Misinformation ahead of US midterm elections: “My biggest concern is the division over what counts as fact”

taz: Ms. DiResta, you have studied disinformation and propaganda in politics for years at the Stanford Internet Observatory. The institute was closed under pressure from Washington. How did that happen?

Renée DiResta: Our work was about election integrity. We monitored the 2020 and 2022 elections and reported our findings. This was initially portrayed by a man named Mike Benz as a censorship plot against right-wing voices. Benz was an employee of the Trump administration and denied the results of the 2020 election. He also briefed Republican members of Congress who he expected would be receptive to these allegations. After the Republicans won the majority in the House of Representatives in November 2022, they began investigating us. There was a congressional hearing, and a few days later we received a letter asking us to provide our emails to the government. At the same time, we were sued by America First Legal, the right-wing legal organization founded by Stephen Miller. After that, our organization was phased out.

taz: You wrote a book about your findings from the project and these incidents. The central thesis of the book is that disinformation arises from the interaction of influencers, algorithms and the crowd (the audience). How do these three elements work together?

DiResta: You can't look at these elements in isolation. Influencers such as Trump or Musk have a large following. Your content is aimed at both this following – the crowd – and the algorithm. For example, if you publish something on Instagram that your target group should actually love, but don't prepare it in such a way that the algorithm prefers it, hardly anyone will see it. In addition, your own content is increasingly no longer automatically displayed to your followers. Instead, they compete with all other potentially suitable content – ​​so-called unrelated content – ​​to appear in users' feeds. This content now makes up 40 percent of what you see in your feed. The algorithm, in turn, is based on how the crowd reacts to content. When a post is played, what matters is how much attention it receives in the first hour. It depends on whether it will be shown to more people. The algorithm therefore relies on feedback from the crowd. At the same time, the algorithm decides whether an influencer’s content becomes visible at all. After all, the crowd itself influences the influencers. The crowd can apply pressure: Why don't you talk about Gaza? Why did you express this opinion? This phenomenon is called “audience capture”: the crowd can influence what an influencer writes and talks about because they are afraid of losing their following.

taz: When you were at Stanford, your team flagged disinformation and informed the platforms about it. They then reacted to it - or not. What was the response?

DiResta: Our report says that the platforms did nothing in 60 percent of cases. So that wasn't particularly effective.

researches disinformation and propaganda in US politics at Georgetown University in Washington. Until 2024, she will work at the Stanford Internet Observatory, a watchdog group dedicated to combating disinformation. The lab was closed under pressure from Republican politicians. Most recently, she wrote the book “Invisible Rulers – The People Who Turn Lies Into Reality” about her experiences and insights.

taz: Do the platforms still have effective mechanisms to combat disinformation in their own ecosystems and moderate content?

DiResta: There are still internal election integrity teams. Your job is to ensure that posts that violate the respective guidelines do not go viral. Most of the time, such posts are flagged. When content is actually removed, it is mostly due to attempted fraud or threats of violence - for example in the run-up to January 6th. While my team did not specifically research January 6th, we did investigate which of the content we reported was removed. It was about 10 percent. The content removed was primarily about identity abuse, fraud and phishing attempts. Posts from people claiming the election was rigged were rarely deleted.

taz: The lie about the allegedly stolen election was probably the most consequential disinformation of the past election cycles. What do you think is the most dangerous misinformation impacting the current midterm election campaign?

DiResta: That's an interesting question. I now run a bot on X that writes community notes - a user-based system for self-regulation. Last year, this program was opened to AI-powered authors. This year I decided to build such a bot myself. Why I relate this to your question: I can now see what corrections are being requested - what users are reporting and what they want corrected. It's mostly pretty predictable. The details change, but the basic narrative patterns remain the same. For example, I receive many inquiries about allegedly illegal votes from undocumented immigrants. Others revolve around the supposed uncertainty of elections due to postal voting. Over the past week there have been many inquiries about James Talarico and his alleged commitment to a psychiatric facility. In addition to the question of which topics are important to people and which rumors are spread, the most important thing is that even very solid corrections are often devalued. You can, for example, present information from a government agency or official statistics on illegal votes. Nevertheless, other users rate the correction negatively. They say it is an opinion, the source is not reliable enough or the information is incomplete. I was surprised at how often the debate ultimately comes down to what counts as fact. So I don't have a concrete answer as to which individual rumor is the biggest piece of misinformation. My biggest concern is the huge public divide over what is considered fact and what sources or evidence is considered acceptable.

taz: You say that the most common misinformation is largely immune to corrections by government agencies, traditional media or fact-checkers. Are there any effective ways to combat disinformation and propaganda?

DiResta: When the audience trusts the sender, they are receptive to correction. The problem is that this trust is often broken. When people distrust the institutions that offer corrections, rebuttals and fact-checks are ineffective. That's why I argue in my book that we need more than fact checks. We need credible communicators who build relationships within communities, media literacy that helps people identify recurring rumors and propaganda techniques, and platforms that don't constantly reward sensationalism and outrage.

taz: In your book you write about the psychology of rumors and disinformation. Which human impulses do influencers and algorithms appeal to?

DiResta: Identity. Belonging. Companionship. The feeling of supporting the same cause. People want to feel like they are part of something meaningful and know something that is hidden from others. Rumors invite participation: people share stories, examine supposed evidence, connect dots, and work together to make sense of events. Algorithmic amplification and curation have vastly accelerated this ancient human process. Influencers give people a story to get excited about and give them a sense of community. That feeling of being involved is rewarding. You're no longer just consuming a story, you're helping to ensure that other people also learn the "truth."

taz: AI plays an immense role in this election campaign. How has it changed the interaction between influencers, algorithms and the crowd? Has it made combating disinformation infinitely more difficult?

DiResta: AI has not fundamentally changed the interaction between influencers, algorithms and crowd, but it has changed the economic conditions. The technology is now very, very powerful. It has become extremely difficult to recognize what is real. At the same time, producing such content is cheap, which is why there is a lot more of it. However, checking them is becoming more and more complex. Fighting disinformation is simply no longer worth it.