Americans are worried about the effect of artificial intelligence on elections, polls show, but the public probably doesn’t understand its full influence on what they experience every day, says an academic who studies the technology.

There have been obvious examples of AI-generated misinformation such as fake audio of President Joe BidenA fake video about voting irregularitiesor memes intended to generate excitement or spread propaganda. AI is also common on a regular basis generate legitimate campaign messagessuch as phone calls and text messages.

But behind these public examples are the “unseen” jobs AI is doing in elections, said Cody Buntain, an assistant professor at the University of Maryland’s College of Information, most prominent in determining the nature of your social media streams. .

“The systems that determine what piece of content is put in front of you, that’s AI at work,” Buntain said. “From the TikTok For You page, X’s feed or profile page to the Facebook feed. Everything is driven by AI.”

Buntain currently teaches a course on how AI is transforming politics, and he said one of the biggest places AI has had an impact is in things we don’t generally see, like the “information diet.”

In a Pew Research Center survey of nearly 10,000 Americans across the political spectrum, released in Septembera sense of unease about the role of artificial intelligence in the presidential election was shared almost equally by both Democrats and Republicans. The poll found that 41 percent of Republicans and 39 percent of Democrats believe artificial intelligence is being used “mostly for harm” during the campaign. Similarly, 56 percent of Republicans and 58 percent of Democrats feel “very concerned” about AI influencing elections.

A separate Pew studyalso published in September, found that many Americans cite social media as their primary source of news.

While general sentiment about AI involvement in elections is negative, most Americans likely don’t understand the full scope of how the technologies are being used by campaigns and outside forces, Buntain said. They probably don’t understand how they design your social media to feed your existing opinions and preconceptions.

Algorithms are built to promote angry and emotional content in streams, which can contribute to information silos and echo chambers.

Echo chambers aren’t an inherently bad thing — they can bring a sense of safety and community, Buntain said. And while there is algorithmic sorting on social media, people tend to self-sort into the streams they identify with. Lately, more conservatives are flocking to X after Elon Musk acquired the platform, and more liberal people are spending their time on TikTok, for example.

“In general, actually, the echo chambers in your offline world are much more echoic than the online echo chambers,” he said.

But the ad campaign is another system using AI “unseen” for more than a decade, Buntain said. Although AI has seemingly only been prominent for a few years — mostly since the launch of ChatGPT in 2022 — this type of search, classification, and targeted advertising of information has long been a tool of political campaigns.

The 2012 Obama for America campaign used data, technology and analytics to better reach American television audiences. This type of search, classification, and targeted advertising of information is the foundation of many AI systems today, and the strategies used by the Obama campaign were further defined and implemented for the 2016 and 2020 elections.

Today’s AI algorithms can mine information about you well beyond general demographics like age and gender to include unique interests and affiliations. This information is then used by campaigns for targeted advertising to almost all of your online spaces.

Outside of these “unseen” AI jobs, Buntain focused on the potential damage that Pew study participants were likely concerned about. People are often worried about inequalities and misinformation perpetuated by AI. They are also concerned about whether they can trust information provided by AI systems such as chatbots. Many are probably worried about whether they are connecting with a real person or a bot throughout the campaign cycle.

People are rightfully worried about these AI strategies and systems playing a role in the election, but Buntain worries about the ways AI can be used in the coming days, especially if it’s a very close race.

“AI tools will allow people to very quickly create content that makes things worse,” he said. “Five years ago, you could still make some kind of disinformation content, but it would take longer and be much more expensive.”

If you’re not a technologist, there’s a lot about artificial intelligence that probably mystifies you and amplifies societal concerns you already had, Buntain said.

“Is this all just a chat bot behind the scenes trying to get us to donate or trying to piss us off?” Buntain said. “I think that concern about, you know, ‘is it a genuine actor’ is a concern that AI really amplifies, but it’s a concern that’s certainly been around since 2016.”

Buntain hopes that the public’s perception of artificial intelligence will change over time. He believes that anxieties about it, particularly about its role in elections, are driven by wider societal issues such as the economy, a sense of security and the ability to trust information.

“I think we’re in a world that’s increasingly online, but still isolated, it makes us a little bit primed to … be negative about how these new technologies probably aren’t going to help us the way we thought they would” , he said.