Layer Attribution Framework for Diagnosing AI Agent Behavior

This framework distinguishes between agent behaviors arising from the computational substrate, such as architecture and memory, and those shaped by the behavioral modulation layer, including objectives and governance. It provides a method to evaluate surrogate validity and identify human-AI divergence in complex systems.