In 2021, President- elect Donald Trump filed a lawsuit against tech giants Google, Twitter, and Facebook, alleging that he was a victim of censorship. The fact that social media platform can block accounts or censor speeches raises the critical question of whether social media platforms have the power to influence political opinions on a global scale. Tauhid Zaman, Associate Professor at the Yale School of Management, shared how the manipulation of online social media constitutes a national security threat: “We don’t need tanks and bombs; we need Tweets and Bots.” Dr. Zaman spoke at Sasin Research Seminar titled “Social Media Suspensions and Shadow Banning: Political Bias or Genuine Disinformation Control?” Dr. Zaman talked about two kinds of censorship on social media, including suspension and shadow banning. Suspension on social media is the removal of a user’s account off the platform usually used on the most dangerous accounts like terrorists or human trafficking accounts. Shadow banning is a subtle mechanism for censoring people, used by social media platform where certain content is less visible to users, often without their knowledge. “A form of shadow banning is when a platform just very quietly removes some of the posts, so users will see less tweets or make their content disappear without them knowing,” said Dr. Zaman. Having worked in the operations management group at Yale for a decade, Dr. Zaman identified information operations as the capabilities to protect information environments or the online environment. He identified steps on how people working in this field can monitor social media content online:

• Monitor: Deciding what to monitor and collect relevant users and posts (C.E. Marks and T. Zaman, Operations Research (2022)
• Identify: Identify the threats whether it is a person spreading rumors, terrorists, Bots artificially shifting the debate online
• Assess: quantify the impact of influence campaigns
• Counter: counter the impact of influence campaigns

Political Bias in Twitter Suspensions Dr. Zaman’s research team tracked 9,000 twitter users in 2020, who used political hashtags in #VoteBidenHarris2020 #Trump2020 with a random sample of 4,500 democrats and Trump supporters. After seven months, they checked who was suspended on Twitter. As Dr. Zaman shared in an October 2024 interview with Yale Insights, “Accounts sharing pro-Trump or conservative hashtags were suspended at a significantly higher rate than those sharing pro-Biden or liberal hashtags—they were about 4.4 times more likely to be suspended. At face value, this pattern could be interpreted as evidence of bias against conservatives.” However, an alternative theory is that the platform’s suspension policies are non-political, focusing on violations of terms of service like misinformation or hate speech, regardless of political party. This highlights the need for further analysis of the quality of news shared on Twitter, as understanding the reasons behind such suspensions is crucial to determining whether political content is being unfairly targeted or if these suspensions are based on other non-political criteria. Dr. Zaman also explained that Twitter’s effort to remain politically neutral wasn’t due to the platform’s intentional or ideological policies, it was a result of the polar differences in the media consumption patterns of the political groups. “If one half of the country believes in one set of news and the other half believes in other set of news I don’t think that country can last-you’re going to have fighting, and insurrections and oh, we already have that, it already happened,” he said. Manipulating Opinions with Shadow Banning Dr. Zaman’s research further explores how social media influences can manipulate political opinions using models like the DeGroot and Bounded Confidence Models. According to “Shaping Opinions in Social Networks with Shadow Banning,” co-authored by Yen-Shao Chen and Dr. Zaman, the DeGroot Model describes how opinions move and “measures how much a single post can shift one’s opinion, leading to opinion consensus on most networks.” The Bounded Confidence Model introduces a tolerance range, explaining why polarization persists when opinions lie outside an individual’s confidence interval, potentially triggering a backfire effect. To shape opinion dynamics, platforms use shadow banning, reducing tweet visibility to control how opinions spread, with the strength of shadow banning influencing the rate of change in opinion objectives like adjusting the opinion mean (push everyone to support a political extreme) or opinion variance (creating consensus or polarization). By optimizing these controls through linear programming, platforms can subtly manipulate public discourse while appearing unbiased. Through these models, Dr. Zaman’s research team developed a simulation of a social network to study shadow banning’s effects. They demonstrated that shadow banning could shift users’ opinions and alter polarization levels. Interestingly, even when the goal was to steer collective sentiment left or right, the moderation tactic appeared neutral. Dr. Zaman found that opinions could be shifted by simultaneously lowering the visibility of accounts on both sides of a debate, creating the illusion of impartiality. He cautioned that while social media may seem impartial when employing shadow banning, people’s opinions will inevitably shift, and by the time the effects become apparent, it may already be too late to address them (Chen Y-S, Zaman, T., 2024). “It’s like frog in a pond of water that’s boiling. In the beginning the water is warm, it’s fine, no problem, you’re relaxing, and suddenly you are cooked,” said Dr. Zaman, “If you don’t watch your platform and measure them the right way, you can be cooked very seriously, because of shadow banning.”   Citations:

1. Chen Y-S, Zaman, T. (2024) Shaping opinions in social networks with shadow banning. PLOS ONE 19(3): e0299977. https://doi.org/10.1371/journal.pone.0299977
2. Mosleh, M., Yang, Q., Zaman, T., Pennycook, G., & Rand, D. G. (2022, April 8). Unbiased misinformation policies sanction conservatives more than liberals. https://doi.org/10.31234/osf.io/ay9q5
3. Zaman, T. (2024, October 15). Do social media platforms suspend conservatives more? Yale Insights. Retrieved from https://insights.som.yale.edu/insights/do-social-media-platforms-suspend-conservatives-more