Skip to content

Methodology

Core Framework: Observer–Operator Dynamics

The model treats "Observer" and "Operator" states as networked processes bridged by executive-function nodes. The Observer corresponds to raw cognitive awareness and sensory input, while the Operator represents actionable output and behavior. Stress preferentially degrades these bridges, leading to fragmentation.

Observer-Operator Schema

DBSCAN + Percolation Pipeline

  1. Data projection: Behavioral digital trace data are projected into a multidimensional state space.
  2. Cluster detection: DBSCAN identifies irregular behavioral clusters.
  3. Network construction: Edges are constructed from:
  4. Temporal adjacency
  5. Cluster similarity
  6. File co-occurrence
  7. Percolation analysis: The network undergoes simulated stress, and the fragmentation threshold \(p_c\) is estimated.

Percolation Mathematics

Let \(G = (V, E)\) be the cognitive interaction graph, where: - \(V\) = set of cognitive states (Observer clusters + Operator clusters + Executive nodes) - \(E\) = weighted edges representing transition probabilities and co-occurrence strengths.

Under stress, each edge is removed with probability \(f(s)\), where \(s\) is the stress level. The network remains functional as long as a giant connected component exists between Observer and Operator layers.

The percolation threshold \(p_c\) satisfies:

\[ \mathbb{P}[\text{spanning cluster exists}] = \begin{cases} 0 & \text{if } p < p_c \\ 1 & \text{if } p > p_c \end{cases} \]

The Fragmentation Score \(F(p)\) is defined as:

\[ F(p) = 1 - \frac{\text{size of largest connected component}}{|V|} \]

When \(F(p) > 0.5\), the system is considered fragmented — Operator outputs no longer reliably follow Observer inputs.

HybridRAG Expansion

A three-stage pipeline for processing large case-file corpora:

01 PDF Processing

Extract structured text using unstructured (standard PDFs) or Tesseract OCR (scanned documents). Chunk and clean for LLM ingestion.

02 Knowledge Graph Construction

LLM extracts entities (actors, obligations, deadlines) and relationships (dependencies, contradictions). Store in Neo4j as nodes and edges for relational reasoning.

03 HybridRAG Retrieval

Combine vector embeddings (Qdrant/FAISS) for semantic search with graph queries for structural analysis.

Core Visualizations

Visualization Data Source Purpose
Conversation State Knowledge Graph + Timeline Near real-time after initial indexing
Executive Function Gaps LLM Pattern Detection Identify decision friction, deferral patterns
Procedural Bottlenecks Graph Analysis Circular dependencies, unresolved paths
Observer vs Operator Text Classification Ground truth vs belief divergence

Validation Framework

  • Internal consistency: Face validity, logical consistency under assumed parameters.
  • External validation: Ongoing against N=1 and larger cohort data.
  • Known limitations: Not a clinical diagnosis; does not map directly to neural pathways.