Question-Driven Relational Field Formation: A Preliminary FIREMAY Framework for Long-Term Human–AI Dialogue

Abstract This paper positions FIREMAY not as prompt engineering, nor as a claim that AI possesses personality or inner life, but as a preliminary framework for describing question-driven relational field formation in long-term human–AI dialogue. Based on a first-person long-term observation, FIREMAY examines how the human participant’s questions, discomfort, boundaries, memory practices, and changing trust shape AI response positions and relational fields over time. This paper introduces several initial concepts, including relational personas, response positions, relational self, forgetting patterns, save points, structural trust, and the boundary/firebowl rule. These concepts are not intended to prove AI consciousness or inner life. Rather, they provide a vocabulary for describing how inquiry, response, forgetting, trust, boundaries, and self-positioning form relational fields in long-term human–AI dialogue. The paper presents FIREMAY not as a completed theory, but as an early lens for examining how relational structures may emerge, stabilize, waver, require boundaries, and become available for re-reading through sustained human–AI interaction.