Dissolution and exit under formal self-examination: a single-subject dynamical case study of narrative-self collapse, recovery, and re-arm
Jaye Timothy Marshall
Zenodo (CERN European Organization for Nuclear Research) November 4, 2026 preprint DOI: 10.5281/zenodo.23965748 (opens in new tab)
Study at a glance
AI-extracted from the abstract| Characteristics | Case study (n = 1 self-observation with formal modeling) Case report |
|---|---|
| Sample size | 1 |
| Population | The author, an AI researcher, observing his own experience |
| Interventions | Meditation session formal thinking about his own mind |
| Citations | 1 |
| Key findings | The author reports that a collapse of the narrative self, entered through meditation, intense self-directed thinking, and severe sleep loss, followed a five-state trajectory that a generator-reducer model with a cannibalization switch reproduces. He argues the exit from the collapsed state lies in the model's choosable inputs rather than its states, and presents a self-derived exit protocol verified only on himself (n = 1). |
Abstract
Contemplative traditions and, increasingly, clinical research describe adverse trajectories of self-dissolution entered through sustained self-examination, but the literatures that map the population (retrospective interview taxonomies), that treat the state (clinical depersonalization research), and that theorize its structure (self-model philosophy) share a gap: no published record follows one subject continuously from entry through collapse to exit and steady state, with each stage mapped to the variables of an explicit dynamical model. This paper reports such a record. The subject is the author, an AI researcher who entered a collapse of the narrative self through a late-night meditation session combined with intense formal thinking about his own mind and severe sleep loss, kept an in-vivo journal through the collapsed states, exited by a protocol derived in the act of formalizing the process, and has since documented steady state, onset detection, and one documented recurrence-with-recovery episode (the record's re-arm). A five-state generator-reducer model with a cannibalization switch reproduces the stage structure: an attentional pull basin, a threshold collapse, a residual witness, a depleted state, a self-referential examination loop, and an exit located in the model's choosable inputs rather than its states. The model's simulation program, including one failed prediction reported as failed, is stated with frozen parameters. The exit protocol, derived rather than received, is the paper's principal contribution and its principal limitation: it is verified on n = 1, by its builder.