Abstract

Current Generative AI interfaces often sacrifice transparency and human authorship through opaque, single-prompt "magic button" interactions. This paper introduces Hierarchical Creative Control (HCC), an architecture for lyric generation that restores agency via structured, multi-level collaboration. HCC organizes creative decisions across global, segmental, and micro-segmental levels, using a state-inheritance mechanism where attributes propagate dynamically unless explicitly overridden. Unlike static templates, the architecture employs a deterministic Context Resolver to translate high-level abstractions into executable constraints. Grounded in Adaptive Structuration Theory, the study positions co-creation as a socio-technical process, contributing a framework where accountability and provenance are embedded architectural features rather than post-hoc evaluations.

Download

Share

COinS