epistemic research

KG Emergence

A knowledge-graph system that ingests contested claims as noisy observations, then discovers whether recurring structural patterns ("graphlets") emerge across independent traditions and eras — and uses them to predict missing evidence.

Qdrant + Neo4jAnti-overfit by designSource-as-observationGraphlet recurrenceLatent-edge prediction

Status: v0 scaffold. The schema, ontology, experiment protocol, and a candidate-graphlet generator are specified; the Qdrant/Neo4j wiring and the benchmark corpus are not yet built. These pages are the design of record.

The governing idea: treat every source — scripture, court record, intelligence leak, occult tract — as a noisy observation, never as truth. Decompose claims, keep confidence and salience separate, and let initially-dismissed hypotheses gain credibility only through independent recurrence. Structure that survives that gauntlet is signal; structure that doesn't is discarded.

source repository →

Cross-tradition motif overlay
Interactive Visualization
Live D3 force graph of the extracted traditions, laid out on the abstract→concrete axis. Each colored cluster is one tradition's role-graph; glowing white hubs are structural motifs that recur across independent traditions — the emergence signal.
explore the graph →
Graphlet recurrence matrix
Interactive Visualization
Traditions × named graphlet templates (whispering adversary, manufactured criminality, legitimacy laundering…). A heatmap of which structural motifs recur in which traditions — the design's "View C". The adversarial-institution motifs stay dark even with a modern conspiracy family added — an honest negative result.
open the matrix →
In/out-of-sample trend
Quantitative Research
The "am I overfitting / is there any signal?" curve: in-sample vs out-of-sample lift over the null across every recorded run. Right now lift hovers near 1 (chance) — the honest state of the evidence.
see the backtest →
Confidence / salience quadrant
Interactive Visualization
Every extracted claim placed on confidence × salience (core facts / watchlist / context / archive). Makes the "keep-and-watch the wild claims" principle visible — and lets you flag bad extractions and export an annotation patch.
explore the claims →
Topics & Roadmap
Map
What the seed corpus covers (institutional capture, controlled opposition, PSYWAR/occult, and the method), the abstract→concrete axis, and what the structure says we still need to add — modern conspiracy, alt-chronology, the reflexive sovcit layer, embeddings, scale.
read →
Universal Epistemic Ontology
Exploratory Framework
The neutral schema: node and edge types, the source-as-observation model, Cromwell-rule probabilities, bridge-quality taxonomy, and the anti-overfit rules that stop symbolic similarity from masquerading as causation.
read →
Graphlet Generator
Algorithm Design
How candidate structural motifs are found without knowing the question in advance: edge-dominance scoring, ego-neighborhood extraction, canonicalization into role families, recurrence scoring, and latent missing-edge prediction.
read →
Qdrant + Neo4j Architecture
System Architecture
Division of labor: Qdrant proposes semantic candidates (dense + sparse), Neo4j decides typed meaning and topology. Collections, hybrid retrieval, and six exploration views.
read →
Experiment Protocol
Falsification Protocol
A trading-grade validation harness: train/validation/test/holdout splits, six null models, and metrics for graphlet recurrence and out-of-sample latent-edge prediction — so emergence has to beat chance.
read →
Benchmark Corpus Plan
Data Plan
Building the corpus across eight data universes and six tradition buckets while preventing source-family leakage, so cross-cultural recurrence reflects real convergence rather than copying.
read →
Origin Archive
Provenance
The visible reasoning trail that produced the schema and architecture, including the distinctions the system is built to preserve (a tradition is not its institutions; a concept is not its nearest analogue).
read →
Generated by scripts/build_site.py from docs/*.md. Single source of truth: the Markdown. Do not hand-edit this HTML.