Neuro-Entropic Extraction Simulator

Disclaimer

This repository contains the raw research code for the manuscript "Quantifying Cognitive Collapse". The scripts were developed for academic simulation and data analysis purposes. As an initial formulation, the code is optimized for mathematical experimentation rather than production efficiency.

Project Description

This project simulates the cognitive dynamics of human attention under algorithmic influence using a multi-layered mathematical model. We combine stochastic differential equations (SDEs), stability analysis, and empirical data validation to quantify how recommendation algorithms extract cognitive bandwidth and reshape neural pathways.

Key Findings

1. Entropy Gap: Social media text (Twitter) exhibits higher unigram entropy than classical literature when measured at fixed word-length windows, suggesting algorithmic media creates a "chaotic" rather than "structured" cognitive environment.

2. Optimal Exploitation Threshold: Parameter sensitivity analysis reveals that algorithms have an incentive to maintain users in a "semi-stupid" state—maximizing capital extraction without triggering complete cognitive collapse.

3. Intervention Efficacy: Simulated interventions (noise reduction + structured resonance) can reverse cognitive collapse trajectories, demonstrating the potential for engineered countermeasures.

Visual Results

Stochastic Paths (SDE Simulation)

Multiple Monte Carlo paths showing cognitive state trajectories under noise. Each path represents an individual's cognitive bandwidth evolution.

Entropy Distribution Comparison

Shannon entropy distributions for Twitter vs. Literature at fixed 40-word windows. The counter-intuitive result (Twitter > Literature) suggests unigram entropy alone is insufficient to capture structural complexity.

Potential Landscape

Potential energy landscape V(B) under different algorithm intensities (α). As α increases, the stable "well" disappears, indicating system collapse.

Lyapunov Exponent

Lyapunov exponent λ vs. algorithm intensity α. The crossing point (λ ≈ 0) marks the cognitive collapse threshold.

Zipf's Law Analysis

Word frequency distributions follow Zipf's law for both corpora, suggesting structural differences manifest at semantic/syntactic levels rather than word frequency.

Intervention Simulation

Demonstration of cognitive recovery through engineered intervention (noise reduction + periodic driving force) at t=40.

File Structure

Core Simulation

• scipy.integrate.odeint.py: Main ODE solver implementing the neuro-entropic model with adaptive algorithm policy
• stochastic_simulation.py: SDE solver using Euler-Maruyama method with hysteresis effects
• stability_analysis.py: Potential landscape and Lyapunov exponent calculations

Data Analysis

• data_calibration_v3.py: Entropy comparison between Twitter and Literature using Project Gutenberg texts
• structural_analysis.py: Zipf's law analysis and high-frequency word comparison

Intervention Studies

• intervention_simulation.py: Simulates the effect of noise reduction and structured resonance on cognitive recovery

Utilities

• generate_all_figures.py: Batch script to generate all publication-quality figures

Dependencies

pip install numpy scipy matplotlib seaborn pandas nltk

Quick Start

1. Run baseline simulation:

   python scipy.integrate.odeint.py

2. Generate all figures:

   python generate_all_figures.py

3. Run entropy analysis:

   python data_calibration_v3.py

Research Context

This work bridges computational neuroscience, information theory, and critical algorithm studies. The model formalizes the hypothesis that recommendation algorithms create a "neuro-entropic extraction" process, where user attention is systematically converted into digital capital while degrading cognitive bandwidth.

Citation

If you use this code in your research, please cite:

Neuro-Entropic Extraction Simulator v0.1
Quantifying Cognitive Collapse: A Mathematical Model of Algorithmic Attention Extraction

License

Research code—use at your own discretion. This is academic exploration code, not production software.