With the empirical core complete, the paper’s theoretical architecture was finalized. The structure runs eight sections: Introduction → Related Work → Theoretical Framework → Research Design → Results → Discussion → Conclusion → References (64 entries, six appendices).
Two theoretical moves were added in revision:
Peirce as methodological foundation. The project’s logic is abductive in the Peircean sense: frontier claims function as “surprising facts” that generate new hypotheses about the taxonomy. The question implicit in every hard case is: which refinement to the framework, if true, would make this case less surprising? This is Peirce’s retroduction — the only form of inference that introduces genuinely new ideas into a system. Popper, Lakatos, and Kuhn explain how hypotheses are tested; Peirce explains how they are generated. The paper needs both.
Building on Quattrociocchi et al. The paper explicitly grounds itself in two 2025 works whose diagnosis it shares and extends: Loru et al. (PNAS, “epistemia”) and Quattrociocchi, Capraro & Perc (arXiv, “epistemological fault lines”). Their contribution is foundational, not adjacent — epistemia is the condition this project takes as its starting point. Post-cognition is offered as a complementary, operative response: structured external intervention that reconstructs the ontological commitments implicit in LLM outputs. The terminological triangle is: metacognition (absent in LLMs, per Loru and Quattrociocchi) / epistemia (the condition they name) / post-cognition (the framework’s mode of operation, built on their diagnosis).
Document: 04_struttura_paper_v2.docx (v3 update). References: 64 entries in epistemologia_ai_bibliografia_definitiva.bib.