Skip to content

README.md

Latest commit

 

History

History
114 lines (87 loc) · 3.54 KB

README.md

File metadata and controls

114 lines (87 loc) · 3.54 KB

World Model Documentation

A system for modeling a person's worldview as an adversarial equilibrium of competing internal tendencies.

Quick Links

What Is This?

The World Model represents a person not as a static profile, but as a dynamic equilibrium of competing internal drives. Seven human tendencies (survival, status, meaning, connection, autonomy, comfort, curiosity) act as agents that:

  1. PROPOSE claims about what matters ("Financial security is foundational")
  2. STAKE observations to support their claims and undermine competitors
  3. WIN or LOSE debates based on evidence
  4. GAIN or LOSE influence based on outcomes

The equilibrium that emerges IS the personality.

Key Insight

Ideas are atomic. Position is relational.

The same observation means different things depending on what you're optimizing for:

  • "Lives paycheck to paycheck at 42"
    • SURVIVAL: PRO (evidence of financial risk)
    • MEANING: PRO (sacrifice for purpose)
    • AUTONOMY: PRO (chose freedom over salary)
    • COMFORT: PRO (evidence of unsustainable intensity)

Quick Start

from world_model import (
    ObservationStore, AgentSet, Arena,
    WorldModel, create_world_model,
)

# Load a person's observations
model = create_world_model("Person", "observations.json")

# Run adversarial debate
arena = Arena()
trees, result = arena.run_full_debate(
    observations=model.observations,
    agents=model.agents,
    rounds=1,
)

# See who won
print(f"Winner: {result.winner}")
print(f"Final allocations: {model.agents}")

ML-Style Training

from world_model.dynamics import Trainer, TrainConfig

config = TrainConfig(
    max_epochs=5,
    convergence_threshold=0.01,
    validation_split=0.2,
)

trainer = Trainer(config)
history, result = trainer.train(
    observations=model.observations,
    agents=model.agents,
)

# Check validation accuracy
print(f"Accuracy: {history.validation_results[-1].accuracy:.1%}")
print(f"P-value: {history.validation_results[-1].p_value:.4f}")

Project Structure

life/
├── world_model/
│   ├── models/           # Core data structures
│   │   ├── observation.py    # Atomic facts
│   │   ├── agent.py          # Human tendencies
│   │   └── tree.py           # Value hierarchies with weight propagation
│   ├── extraction/       # Extract observations from text
│   ├── staking/          # Legacy staking mechanisms
│   ├── dynamics/         # Adversarial competition
│   │   ├── arena.py          # Debate orchestration
│   │   └── trainer.py        # ML-style training loop
│   └── storage/          # Persistence (JSON, Firestore)
├── api/                  # FastAPI service
├── docs/                 # Documentation
└── test_*.py             # Test scripts

Validation Results

First training run on 165 observations:

Metric Value
Accuracy 27.3%
Baseline 7.1% (1/7 random)
P-value 0.001
Significant Yes

The model predicts which tendency "owns" an observation nearly 4x better than chance.