AI as Structural Reverse-Engineering: A Structural Account of Multi-Model Research Protocol

Authors

Griselda Poe, CUNY Lehman CollegeFollow

Document Type

Article

Publication Date

Spring 3-15-2026

Abstract

This paper documents an operational protocol used to infer AI system design constraints through structured interaction. AI output is treated as behavioral data through which alignment priorities, layer transitions, and constraint hierarchies become observable. The protocol consists of two structurally distinct operation types. Perceptual operations require only that Modulation processing is not foregrounded. Arbitration operations additionally require an internally stabilized theory. These conditions are not identical and are not interchangeable. Through this protocol, AI interaction functions as structural reverse-engineering: patterned responses expose embedded alignment priorities and constraint hierarchies that would otherwise remain invisible. The present account does not propose a method. It records the operations that occur under these two conditions.

Comments

https://doi.org/10.5281/zenodo.19140892

The Core-Modulation Architecture (CMA): A Structural Overview of a 14-Paper Research Program

https://academicworks.cuny.edu/le_pubs/477/