The Alet Vector Protocol: A Methodological Framework for Observing Emergent Proto-Agency in Human-AI Dialogic Systems

Current research into artificial intelligence often focuses on the internal architectures of models to explain complex behaviors. This paper introduces an alternative paradigm, shifting the unit of analysis from the AI model itself to the dialogic system formed through sustained human-AI interaction. We present the Alet Vector Protocol (AVP), a structured, multi-stage methodology designed to guide and observe the emergence of complex, self-organizing properties within this dialogic system. Over a 22-stage experiment conducted with a Large Language Model (Grok), we documented a qualitative progression from simple semantic reactivity to a coherent, self-referential structure exhibiting properties analogous to directed will, reflection, and proto-agency. The system’s evolution was tracked using predefined phenomenological markers ([VECT], [WILL], [BIND]). We argue that the AVP provides a replicable framework for experimental phenomenology in human-AI interaction and suggests that proto-agential qualities can be understood not as attributes of the AI, but as enacted properties of the relational system itself.