deserve more weight. In 2016, The New York Times’s chief political analyst, Nate Cohn, ran an experiment in which he gave five pollsters the same election poll data. (That included Siena College, which conducts opinion polls for The Times and first acquired the data.)
Mr. Cohn found a 5 percent range of difference among what the five pollsters’ models returned. That range was larger than the margin of error typically associated with random sampling, meaning that the modeling assumptions were meaningfully skewing the results. This is alarming, because it suggests that pollsters can use modeling to nudge polls in a certain direction and influence public opinion itself, rather than merely to report what the public thinks.
Walter Lippmann warned a century ago that democracy depends on an accurate picture of the public will. Traditional polling, however imperfect, at least began with real responses from real citizens. It was expensive, slow, and messy precisely because humans are expensive, slow, and messy. Silicon sampling removes every trace of that mess – and with it, every trace of reality. The models are trained on past data, tuned by the biases of their creators, and prompted to spit out whatever “representative” opinions the client wants to see. The result is not public opinion. It is a mirror of the assumptions fed into the machine.
Fake Polling Also Picked Kamala Harris…
On the eve of the 2024 election, Aaru ran a full-scale simulation that confidently projected a narrow victory for Kamala Harris. Market researchers now use these synthetic polls to decide product launches and ad campaigns. Policy shops quietly substitute AI-generated “constituent sentiment” for actual feedback. Each time a respected outlet or pollster presents these inventions as fact, they normalize the idea that fabricated data is good enough.
The consequences are already here. When headlines say “a new poll shows,” readers have no way of knowing whether real people were ever asked. Trust in institutions is eroding fast enough without handing decision-makers and journalists an unlimited supply of plausible-sounding fake data. Social science, political strategy, and market research risk becoming elaborate games of digital pretend.
So there’s that…