TikTok's Explainable Recommendations: UX for Algorithmic Transparency

AI · 5 min read

TikTok’s explainability feature places a lightweight chip under each video that explains the top three signals used for recommendation—followed topics, recent interactions, and watch patterns. Tapping the chip reveals a compact modal with toggleable controls to de-prioritize signals and a brief educational snippet about how recommendations work. The UI is intentionally simple to avoid cognitive overload while giving users meaningful agency.

The product team prioritized signal surfacing that users can meaningfully act on. For example, if a recommendation is driven by a past liked sound, the user can opt to stop recommendations tied to that sound without blocking creators. This granular control supports experimentation while preserving content variety. The design uses progressive disclosure to keep the default surface clean.

From a design ethics perspective, TikTok’s approach is notable: it provides actionable controls rather than slogans. The challenge ahead is scaling these controls so they remain comprehensible as models get more complex. For UX teams, the lesson is that explainability succeeds when paired with simple, reversible actions.