# Recommendation Systems **Domain:** Machine Learning / Attention Governance **Doc Type:** System Class **Maturity:** Developed **Related:** [[Attention Economy]], [[Reward Prediction Error]], [[Variable-Ratio Schedule]], [[Objective Function]], [[Schedule and Loop]] ## Definition Recommendation systems rank and select items for a person using behavioral traces, content signals and continuously updated models. ## Schedule–Loop Context They embody the merger: cybernetic instrumentation constructs a model of the user while behaviorist objectives shape what is delivered and when. The constitutional question is who chooses the objective function and whether the subject can inspect or alter it. ## Backlinks - [[Schedule and Loop]] - [[articles/Schedule and Loop|Schedule and Loop]] - [[Russia and Prussia Kybernetiks]] ## Generative Retrieval and the Catalog Boundary [[articles/The Skeuomorphic Interval|The Skeuomorphic Interval]] follows recommendation from retrieve-and-rank into generative retrieval. The catalog boundary dissolves when a generated semantic identifier resolves to a synthesis instruction rather than a stored item. At that point recommendation participates in constructing the available artifact or environment, and the distinction between predicting and supplying a preference becomes constitutionally important. [[wiki/Provisioned Environment|Provisioned Environment]] names the larger architecture. [[wiki/Self-Confirming Representation|Self-Confirming Representation]] names the feedback failure in which behavior shaped by prior recommendations is treated as independent confirmation of the model.