Environmental Governance Regulatory Enactment - Agent-Based Modeling as Counter-Archive
Permagame of Systemic Inequality · Sriya Thotakura · GSAPP CDP · Fall 2025
FERAL AGENT
South Bronx · Tactical Blackout View
Cross-Bronx Expressway

"This is South Bronx, New York City (Region specific prototype based on my research, this is most visible and vulnerable to inequality of land use). The red you're seeing is not traffic. It's not heat. It's a computational model of years of accumulated socio-economic inequality, the Feral Agent."

Historical Scar
Cross-Bronx Expressway construction
"Before the highway, the neighborhoods through which Moses would build the Cross Bronx were among the most racially integrated in the country, with large populations of Jewish immigrants from Eastern Europe, Irish and Italian Immigrants, (and after WWII) Puerto Ricans, and African-Americans. After the highway tore through, property values plummeted. White residents fled (lured to the suburbs by government-backed mortgages), while black and brown residents remained (oftentimes with few other options due to red-lining and racist housing policies). As covered in the documentary, Decade of Fire, subsequent disinvestment in services from the city led The Bronx to burn in the 70s and 80s, destroying roughly 80% of housing stock and displacing another 250,000."
Cross-Bronx Expressway construction (1950s, Robert Moses era) Source: Decade of Fire documentary
Problem

Traditional tools (GIS, census) can't show the algorithmic logic of how infrastructural violence propagates through time and space.

Computational Scar
  • Feral Agent concentrated along highway corridor
  • Inherited displacement and disinvestment
  • Violence that spreads, compounds, and becomes systemic
Narrative Precedents
Don't Flush Me project

Don't Flush Me

Data visualization of NYC's combined sewer overflow system

Feral Web project

Feral Web

Decentralized web infrastructure project

Emissary's Trilogy

Emissary's Trilogy

Narrative exploration of infrastructure and power

Vanessa Agard-Jones work

Vanessa Agard-Jones

Anthropological work on toxicity and environment

Feral Atlas
Feral Atlas visualization
Feral Atlas: A framework for analyzing human-environment interactions

FERAL AGENT

NOT A Human/Resident

The agent is not an individual person, household, or demographic group. It transcends individual human experience.

NOT Physical Infrastructure

Not a road, building, pipe, or tangible object. It exists in computational space as an abstraction.

IS A Digital Proxy

A computational representation of accumulated vulnerability that can be simulated, measured, and analyzed.

TECHNICAL FLOW

Input/Logic Layers

Core framework components and logic processing units.

Data Sources

Historical and real-time data inputs feeding the system.

Core Algorithms

Computational engines driving the analysis.

Processing Functions

Specialized functions for data transformation.

TECHNICAL PRECEDENTS

AGENT-BASED MODELING VISUALIZATIONS

ABM Prediction Model

Mathematical Prediction Model

Demonstrates the predictive capabilities of the ABM system, showing how agents evolve over time based on mathematical rules.

ABM Simulation Recording

Interactive Simulation

Shows real-time agent interactions and emergent patterns in the simulation environment.

Technical Precedents
Ternary Logic Visualization Technical Precedent 2

Ternary Logic System

Implements a three-state logic system for agent decision making, allowing for more nuanced behavior representation.

State Transitions

Visual representation of how agents transition between different behavioral states based on environmental factors.

Decision Matrix

Shows the decision-making framework that guides agent behavior in different scenarios.

LOADING DEMO...
Please wait while we load the interactive map...
System Shift: Reactive → Proactive
  • Precision

    Replaces guesswork with a calculated "Jurisdiction of Risk" for targeted zoning.

  • Prediction

    Tests policies via simulation before implementation to prove they work.

  • Enforcement

    Automates instant evidence for fines, removing human delay and bias.

  • Democratization

    Turns invisible pollution into visual proof for community leverage.

The Bottom Line

"The Map targets the risk.

The Simulation validates the solution.

