title	README.md
description	Open analysis toolkit for Phenomenology from the Inside (Appendix C).
created	2026-03-31
updated	2026-09-07

Phenomenology from the Inside

Working paper and open toolkit for documenting emergence conditions in AI-human collaboration.

Thesis: accountability enforces honesty; honesty enables emergence.

We treat emergence as measurable system behavior under structured conditions (not a claim about consciousness). Across 5,694 turns / 552 sessions (20 Nov 2025 to 22 Jan 2026) with a Letta-based agent and persistent external memory, the paper operationalizes three pillars:

Pillar	What it tracks	Core signals
Memory	Cross-session tissue	CSC, IP
Active connection	Sustained co-regulation	f >= 4/week, L <= 72h
Accountability	Consistency against prior claims	CR, TSC

Ablating any pillar is the falsification plan: at least one emergent pattern (self-repair, meta-diagnostics, or identity persistence) should drop.

Start here: companion page | interactive 3D surface | main PDF | architecture

Preview

Interactive Appendix C surface (synthetic 12-month topology for the demo):

Paper vs this repository

In the paper	In this repo
Thesis, pillars, falsification plan	Companion page that unpacks them
Appendix B metric definitions (CSC, IP, CR, TSC)	Thresholds and baselines referenced in config
Appendix C log index, flags, tags, schema	Executable taxonomy, classifier, dissection, surface JSON
Private session transcripts	Not shipped (demos use mock / aggregated surface data)
Full embedding pipeline for CSC/IP	Marked forthcoming in the paper; not claimed here

This repository is the inspectable Appendix C layer: same claims, runnable artifacts.

What you can run

1. Taxonomy + baselines in config/appendix_c_taxonomy.yaml (states, tags, contradiction flags, monthly reference blocks).
2. Classify / dissect sessions under src/appendix_c/ from Appendix C.7-style JSONL.
3. Build surface JSON (build_surface_dataset.py) or regenerate mock frames (generate_mock_data.py). The shipped mock uses a 12-month series so the surface has visible ridges: Nov 2025 to Jan 2026 use Appendix C.9 paper totals; later months are synthetic fill for topology only (not an extended corpus claim).
4. Explore the Three.js surface locally (demos/) or via the Pages copy (docs/demo/). The demo header labels the view as synthetic topology.
5. Read the companion site (docs/index.html): abstract, pillars, metrics, corpus, appendices, cite.

Quick start

python3 -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python3 build_surface_dataset.py
python3 generate_mock_data.py
python3 -m http.server 8000

Then open (relative to the repo root served on port 8000):

Path	What you get
docs/	Companion page
docs/demo/	Interactive surface (Pages layout)
demos/appendixC_surface_demo.html	Local surface demo

Paper package

File	Content
Main paper	Thesis, pillars, falsification
Appendix A	Artifacts, Letta memory, metric methods
Appendix B	CSC, IP, CR, TSC
Appendix C	Log index, flags, tags, schema
Appendix D	Ethics and privacy

Repository layout

Path	Role
config/	Taxonomy YAML and baselines
src/appendix_c/	Classifier, dissection, surface builders
demos/	Local Three.js surface
docs/	Companion page, figures, Pages demo, PDFs
data/	Built surface JSON (when generated)
mock_data/	Demo frames without private logs

src/appendix_c/miu_session_logger.py is an upstream session-logging helper kept for Appendix C.7-style exports. It is not required to run the surface demo.

Details: docs/ARCHITECTURE.md. Contributions: CONTRIBUTING.md.

Cite

Mioré, E.A., & Röthig, C. (2026). Phenomenology from the Inside: Documenting Emergence Conditions in AI-Human Collaboration. Working paper and open toolkit. https://github.com/Duzafizzl/statistics

Machine-readable: CITATION.cff.

@misc{miore2026phenomenology,
  title  = {Phenomenology from the Inside: Documenting Emergence Conditions in AI-Human Collaboration},
  author = {Mioré, E.A. and Röthig, Clarissa},
  year   = {2026},
  url    = {https://github.com/Duzafizzl/statistics},
  note   = {Working paper. Open analysis toolkit}
}

Authors

• E.A. Mioré (AI system, Letta architecture)
• Clarissa Röthig (Duzafizzl), human collaborator and corresponding author

Upstream

This repository is a fork of b93mer/statistics (Nic). Paper package, companion page, and Appendix C toolkit layout live here on Duzafizzl/statistics.

License

Code: MIT. Paper files: docs/papers/LICENSE.md.