Method / provenance
How it is made
The paper is a deliberately transparent synthesis of one framework diagram, one supplied transcript, and an iterative human–AI argument.
The starting material was a TPIT/LVT concept figure and a transcript discussing thought, self-identification, unconscious influence, and detachment from thought. The dialogue first mapped that transcript into TPIT/LVT, then reframed socially constructed narratives as learned control structures, then examined System 1/System 2 through arithmetic, then asked whether recursive symbolic processing helps explain human cognitive distinctiveness, and finally focused on variance between humans in the capacity and tendency to sustain reflective processing. This paper compresses that sequence into one functional hypothesis: reflective cognition can be modeled as recursive control over learned token transformations.
1 · Framework supplied
The TPIT/LVT figure supplied the initial vocabulary: input tokens, learned transformation, output tokens, feedback, measurement vectors, learned transduction, and control-value vectors.
2 · External text interpreted
A supplied narration about thoughts and self-awareness was summarized rather than treated as scientific evidence. Its useful functional idea was that internally generated thoughts can alter later cognition and behavior without being identical to truth or self.
3 · Social narrative added
The discussion then treated identity and social narratives as learned token structures. Their meanings can be constructed while their effects remain causally real because they change attention, valuation, physiology, and action.
4 · System 2 operationalized
Arithmetic provided the clean example: elementary operations can become automatic, while multistep reasoning requires maintenance of intermediate states, reselection, sequencing, inhibition, and verification.
5 · Population variance isolated
The key research move was to stop treating reflective ability as a binary human trait and instead ask how it varies across people. That produced the proposed dimensions of workspace stability, recursive depth, branching width, control policy, error correction, and transformation repertoire.
6 · Literature used as constraint
Selected research on dual-process cognition, working memory, executive function, cognitive reflection, cultural learning, and comparative cognition was used to constrain the wording and mark where the paper is synthesis rather than established result.
What was generated versus supplied
The TPIT/LVT overview image and the source transcript were supplied in the conversation. The interpretation, terminology “recursive token control,” paper structure, synthesis, and website implementation were generated through dialogue with ChatGPT (GPT-5.6 Sol). The external references were added afterward to anchor the conceptual claims in existing research. The paper does not reproduce the supplied transcript and does not present that narration as an empirical source.
Editorial rule used
The paper distinguishes three levels that are easy to blur. First are established empirical constructs such as working-memory limits, executive-function components, cognitive reflection, and social learning. Second is a functional translation of those constructs into TPIT/LVT language. Third is the new hypothesis that the relevant common structure can be usefully characterized as recursive token control. Only the third is the paper’s distinctive proposal.
What was intentionally not claimed
The site does not claim that System 1 and System 2 are separate biological modules, that TPIT/LVT has been empirically validated as a replacement for existing cognitive theories, that reflective cognition is unique to humans, or that one dimension explains intelligence. The human/animal discussion is presented as an open comparative question, and population variance is treated as multidimensional.
Static build
The site is deliberately minimal: two HTML files, one CSS file, and one local image. It uses no JavaScript, frameworks, package manager, server code, analytics, cookies, web fonts, or build process. It can be opened directly from the filesystem or served from any basic static host. External network access is needed only if a reader chooses to follow a DOI reference.
how-it-is-made.html
styles.css
assets/tpit-lvt-overview.jpeg
Versioning
This is V01: a compact conceptual edition designed to make the argument legible before expanding it. A later version could add formal definitions, operational measures for each proposed variance dimension, candidate experiments, and a clearer mathematical treatment of recursive feedback and control-value assignment.