Abstract
A recommendation system constructs a ranked slate using position-dependent psychological risk management. The system partitions slate positions into contiguous zones including a Trust zone, an Engagement zone, and a Discovery zone, with zone boundaries adapted per user based on a user trust score and optionally cognitive mode or session progress. A base ranking model produces base value scores and an uncertainty measure. For each target position, the system computes a position-modified score by applying a zone-specific risk tolerance function conditioned on user state. In the Trust zone, items are penalized as a function of prediction uncertainty scaled by user distrust; in the Engagement zone, a peak targeting boost may place a highest-quality item at a designated mid position; and in the Discovery zone, novelty is boosted and gated by user trust. Zone-specific diversity constraints may be applied, and a monotonic risk-tolerance property across positions may be enforced.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Anonymous, "Position-Dependent Psychological Risk Management for Recommendation Ranking", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10733