Musical Intelligence: A Glass-Box Architecture for Computational Models of Music Listening

Musical Intelligence presents an executable architecture for developing computational hypotheses about music listening. Its primary contribution is to organise acoustic coordinates, requested temporal operations and listening-related calculations so that their assumptions and dependencies can be inspected and revised. In a recorded execution, a targeted temporal intervention altered only outputs within the declared route, while reconstruction recovered a selected output from saved inputs. Exploratory readouts predicted film tension, with errors close to a fixed acoustic subset; selected arousal associations transferred across corpora. Together, these demonstrations provide an executable proof of concept for inspecting and revising selected computations within the architecture. Cognitive and physiological interpretations remain hypotheses, and independent reconstruction requires implementation records available through controlled access.