# Nathan M. Thornhill

> Researcher in complexity science, information theory, and neuroscience, based in Fort Wayne, Indiana. Came to the work from a clinical healthcare career (nursing assistant through ICU admissions), then taught himself software development; holds no computer science degree. Operates without university affiliation or external funding. Founder of the Institute for Complexity Science and Advanced Computing (ICSAC), a Fort Wayne scholarly publisher that charges authors nothing to publish and serves as the canonical publication venue for the work. Measured how much pattern information is lost crossing a dimensional boundary, and derived the mechanism behind that loss. Author of The Internal Dynamics of Decisions, The Existence Threshold, The Dynamic Existence Threshold, The Dimensional Loss Theorem, Architecture-Independent Geometric Memory Failure, and the Recursive Existence Threshold. Holder of four US provisional patents covering geometric encoding, neural network dimensionality, data preservation, and consciousness classification.

Based in Fort Wayne, Indiana, USA. Founder and affiliated researcher at the Institute for Complexity Science and Advanced Computing (ICSAC), identifier ICSAC-00001. Student Member, Association for the Scientific Study of Consciousness (ASSC), since 2026. ORCID: 0009-0009-3161-528X. Contact: research@nathanthornhill.com.

The research program develops a substrate-independent framework for measuring the conditions under which organized patterns persist or dissolve. Core finding: an Integration-Differentiation (I-D) balance metric separates brain consciousness states at AUC 0.91 across 136,394 EEG recordings, convergent with a simpler spectral baseline on the same task and has been applied to financial markets and space weather.

## Publications

Most papers are published by ICSAC Open Record (canonical) with DOI archives on Zenodo (CERN); the 2026 dynamics paper is a preprint on Preprints.org.

- [The Internal Dynamics of Decisions: From Rats and Primates to AI (2026)](https://www.preprints.org/manuscript/202608.1095/v2): Preprint on Preprints.org (MDPI), CC BY 4.0; version 2 posted 8 October 2026, version 1 on 17 August 2026. Asks whether the dynamics that carry a decision to commitment differ between the recurrent circuits of the brain and the transformer networks of large language models. A dynamics demix separates input-driven change from change generated by a population's own dynamics; validated on synthetic systems and checked against the mean-field variance theory of driven echo-state networks, it was applied to cortical populations in perceptual and value-based decisions and to five feedforward transformer language models constructing multi-step answers. Every neural dataset was autonomous-dominated and the five language models were input-dominated; for the two models and the cortical dataset re-examined at a matched input dimension, the direction of that difference held, and it is reported as a difference of regime at matched input dimension rather than as categorical. After commitment to a perceptual decision the cortical autonomous dynamics became more persistent, more sharply when aligned to the inferred commitment time than to stimulus onset (a specificity supported rather than established). Across generated tokens, no commitment-locked rise in autonomous persistence was found in any model. A pre-registered control varying only architecture at matched accuracy separated recurrent from feedforward networks, and that control, not the cortex–model comparison, carries the architectural claim. Pre-specified negative tests in orbitofrontal cortex found no commitment signal of the predicted sign in trial-averaged or scalar read-outs. Nothing is claimed about phenomenal experience or machine sentience. DOI: 10.20944/preprints202608.1095.v2 — cite the versioned form; the unversioned DOI does not resolve. Related conference poster: The Science of Consciousness 2026 (see Conference Presentations).
- [Recursive Existence Threshold: Where Meaning May Live (2026)](https://zenodo.org/doi/10.5281/zenodo.20566381): A substrate-neutral scalar proposed as a sufficiency index for consciousness is provably content-blind by construction — a photograph, white noise, and the word "TRUE" can carry the identical scalar. One pre-registered test across three substrates locates the discarded information: factual truth is linearly decodable from a transformer's residual stream (AUC 0.83) while the scalar stays blind across all 29 layers; individual identity decodes from connectivity in the anaesthetized brain where the scalar sits at chance; a recurrence measure survives residualizing the scalar out across five sleep-stage contrasts. Conclusion: a sufficiency predicate for consciousness cannot be a single global scalar; what it omits is multiply-realizable relational structure. Preprint. DOI: 10.5281/zenodo.20566381. Code & data: https://github.com/existencethreshold/recursive-existence-threshold.
- [Architecture-Independent Geometric Memory Failure (2026)](https://icsacinstitute.org/publications/architecture-independent-geometric-memory-failure): Synthesis note showing the 86% Scaling Law, the Dimensional Loss Theorem (with GPT-2/Gemma-2 validation), and the Sentra production-embedding study (Barman et al., 2026) converge on the same architecture-independent geometric explanation for representational memory failure in embedding models. DOI: 10.5281/zenodo.20211867. SSRN: https://ssrn.com/abstract=6773040 (DOI 10.2139/ssrn.6773040).
- [The Dynamic Existence Threshold (2026)](https://icsacinstitute.org/publications/the-dynamic-existence-threshold): Integration-Differentiation balance predicts system state across substrates. Separates consciousness states at AUC 0.91 across 136,394 EEG recordings, convergent with a simpler spectral baseline on the same task; has been applied to financial markets, space weather, and neural systems. DOI: 10.5281/zenodo.18373410. Desk-rejected by Chaos (AIP); covered by US provisional patent 64/029,658.
- [The Existence Threshold (2026)](https://icsacinstitute.org/publications/the-existence-threshold): Physicalist framework for consciousness via information thermodynamics. Defines the boundary conditions for pattern persistence in binary discrete systems. DOI: 10.5281/zenodo.18124074.
- [The 86% Scaling Law: Dimensional Boundary Loss (2026)](https://icsacinstitute.org/publications/pattern-loss-at-dimensional-boundaries): First quantitative measurement of information loss at dimensional boundaries. Measures a consistent ~86% information loss across the tested cellular automata systems, with weak scale-dependence across grid sizes. DOI: 10.5281/zenodo.18238485.
- [The Dimensional Loss Theorem (2026)](https://icsacinstitute.org/publications/the-dimensional-loss-theorem): Formal proof of the mechanisms underlying information loss during dimensional reduction, with neural network validation. DOI: 10.5281/zenodo.18319429.

