Fighting Social Media Manipulation in Elections: A Four-Step Plan

As the 2020 U.S. presidential election approaches, researchers are increasingly concerned about the threat of social media manipulation, a problem that has plagued democratic processes since 2016. Two MIT professors have proposed a comprehensive research framework designed to measure, analyze, and defend against this growing threat to electoral integrity.

MIT professor of IT and marketing Sinan Aral and associate professor of marketing Dean Eckles outlined their four-step process in the August 30 edition of Science magazine, emphasizing that “without an organized research agenda that informs policy, democracies will remain vulnerable to foreign and domestic attacks.”

Their proposal comes at a critical time, as social media platforms like Facebook and Twitter continue to grapple with the spread of misinformation. With just one year remaining until the next presidential election, the question has shifted from whether manipulation will occur to how effectively it can be countered.

The researchers’ first recommendation focuses on cataloguing exposures to manipulation. This involves documenting what messages were shared across platforms, what advertising methods were employed, which social media sites were utilized, and how content spread among users. The professors emphasize that analyzing the content itself is crucial for understanding its effectiveness in influencing voters.

The second step involves combining exposure data with voting behavior records. While previous attempts to connect social media accounts with public voting records relied on self-reported profile information, Aral and Eckles suggest leveraging location data already collected by social media companies. This could include matching voter registration with home addresses derived from mobile location information—data already used extensively for marketing purposes.

For their third recommendation, the MIT professors address the challenge of accurately assessing the effectiveness of manipulative messages. They propose adapting the A/B testing and algorithm evaluation methods already employed by social media companies to measure the impact of exposure to misinformation on voter behavior.

The final step involves calculating the consequences of changes in voting behavior. By combining predicted voter behavior—both with and without exposure to misinformation—with geographic and demographic data specific to an election, researchers could develop more accurate estimates of vote totals in particular regions.

While their proposal involves handling sensitive data, Aral and Eckles are mindful of privacy concerns. They do not advocate for unlimited access to personal information but rather encourage collaboration with social media platforms and the development of “privacy-preserving methods for record linkage and content analysis” that balance privacy protection with research needs.

“Achieving a scientific understanding of the effects of social media manipulation on elections is an important civic duty,” write the researchers. “The sooner we begin a public discussion of the trade-offs between privacy, free speech, and democracy that arise from the pursuit of this science, the sooner we can realize a path forward.”

The proposal reflects growing concerns within the academic community about the vulnerability of democratic processes to digital manipulation. As social media’s influence on public opinion continues to expand, developing robust methodologies to identify, measure, and counter misinformation campaigns has become increasingly urgent.

With major technology companies facing mounting pressure to address these issues, the MIT researchers’ framework offers a potential roadmap for collaborative efforts between academia, industry, and policymakers to safeguard electoral integrity in an increasingly digital world.