TikTok says it will adjust its recommendation algorithm to avoid negative reinforcement and let users pick words and hashtags to remove videos from their feed
Liza Lin / Wall Street Journal :
Context & Ripple Effects
TikTok had already expanded its Community Guidelines and well-being tools around harassment, self-harm, and violence. The new feed controls move that safety work into the recommendation experience, giving people a way to signal content they do not want surfaced.
The change also extends TikTok's prior use of user-facing controls, including comment-approval and anti-harassment prompts, from interactions around a post to the selection of videos in the feed.
First-order effects
- TikTok users can exclude videos associated with chosen words and hashtags, while TikTok's recommendation system is adjusted to avoid reinforcing unwanted viewing patterns.
- Creators whose distribution depends on particular terms or hashtags can lose reach among users who opt to filter those signals.
Second-order effects
- TikTok's safety and product teams must treat user exclusions as an input to feed ranking, linking moderation-oriented controls more directly to recommendation design.
- Other short-video platforms face a clearer product benchmark: feed personalization now includes tools for reducing exposure, not only tools for finding more content.
Third-order effects
- If such controls become standard, responsibility for harmful recommendation loops shifts from rules enforcement alone toward auditable choices in how ranking systems interpret negative feedback.
- The pattern points to a feed-governance model that combines platform safeguards with individual preference controls, rather than relying on either alone.
The trend: Social platforms are recasting digital well-being as a recommendation-system design problem, with user controls supplementing content rules.