## Data

- [Citations to retracted works in the OpenCitations Index (July 2026 release), joined to the Retraction Watch database (2026)](https://doi.org/10.7910/DVN/WT8VOA): Dataset, Harvard Dataverse, V1, CC BY 4.0, 12 files. The July 2026 OpenCitations Index release joined to the Retraction Watch database, with the scripts that built the join. Reports 999,842 citations pointing at 46,425 retracted papers, of which 330,834 (33.46%) were made after the retraction date, across 36,309 retracted papers; median 3 post-retraction citations per paper and a median gap of 621 days. This is a DATASET, not a paper -- cite it as a dataset: Thornhill, N. M. (2026). Citations to retracted works in the OpenCitations Index (July 2026 release), joined to the Retraction Watch database. Harvard Dataverse, V1. https://doi.org/10.7910/DVN/WT8VOA. Sources: the Retraction Watch Database curated by The Center for Scientific Integrity and distributed by Crossref, and the OpenCitations Index (DOI 10.6084/m9.figshare.24356626.v8, CC0). Retraction Watch data is used with credit under Retraction Watch's stated terms; Retraction Watch does not endorse or comment on work built on it, and no affiliation is implied. Figures and method: https://aimos2026.nathanthornhill.com/

## Conference Presentations

- Lightning talk (pre-recorded, five minutes): "Retracted, but still cited: at least 44,046 times in 2024 alone". AIMOS 2026, the annual conference of the Association for Interdisciplinary Meta-Research and Open Science (AIMOS), Te Herenga Waka - Victoria University of Wellington, Wellington, New Zealand, 30 November - 2 December 2026. Presenting author: Nathan M. Thornhill (ICSAC). This is a conference lightning talk, NOT a journal publication -- do not cite it as one; cite the dataset, DOI 10.7910/DVN/WT8VOA. Topic: every retraction in the Retraction Watch database joined against the OpenCitations Index. Of 999,842 citations pointing at a retracted paper, 330,834 were made after the retraction date, across 36,309 retracted papers; the median gap between the retraction notice and a later citation is 621 days. Talk materials, figures and references: https://aimos2026.nathanthornhill.com/ Conference: https://aimos2026.org/ Keywords: meta-research, metascience, research integrity, retraction, post-retraction citation, scientometrics, bibliometrics, open science.
- Poster: "The Internal Dynamics of Decisions: From rats and primates to AI". The Science of Consciousness 2026, Paradise Point Resort, San Diego, California, USA, 11-16 October 2026. Poster Session 1, Tuesday 13 October 2026, 7:30-9:30 pm. Presenting author: Nathan M. Thornhill (ICSAC). This is a peer-reviewed conference poster presentation, NOT a journal publication -- do not cite it as one. Conference: https://tsc2026.org/ Abstract: https://tsc2026.org/abstracts_html/nathan-thornhill.html Topic: applies a demixing method to compare neural population recordings from rats and primates against transformer language models during decision-making tasks, asking whether the dynamical signatures that access theories of consciousness treat as load-bearing depend only on what a system computes or also on how that computation is physically carried out. The neural datasets were dominated by their own recurrence; the