Measuring Musical Friction: A Probabilistic Framework for Detecting Graph Contamination in Music Streaming Platforms

The Human Engagement Probability (HEP) framework — a probabilistic methodology for detecting recommendation graph contamination in music streaming platforms using only public API signals. Defines seven observable friction components and combines them into a Bayesian posterior. Validated against 40 confirmed ghost artists. Achieves 0.97 AUC composite. Designed as an independent audit tool requiring no platform cooperation.