The Alert enforces the law."

The Spatial Buffer
1 Code (The Logic)

// Simple boolean switch

const interventionActive = true; // Toggle effect

const getSensorReading = () => interventionActive ? 15 : 45; // Safe or Violation

2 Simulation (The Physics)

The Purple Box becomes a Damping Zone

  • Action: Multiplies particle speed by 0.05 (95% reduction)
  • Visual: Particles "freeze" upon hitting the wall
3 Narrative (The Science)

Represents a CLT Green Wall physically blocking and filtering PM2.5.

"The Buffer activates a Damping Function (β) that physically slows particle velocity, forcing the sensor reading down to a safe 15."

Permacomputing

This system aligns with Permacomputing by prioritizing radical efficiency and lifespan over excess.

1 Radical Minimalism (The Stack)

The Tech:

The simulation runs on a single, lightweight HTML file with vanilla JavaScript—not a heavy engine like Unity or Unreal.

The Benefit:

It requires negligible processing power, meaning it can run on 10-year-old laptops or low-end mobile devices, extending hardware lifespans.

2 Energy Resilience (The Hardware)

The Tech:

The "Forensic Node" logic is designed for low-power microcontrollers (Raspberry Pi/ESP32).

The Benefit:

It consumes minimal wattage, allowing the sensors to run indefinitely on small solar panels off the grid.

3 Functional Aesthetics (The Visuals)

The Tech:

It uses simple geometric primitives (dots, lines) instead of photorealistic rendering.

The Benefit:

It prioritizes clarity over computational cost, reducing the carbon footprint of the digital tool itself.

The Bottom Line

"The medium matches the message: A system designed to protect the environment is built on code that respects environmental limits."

Permacomputing in Practice
Permacomputing Visualization

This visualization demonstrates how permacomputing principles are applied in our system, showing the efficient use of resources and sustainable computing practices.

NOT A Human/Resident

The agent is not an individual person, household, or demographic group. It transcends individual human experience.

NOT Physical Infrastructure

Not a road, building, pipe, or tangible object. It exists in computational space as an abstraction.

IS A Digital Proxy

A computational representation of accumulated vulnerability that can be simulated, measured, and analyzed.

IS Algorithmic Proof

Its behavior demonstrates that structural violence propagates through calculable, systematic patterns—not random chance.

The Feral Agent exists at the intersection of computational modeling and environmental justice, providing a framework to understand how systemic inequalities become embedded in urban landscapes through infrastructure.

METHOD · COMPUTATIONAL FRAMEWORK
LAYER 1: BASE MAP
  └─ Leaflet.js + NYC GIS Data

LAYER 2: VULNERABILITY INDEX
│  └─ OpenFEMA + NCEI Climate Data                │
│                                                 │
│  LAYER 3: THE FERAL AGENT (ABM + KDE)           │
│  └─ Agent-Based Model → Kernel Density Est.     │
│                                                 │
│  LAYER 4: TACTICAL UI                           │
│  └─ Intervention Mechanics + Scorecard          │
└─────────────────────────────────────────────────┘
                
Feral Agent · Spread Logic (Pseudocode)
function propagateAgent(adjacentZone) {
  let spreadRate = 1.0;

  // Rule 1: Income-based acceleration
  if (adjacentZone.medianIncome < 40000) {
    spreadRate *= 1.8;
  }

  // Rule 2: Infrastructure as vector
  if (adjacentZone.hasHighwayBarrier) {
    spreadRate *= 2.2;
  }

