AI Generated Image
TikTok users don't have as much agency over their FYPs as they think
Ars Technica • • science
Key Points:
- Northwestern University researchers conducted an audit of TikTok's For You Page (FYP) algorithm using 90 bot accounts to test how effectively negative feedback influences content recommendations.
- Their findings show that the "not interested" button reduces unwanted content by about 84%, while simply skipping videos only reduces it by 48%; however, the "not interested" option is often hidden from users.
- The algorithm responds to negative feedback initially by reducing certain content but gradually reintroduces it unless users consistently provide the same negative signals, indicating a relapse effect.
- The study highlights that user behavior, such as re-engaging with unwanted content, can override negative feedback, suggesting that continuous and active feedback is necessary to curate the FYP effectively.
- Researchers emphasize the limitations of aggregated data access for studying personalization and hope to validate their findings with real user data in the future, noting that platform design fundamentally shapes user experience.
