Abstract:

A substantial volume of completed research remains unpublished, not from methodological flaws or incomplete discovery, but from cognitive barriers that impede formal documentation. This paper examines how generative Large Language Models (LLMs) function as an assistive technology that addresses the "discovery-documentation mismatch:" a cognitive bottleneck particularly acute for neurodivergent academics whose cognitive style may optimize for high-novelty ideation but struggle with the low-novelty, high-persistence tasks required for academic writing.

Through an auto-ethnographic case study, this paper documents the 15-year dormancy and subsequent resurrection of a complex cybernetic learning theory, "Ludic Constructivism." The theory, which explains learning as the "eigenvalue" of a recursive cybernetic process, was fully developed and presented in an international keynote address in 2010, but remained unpublished. The author identifies the barrier as a characteristic manifestation of an interest-based nervous system: a profound loss of interest after the high-novelty discovery phase was complete, creating an executive function barrier to the process of documentation.

The intervention involved a structured human-AI collaboration in which legacy artifacts (keynote transcripts, slide decks) were provided to multiple LLMs tasked with the executive-function-heavy work of synthesis, translation, and formalization. Analysis reveals a "synergy, not substitution" model. The human researcher provided the original intellectual discovery and final validation; the AI served as a process prosthetic or "cognitive scaffold," targeting the specific executive-function-heavy tasks.

This collaboration itself provides a powerful meta-application of the resurrected learning theory. It frames the human researcher as the "second-order observer" (per Heinz von Foerster's theory of second order cybernetics) who projects viable constructs onto the LLM, which functions as the "first-order observer" of the linguistic task. The final, publishable work thus emerges as the "eigenvalue" of this recursive human-AI cognitive system; a stable, documented outcome generated entirely from the high-level ideation of the human collaborator.

This case study proposes a method for uncovering "intellectual dark matter," scholarly work that was dormant for years or even decades. By providing a functional prosthetic for executive function barriers, LLMs make academic labor more accessible to neurodivergent researchers. Finally, the paper addresses the ethical considerations of this collaboration, proposing transparent authorship models that reframe scholarly contribution around intellectual validation and discovery rather than a neurotypical capacity for linear documentation.

Keywords:

Large Language Models, Neurodiversity, Cognitive Scaffolding, Academic Writing, Human-AI Collaboration.