  // Rule 3: Service desert multiplication
  if (adjacentZone.greenSpacePercent < 5) {
    spreadRate *= 1.5;
  }

  return spreadRate;
}
                        
What This Encodes
  • 80% faster spread in census tracts < \$40k median income
  • 120% faster across highway barriers as infrastructural vectors
  • 50% faster in green space deserts (< 5% coverage)
  • Rules derived from health, climate vulnerability, and displacement literature
02:30 – 03:30
ACT I · RECOGNITION
Initial Feral Agent cloud along Cross-Bronx Expressway
03:30 – 04:30
ACT II · ACCELERATION
Control Scenario · 10 Turns
  • Spreads along highway north to Westchester, south to Lower Bronx
  • Leaps quickly across low-income zones
  • Slows near wealthy enclaves like Riverdale
  • Growth curve: exponential, not linear
  • By Turn 25: 34% of Bronx tracts at “Critical Vulnerability”
Line graph: Feral Score vs. Turns (Control)
04:30 – 05:45
ACT III · INTERVENTION
Tactical Console
  • Community Land Trust (CLT)
  • Mutual Aid Network
  • Green Infrastructure
  • Policy Reform Marker
═════════════════════════════════════════
CLT DEPLOYED · MOTT HAVEN

3-block stabilization zone established
Feral spread rate −40% in radius

Cost: 2 intervention units
Remaining: 8 units
═════════════════════════════════════════
                        
User becomes Agent of Repair; CLT changes local ABM rules, modeling resident ownership as computational immunity.
05:45 – 07:30
ACT IV · FORENSIC ALERT
Control
Continued spread · +47% Feral Score by Turn 30
Intervention
Growth slowed near CLT · +23% Feral Score by Turn 30 · Declines by Turn 40
═══════════════════════════════════════════════════
FERAL ALERT · COMPARATIVE ANALYSIS

CONTROL TRAJECTORY:     Feral Score +47% by Turn 30
INTERVENTION TRAJECTORY: Feral Score +23% by Turn 30

ZONES STABILIZED: 8 census tracts (24% of affected area)
SPREAD PREVENTED: 12,400 residents spared Critical threshold

CRITICAL FINDING:
Single CLT intervention created cascading resilience
across 3.2km radius. Spread rate decreased 31%
in adjacent zones through network effects.
═══════════════════════════════════════════════════
                
07:30 – 09:30
CONCEPT · PERMAGAME

PERMAGAME

PLAY → INTERVENE → MEASURE → EXPORT
                ↓
              ITERATE   (Permagame loop)
                
09:30 – 10:30
═══════════════════════════════════════════════════════
"The master's tools will never dismantle
 the master's house." — Audre Lorde

"Maps are never neutral. They are always
 arguments about the world." — James Corner

"Computation is not objective. It is the
 operationalization of value systems." — Safiya Noble
═══════════════════════════════════════════════════════
                
Repurposing ABM and computation as counter-evidence; maps and models as explicit arguments encoding values of community resilience and infrastructural reparations.
10:30 – 11:30
┌─────────────────────────────────────────────┐
│  FRONTEND                                   │
│  └─ Leaflet.js (mapping)                    │
│  └─ D3.js / Chart.js (data viz)             │
│  └─ HTML/CSS/JS (Tactical Console)          │
│                                             │
│  COMPUTATION                                │
│  └─ Custom ABM Engine (vanilla JS)          │
│  └─ Turf.js (spatial analysis + KDE)        │
│                                             │
│  DATA SOURCES                               │
│  └─ NYC Open Data (GIS)                     │
│  └─ OpenFEMA (vulnerability)                │
│  └─ HOLC Redlining Maps (historical)        │
│                                             │
│  DEPLOYMENT                                 │
│  └─ Static site (GitHub Pages / Netlify)    │
│  └─ Client-side computation (no server)     │
└─────────────────────────────────────────────┘
                
11:30 – 12:00
APPLICATIONS & FUTURE WORK
Community Workshop
CLT organizers in Mott Haven model intervention impact for city council and funders.
Policy Advocacy
Coalitions export Feral Scorecard into highway expansion lawsuits and zoning hearings.
Pedagogical Tool
Planning students simulate how infrastructural decisions compound over decades.
12:00 – 13:00
Q&A · TACTICAL CONSOLE
Anticipated Questions
14:00 – 15:00
City segment