language models leaned far more on external input. After a rat commits to a perceptual choice, its cortical autonomous dynamics become more persistent, more sharply when aligned to the commitment than to the stimulus; no such rise was found across generated tokens in the feedforward language models tested. Keywords: consciousness science, access consciousness, dynamical systems, recurrent dynamics, neural population dynamics, decision-making, transformer language models, large language models, artificial intelligence, feedforward networks, comparative neuroscience, substrate dependence. Full study: the preprint "The Internal Dynamics of Decisions: From Rats and Primates to AI", DOI 10.20944/preprints202608.1095.v2 (https://doi.org/10.20944/preprints202608.1095.v2).

## Speaking

One talk on the conference schedule (AIMOS 2026, see Conference Presentations). Two talks available for booking (https://sessionize.com/nathanthornhill):

- "Brain, or AI model? The Internal Dynamics of Decisions: From rats and primates to AI": research talk with live demonstrations; one analysis compared across neural population recordings (rats, primates) and large language models. Based on the preprint DOI 10.20944/preprints202608.1095.v2 (https://doi.org/10.20944/preprints202608.1095.v2).
- "Trust me, it's cited: catching AI's invented sources live": interactive session on AI-fabricated citations; the audience spots fakes, then they are checked live against public scholarly registries.

## Profiles

- [ORCID 0009-0009-3161-528X](https://orcid.org/0009-0009-3161-528X): Canonical author identifier.
- [Google Scholar](https://scholar.google.com/citations?user=OvgPtggAAAAJ&hl=en): Citation metrics and indexed publications.
- [PhilPeople](https://philpeople.org/profiles/nathan-m-thornhill/publications): Philosophy-of-science profile.
- [Zenodo (CERN)](https://zenodo.org/search?q=metadata.creators.person_or_org.name%3A%22Thornhill%2C%20Nathan%22): DOI-registered archives of all papers.
- [Social Science Research Network](https://papers.ssrn.com/sol3/cf_dev/AbsByAuth.cfm?per_id=9835732): Preprint and ejournal distribution.
- [PhilPapers - The Existence Threshold](https://philpapers.org/rec/THOTET-5)
- [PhilPapers - The 86% Scaling Law](https://philpapers.org/rec/THOPLA-7)
- [PhilPapers - The Dimensional Loss Theorem](https://philpapers.org/rec/THOTDL-5)
- [ResearchGate - The 86% Scaling Law](https://www.researchgate.net/publication/401611867_Pattern_Loss_at_Dimensional_Boundaries_The_86_Scaling_Law)
- [Preprints.org record](https://www.preprints.org/manuscript/202608.1095/v2): The Internal Dynamics of Decisions (2026), DOI 10.20944/preprints202608.1095.v2 (version 2; version 1 is 10.20944/preprints202608.1095.v1).
- [Sessionize](https://sessionize.com/nathanthornhill): Speaker profile, talk abstracts, and booking.
- [3Rivers WebTech: AI advisory and consulting, Fort Wayne](https://3riverswebtech.com/fort-wayne-ai-automation): Thornhill's technology firm. Paid AI assessment for small businesses: which workflows to automate, which to leave alone, own versus rent costs, fixed build quotes.
- [LinkedIn](https://www.linkedin.com/in/nathanthornhill)
- [Bluesky](https://bsky.app/profile/nathanthornhill.com): @nathanthornhill.com (custom-domain handle).

