Algorithmic “Gatekeeping”
Social media restrictions can render newsworthy reporting on controversial topics invisible. This is especially so in the case of so-called “shadows bans,” a blanket term for a variety of content reduction practices that social media platforms use to reduce the visibility of social media content without the content creator’s knowledge.
This unit explains how “shadow bans” can restrict the reach of legitimate, newsworthy content and provides five basic recommendations for recognizing and responding to this controversial form of content moderation.
Although major platforms dispute or deny accusations of shadow bans, studies indicate real impacts. For example, Monica Horten, an internet policy analyst, found that Facebook’s “feature block” function created a “sudden plunge” of 93–99 percent in the reach of UK-based Pages, or “not much different from completely unpublishing the page.”
Shadow bans highlight the potential for social media platforms to manipulate public opinion. Tarleton Gillespie has warned that using algorithmic recommendations to reduce the visibility of targeted content has sweeping consequences, influencing “not only what any one user is likely to see but also what society is likely to attend to, take seriously, struggle with, and value.”
Those consequences disproportionately impact members of marginalized communities, online and offline, as Black TikTok creators and LGBTQ+ content creators on Instagram have called out. As a veiled form of content moderation, shadow bans make it more difficult to hold platforms accountable for restricting (already) marginalized voices.
These are special concerns for journalists and newsrooms that cover topics—including police violence, institutional racism, and Palestine—that Facebook, X (Twitter), Instagram, TikTok, and other prominent social media platforms deem “controversial” or at odds with “community standards.”
Some practical tips from the Don’t Delete Art campaign and website are applicable to journalists:
Beware of sudden, sharp drops in the reach of your account. You can use analytics provided by each platform to monitor your account’s reach and, on many platforms, the performance of specific posts.
You can also benchmark your content’s reach against similar posts from other accounts with comparable engagement.
There are also third-party, platform-specific tools designed to track evidence of shadow banning. For example, the Shadowban Scanner extension for the Chrome browser and HiSubway’s Twitter Shadowban Test allow you to monitor content reduction on X/Twitter.
In her study of shadow banning on Facebook, Horten found that posting behavior could trigger a shadow ban, independent of posts’ content.
Avoid posting too frequently, soliciting likes too often, using the same hashtags in every post, and other types of “spammy” behavior. In brief, don’t act like a bot!
Automated content moderation systems were originally deployed to detect spam, and later disinformation. Although much of the controversy around shadow banning focuses on content, automated content moderation systems continue to track behavior as well, reducing the visibility of accounts that engage in online behavior those platforms deem inauthentic, abusive, or manipulative.
Many creators with legitimate reasons to address sensitive topics on social media platforms use “algospeak,” coded language designed to evade algorithmic detection by using substitutes for banned terms.
For example, in their study of TikTok creators who post sensitive content about race, gender, or sex and sexuality, Kendra Calhoun and Alexia Fawcett identified four major categories of linguistic innovations to avoid censorship, including creative use of language (unalive in place of “killed”) and word replacement (using accounting for “sex work”); and substitute terms that play on similar sounds (seggsy for “sexy,” droogs for “drugs”).
Although algospeak may provide many social media users with means to avoid censorship, we fear that its use in promoting journalism might undermine the credibility of the reporting for some readers. It may also violate many newsroom’s standards for clear use of language.
For video, publish potentially sensitive text in the video’s image, rather than in its description. This makes it harder for automated content moderation systems to track and flag the sensitive content.
Of course, the fundamental principles of ethical journalism apply: Content should reflect the cardinal professional values of maximizing transparency and minimizing harm in the course of seeking and reporting truth.
“Sensitive content” warnings can help protect your posts from shadow bans and takedowns. Ryan Sorrell described content warnings as “the single biggest protective mechanism” the Kansas City Defender uses to avoid content reduction or removal.
There are built-in options for adding sensitive content warnings on Instagram, for instance. It’s important to make your warning clear and concise. Avoid generalities, such as “This post may not be suitable for all viewers,” in favor of direct language that clearly represents the content that follows, such as “Sensitive content: gun violence.”
Place the warning prominently. Use a font style and size that is easy to read. Use the post’s caption to contextualize the content.
Preview the content warning and make any necessary adjustments before you post.
