TikTok Algorithm Transparency: Design Audit of the New 'Why This Video' Panel

Tech · 5 min read

TikTok updated the 'Why This Video' panel to provide more granular signals — such as watched duration, shared trends, and creator interactions — in a compact, scannable layout. The panel groups signals into categories (behavioral, content, and network) with iconography and brief tooltips. Designers faced the challenge of making opaque ML features understandable without exposing model internals.

The panel also includes a 'Control My Feed' row where users can boost, hide, or see fewer recommendations based on specific signals. This direct manipulation model improved perceived control; early metrics from the design team showed a drop in feed frustration reports. The panel's copy uses neutral non-technical language, and microcopy explains the limits of personalization and how feedback shapes future recommendations.

The teardown highlights broader implications: transparency without agency can create user anxiety, so pairing explanations with control is essential. For product designers, the case study illustrates balancing regulatory compliance, pedagogy, and product metrics while avoiding overwhelming users with raw model features.