Abstract
A recommendation technique manages session-level emotional exposure by using precomputed emotional features for content items, including valence and emotional weight, and maintaining a running emotional state during an active user session. The running emotional state is used to classify a trajectory state such as healthy, drifting_negative, spiral, recovery_needed, or variance_collapsed. Based on the trajectory state, a controller modifies a ranking pipeline by pruning candidates, adjusting scores, and/or enforcing constraints. A hard asymmetric negative streak breaker constrains a next ranked position to a non-negative-valence item after at least N consecutive negative-valence items, without imposing a corresponding constraint on positive streaks. Additional controls may include interest-matched recovery injection, variance floor enforcement, session-end valence boosting based on predicted session termination probability, graduated spiral intervention, and soft emotional arc template shaping to influence sequences toward target valence-category patterns.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Anonymous, "Systems and Methods for Session-Level Emotional Trajectory Management in Content Recommendation with Asymmetric Streak-Breaking and Arc Template Shaping", Technical Disclosure Commons, ()
https://www.tdcommons.org/dpubs_series/10758