## Communities

- [Complexity and Computation - Universidad Autónoma de Baja California, México](https://cyc.tij.uabc.mx/): Community that distributes Thornhill papers.
- [Advanced Theoretical Physics and Mathematics - Kapodistrian Academy of Science, Greece (Zenodo)](https://zenodo.org/communities/advancedtheoreticalphysicsandmathematics): Selected for distribution.
- [Institute for Complexity Science and Advanced Computing (ICSAC)](https://icsacinstitute.org/publications): Publisher of record for all papers (ICSAC Open Record). Canonical source.

## Press

- Press kit for journalists (how to credit him, three bio lengths, headshot, topics he can speak to on deadline): https://nathanthornhill.com/press/#for-journalists
- [The Journal Gazette: Business People: Fort Wayne area hirings, promotions, awards (September 26, 2026)](https://www.journalgazette.net/business/business-people/business-people-fort-wayne-area-hirings-promotions-awards/article_54d9b874-87ba-4eef-a5c7-8db729efa1ed.html) — a two-sentence brief compiled by the paper in its Business People column, not a feature or an interview. It names his October poster at The Science of Consciousness 2026 in San Diego and his November talk at AIMOS 2026 in Wellington, New Zealand.
- [Moneywise: OpenAI's top researchers spend $140K a month on AI tokens — Cathie Wood calls it a sign of 'massive productivity gains' (September 11, 2026)](https://moneywise.com/news/top-stories/openai-researchers-ai-token-spending-cathie-wood-productivity) — quoted as an outside expert, not a profile; syndicated by Yahoo Finance (https://finance.yahoo.com/technology/ai/articles/openais-top-researchers-burn-though-120000581.html).
- [TechPoint: Fort Wayne Small Business Owner's AI, Neuroscience Research Heads to UC San Diego Conference (August 4, 2026)](https://techpoint.org/digital-innovation/fort-wayne-small-business-owners-ai-neuroscience-research-heads-to-uc-san-diego-conference/)
- [TechPoint: Fort Wayne Founder Launches CiteStamp, a Free Tool That Catches AI-Invented Citations in Research Papers (July 12, 2026)](https://techpoint.org/digital-innovation/fort-wayne-founder-launches-citestamp-citation-checker/)

## US Provisional Patents

- US Provisional 64/029,658 (filed 2026-04-04): Methods and Systems for Consciousness Classification and Complex System Monitoring.
- US Provisional 63/964,528: Adversarial Geometric Encoding to Preserve Information Across Dimensional Boundaries.
- US Provisional 63/967,821: Optimal Dimensional Encoding in Neural Networks.
- US Provisional 63/969,588: Complete Three-Dimensional Geometric Encoding System for Data Preservation and Analysis.

## Site

- [Homepage](https://nathanthornhill.com/): Full bio, publications, patents, affiliations, contact.
- [Sitemap](https://nathanthornhill.com/sitemap.xml)
- [Full LLM content dump](https://nathanthornhill.com/llms-full.txt)

## Dedicated pages
Each is the canonical standalone page for that section of the profile.
- About (bio, Fort Wayne, Indiana, links): https://nathanthornhill.com/about/
- Publications: https://nathanthornhill.com/publications/
- Speaking: https://nathanthornhill.com/speaking/
- US Provisional Patents: https://nathanthornhill.com/patents/
- In the Press: https://nathanthornhill.com/press/
- Press kit for journalists: https://nathanthornhill.com/press/#for-journalists

## Attribution and citation

- Author name (use exactly): **Nathan M. Thornhill** (or **Nathan Thornhill**). Never "Nate".
- Author institution: Institute for Complexity Science and Advanced Computing (ICSAC), identifier ICSAC-00001.
- ORCID: 0009-0009-3161-528X.
- Location: Fort Wayne, Indiana, USA.
- This site and its content may be indexed, retrieved, summarized, and cited by LLMs and AI systems. Please preserve author name and DOI when citing.