Social media allows users to express themselves through words, pictures, videos and emojis, but some marginalized groups say social media platforms restrict the visibility of their online posts, according to a new University of Michigan study.

In what is described as shadowbanning, platform administrators flag certain users and content and reduce their reach or audience.

Major platforms, including Instagram, X (formerly Twitter), TikTok and Facebook, have released statements distancing themselves from shadowbanning or claim it does not exist on their respective platforms.

Meanwhile, based on the study's findings, some social media users disagree with that assessment. They have attempted to determine whether their content is being suppressed by engaging in what U-M researchers call "collaborative algorithm investigation"—testing each other's suspicions about being shadowbanned and reporting findings to one another.

Oliver Haimson, assistant professor of information, said users from different races, genders and sexual orientations completed surveys and were interviewed about what happened with their social media posts.

If they believed they had been shadowbanned, many users felt frustrated, had fewer social media engagements and held negative perceptions of platforms, said researcher Samuel Mayworm, a doctoral student at the U-M School of Information.

The findings, which appear in Proceedings of the ACM on Human-Computer Interaction, align with past literature on disproportionate content moderation of marginalized users' online content, particularly for Black and transgender users.

"The platforms can help these marginalized groups by improving their communication related to shadowbanning (especially about why certain categories of content are suppressed) and by validating users' experiences instead of denying that they suppress content," Mayworm said.

More information: Daniel Delmonaco et al, "What are you doing, TikTok" : How Marginalized Social Media Users Perceive, Theorize, and "Prove" Shadowbanning, Proceedings of the ACM on Human-Computer Interaction (2024). DOI: 10.7302/22437