# Fresh external search notes — 2026-08-16

Fresh search was performed immediately before drafting EML-RKD-05.

## Primary sources

1. Samuel Recht, Ljubica Jovanovic, Pascal Mamassian & Tarryn Balsdon (2022), *Confidence at the limits of human nested cognition*, Neuroscience of Consciousness, niac014.
   - Human nested judgements were above chance through fourth order in a specific perceptual-confidence paradigm.
   - The authors' model includes recursive evidence degradation and multiple noise sources.
   - This is an empirical precedent for attenuation in one human realization, not a universal recursion law.

2. Yury Polyanskiy & Yihong Wu (2015), *Strong data-processing inequalities for channels and Bayesian networks*, arXiv:1508.06025.
   - Reviews/derives channel-dependent contractions stronger than ordinary data-processing.
   - Used as a mathematical precedent for conditional recursive contraction when a genuine Markov/post-processing structure is present.

3. Maxim Raginsky (2014), *Strong data processing inequalities and Phi-Sobolev inequalities for discrete channels*, arXiv:1411.3575.
   - Systematic treatment of optimal SDPI constants for discrete channels.
   - Used as additional primary mathematical background for contraction coefficients.

4. C. Nicolò De Sabbata, Theodore R. Sumers & Thomas L. Griffiths (2024), *Rational Metareasoning for Large Language Models*, arXiv:2410.05563.
   - Models value-of-computation by trading reasoning performance against inference cost.
   - Used as an engineering precedent that additional reasoning/meta-reasoning has a cost dimension.

5. Michael T. Cox et al. (2022), *Computational Metacognition*, arXiv:2201.12885.
   - Used only as an artificial-metacognition architecture precedent.

## Search conclusion
External work supports conditional contraction/noise/cost realizations. It does not establish any universal monotone decay law across arbitrary recursive-knowing systems. The arbitrary-profile theorem in EML-RKD-05 is an internal algebraic construction.