When Conversation History Becomes Infrastructure: Operationalising High-Coherence Interaction State in Human–LLM Dyads
A framework for examining how conversation history shapes later interaction: interpreting shared context, reusing earlier material, retaining corrections and tracking change over time.
What it contributes
The author demonstrates HCIS retrospectively across five selected, naturally occurring human–LLM conversations spanning 94 to 2,730 moves. The analysis distinguishes accurate contextual interpretation and reuse from lasting behavioural adaptation, reporting a profile of separate functions and their trajectories rather than a single score.
Read with care
Five purposively selected cases demonstrate the method rather than establish population prevalence, causal effects or validated thresholds. Independent reliability and validity testing remains necessary. This is the Society founder’s own work; this entry signposts it rather than providing an independent review. Interaction coherence should not be equated with factual accuracy, safety or wellbeing.
Why it belongs here
For researchers and practitioners asking whether a system uses accumulated conversation history effectively, beyond simply retaining or recalling it.
Zenodo lists this as a preprint, version 2.0, published 12 October 2026. Peer-reviewed publication is not claimed. DOI: 10.5281/zenodo.23322029.
Source reading depthZenodo repository abstract and version metadata checked. The full manuscript was not assessed for this source note.
Source check: 15 October 2026. Reading cautions are editorial interpretation, not quotations from the authors.