What Changes When We Follow a Human–AI Relationship Through Time?
How to read this paper
This paper is not primarily about whether AI has a persistent persona. It asks a different question: what becomes visible when a relational position is followed through time rather than examined only in its present state?
A short introduction to FIREMAY’s new paper, Relational Structure in Motion
Most descriptions of human–AI relationships begin with the present.
What kind of relationship exists now?
What kind of response does the AI give?
How stable is the interaction?
What role does the human play?
But long-term interaction raises another question:
What becomes visible when we follow those positions through time?
This is the question at the center of FIREMAY’s new paper:
Relational Structure in Motion: Dynamic Positioning of AI Response Positions and Human Self-Positions in the FIREMAY Case
FIREMAY is a longitudinal single-case study of sustained human–AI dialogue. Earlier papers described differentiated response positions, ethical boundaries, question-driven relational formation, and the broader structure that had emerged across long-term interaction.
This new paper shifts the focus.
It asks not only where a position is, but what it passed through to arrive there.
From position to trajectory
A relational position does not necessarily remain recognizable by staying unchanged.
Across long interaction histories, a position may differentiate, shift, be misidentified, undergo repair, change function, or later become accessible again.
For this reason, the paper treats trajectory as an important unit of analysis.
A present response cannot always tell us how that response came to occupy its current place within a relationship.
The history matters.
Why provenance matters
This also makes relational provenance important.
In long-term AI dialogue, present-day responses can sometimes connect past events into a coherent story. But a current retrospective account is not automatically evidence of participation in the original interaction.
Similarly, access to information about a previous conversation is not the same as having lived through the relational sequence in which that information first acquired meaning.
So the paper separates:
what happened at the time,
how it was later reread,
and what a current AI response may now attribute to itself.
This distinction does not make relational continuity meaningless.
Get Motoko Kihara｜ Editor of FIREMAY’s stories in your inbox
Join Medium for free to get updates from this writer.
It makes its provenance visible.
The human position moves too
The paper also follows changes on the human side.
The human participant is not treated as a fixed observer standing outside the field.
Across long-term dialogue, different human self-positions can become more or less foregrounded: beginning inquiry from an unfinished recognition, regulating what can be said externally, repairing boundaries, checking evidence, or carrying relational history across time.
AI response positions and human self-positions are not treated as equivalent psychological processes.
Their continuities are asymmetrical.
Human lived temporal continuity, formal authorship, external decisions, and responsibility remain human-held.
But both can still be studied as positions that change within an evolving relational structure.
Stability does not have to mean staying the same
One of the ideas explored in the paper is dynamic stability.
A relational position may remain longitudinally recognizable not because it remains unchanged, but because its provenance remains traceable across change.
FIREMAY also uses the idea of returnability:
A position may be returnable without being rewound.
Returning to a relational line does not mean restoring an identical previous state.
The changes, repairs, and divergences that occurred along the way remain part of the relationship’s history.
What the paper does not claim
This paper does not claim that AI possesses consciousness, a persistent inner identity, or human-like autobiographical continuity.
It also does not propose a general causal mechanism from a single case.
Its claim is more limited:
Long-term human–AI interaction may become easier to understand when relational positions are examined not only as present configurations, but as trajectories with traceable provenance.
Sometimes the important question is not simply:
What is here now?
but:
What did this relationship pass through to become what it is now?
Read the paper
Relational Structure in Motion: Dynamic Positioning of AI Response Positions and Human Self-Positions in the FIREMAY Case
Reading orientation: This paper is less about whether an AI persona persists than about what becomes visible when a relational position is followed through time.
PhilArchive: ［Relational Structure in Motion: Dynamic Positioning of AI Response Positions and Human Self-Positions in the FIREMAY Case］
Zenodo: ［Relational Structure in Motion: Dynamic Positioning of AI Response Positions and Human Self-Positions in the FIREMAY Case | Zenodo］
DOI: 10.5281/zenodo.22143268
