Degree

Doctor of Philosophy (PhD)

Department

Music

Document Type

Dissertation

Abstract

Across the nineteenth and twentieth centuries, artistic practice repeatedly absorbed new machines: from mechanical instruments and recording media to electronic studios and software. That progression now reaches machine intelligence that can plan and act inside a live performance loop. Putting intelligence into realtime performance changes how live audiovisual work is created and performed, and therefore how it must be studied.

How can we study human–computer interaction when machine intelligence can act in realtime audiovisual performance? What is missing is a design theory for that study. This dissertation proposes Neural Parallel Intelligence (NPI) as that design theory. NPI rests on four design requirements: parallel cognition, contextual memory, environmental binding (success checked how and where on the intended surface), and inspectable delegation. “Neural” names tiered intelligent orchestration paths (deterministic, local, cloud), not brain simulation. Two principles complete NPI. The Conversational Performance Loop (CPL) is the staged negotiation protocol that keeps human–machine shared control inspectable under load while cutting serial hand-wiring work. The Multiple Hands Principle (MHP) explains why one performer needs bounded parallel helpers (virtual hands) while remaining the authority who states goals, judges results, and can veto.

Implementation evidence appears in patchBAI (NTERNN orchestration, patchWORLD environmental-context stores, patchSCRIPT domain language), which makes those principles inspectable. The doctoral recital song.play (May 7, 2026) at the Louisiana State University (LSU) extended-reality (XR) Studio is the practice site where that design theory is put under live load. The primary research question (PRQ) asks how NPI, completed by CPL and MHP, provides a design theory for studying that interaction in live audiovisual performance. The supporting research question (SRQ) compares Connect, Type, and Talk on matched tasks to ask which modality reduces steps and time to completion, and whether the intelligence-connected modalities (Type and Talk) are faster than manual wiring. Beyond this recital, NPI is one Intelligence Architecture: a reusable shared-control pattern for human judgment and bounded machine intelligence under live load.

Date

7-17-2026

Committee Chair

Stephen Beck

LSU Acknowledgement

1

LSU Accessibility Acknowledgment

1

Download

Available for download on Saturday, July 17, 2027

Share

COinS