Abstract
Systems and methods are described for monitoring recommendation feedback loops and automatically intervening in live ranking. A feedback loop intensity (FLI) for a user and topic is computed as a ratio of a user-conditional topic serving proportion to a global topic serving proportion over a rolling window. An organic interest score is computed from non-recommendation-channel evidence including one or more of search, direct navigation, creator following, or external engagement. An adjusted FLI is generated by dampening the raw FLI as a function of organic interest. When adjusted FLI exceeds configurable thresholds, a controller applies topic-specific, tiered interventions in the ranking pipeline, including expanding alternatives, applying a proportional topic cap, and applying a hard cap with exploration injection. Adjusted FLI values may be aggregated into system-level health metrics including mean adjusted FLI and prevalence above threshold for monitoring and evaluation.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Anonymous, "Feedback Loop Intensity Monitoring and Automatic Intervention for Recommendation Systems", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10732