When Mark Twain was visiting London, a newspaper reported that he had died. Twain famously replied, “The report of my death was an exaggeration.” The same might be said of the public opinion polling industry.
Polling has taken some hits in the last several years—some justified, most not. In fact, media polling was actually pretty darned accurate in the last few Presidential elections. For example, in the 2016 election, the average of public polls predicted Hillary Clinton would get 46.8% of the popular vote—she actually got 48.2%. That’s just 1.4% off. In 2024, media polls missed the final vote by only 2.94% nationwide—and by just 2.2% in the seven swing states.
And as social bots and AI increasingly crowd out the information space, survey research might actually be the most reliable source of where the public stands on critical public policy issues and elections. Surveys actually rely on scientific methodology and are (in most cases) fully transparent. Not so much for AI and bots.
Autonomous bots can flood social media with large numbers of posts expressing support for certain political points of view—creating a false impression of grassroots support. Similarly, AI-generated “deep fakes” can influence public opinion and potentially affect the outcome of elections. You never really know who is behind the bots and deep fakes. “Joe from Nebraska” could actually be Jiehong from Shanghai or Yuri from St. Petersburg.
While the impact of AI on the 2024 election was generally regarded as less than apocalyptic, the danger still lurks. Some public policy responses have been helpful. Specifically, the European Union passed comprehensive legislation mandating that AI-generated content on social media has to be clearly labeled as such. The US could use similar rules.
Until then, for all the criticism it’s weathered, good-old survey research might just be the best way to gauge public sentiment. Thank you, George Gallup.