Amplified by Design: TikTok's Engagement Logic, Misogynistic Content, and the Limits of Platform Governance
April 2026, Linnaeus University, Global Challenges in New Media and Management
Platform governance, Media policy, Digital rights, Academic
TikTok's recommendation system rewards misogynistic content because outrage counts as engagement. This paper argues the problem is the design, and that self-regulation, advertiser standards, and the Digital Services Act all miss it by going after posts instead.
What the paper argues
TikTok passed a billion monthly users in 2023. Behind the entertainment is an advertising business that makes money from attention, and a recommendation system built to keep people watching. The paper’s claim is that misogynistic content does well on TikTok because of that system, not despite it. Outrage is engagement. Agreement is engagement. The algorithm can’t tell the difference and has no reason to care.
Three attempts at governance, one shared blind spot
Platform self-regulation, advertiser standards, and the EU’s Digital Services Act have all tried to deal with this, and all three go after content.
Self-regulation (community guidelines, bans) produces a waterbed effect. Banned creators move to Telegram or BitChute and the system that amplified them stays put.
Advertiser-driven standards like GARM protect brands, not users. They draw the line at illegal content, which leaves most manosphere material comfortably inside the “brand safe” zone.
The DSA is the most serious of the three and still frames the problem as speech to be removed “where proportionate.” It doesn’t ask why the content circulates at scale in the first place.
What would work instead
The paper ends with three shifts. Audit the recommendation system, not the posts: independent, mandatory, with an obligation to show that engagement optimization doesn’t structurally favor harmful content. Measure risk against democratic rights rather than legal minimums. And treat online misogyny as a structural question about who gets pushed out of public life, not as a moderation backlog.
The full paper and the presentation slides are linked above.