Skip to content

DATA_ARCHITECTURE.md

Latest commit

 

History

History
209 lines (178 loc) · 6.23 KB

DATA_ARCHITECTURE.md

File metadata and controls

209 lines (178 loc) · 6.23 KB

DATA ARCHITECTURE: PURE HISTORY vs GAME ADAPTATIONS

ARCHITECTURAL PRINCIPLE

Separation of Concerns: Pure historical facts vs. game-specific adaptations

CURRENT ARCHITECTURE DECISION

pdoom-data (This Repo) = PURE HISTORICAL FACTS

Purpose: Authoritative source of unmodified historical AI safety events Principles:

  • Scholarly accuracy and integrity
  • Useful to other researchers
  • Real probability estimates from actual analysis
  • Unmodified source citations
  • No game-specific simplifications or ratings

Content Standards:

  • Historical accuracy is paramount
  • All claims must be sourced
  • Probability estimates based on real analysis
  • Events described as they actually occurred
  • Suitable for academic or research use

pdoom1-data (Game Repo's Data) = GAME ADAPTATIONS

Purpose: Game-playable versions adapted from pure historical data Principles:

  • Gameplay balance and fun
  • Simplified for player comprehension
  • Game-specific ratings and mechanics
  • Modified for narrative flow
  • Optimized for player engagement

Transformation Process:

pdoom-data (Pure History) 
    |
    | [Fetch and Transform]
    |
    v
pdoom1-data (Game Version)

IMPLEMENTATION STRATEGY

1. Keep pdoom-data PURE

What stays in pdoom-data:

  • Original event descriptions
  • Real dates and actual participants
  • Authentic source citations
  • Scholarly probability estimates
  • Unmodified historical context
  • Research-grade documentation

What gets removed from pdoom-data:

  • game_impacts arrays (these are game-specific)
  • rarity classifications (game mechanic)
  • Simplified/dramatized descriptions
  • Player-facing language
  • Game balance considerations

2. Create Transformation Pipeline

In pdoom1 game engine:

# pdoom1/src/data/historical_adapter.py
class HistoricalEventAdapter:
    def fetch_from_pdoom_data(self):
        """Fetch pure historical events from pdoom-data repo"""
        # Load from JSON exports or API
        
    def transform_for_gameplay(self, historical_event):
        """Transform pure history into game-playable format"""
        # Add game mechanics
        # Simplify language for players
        # Add rarity classifications
        # Balance for gameplay
        
    def generate_game_impacts(self, event):
        """Generate game-specific impact arrays"""
        # Convert historical significance to game variables

3. Proposed File Structure Changes

pdoom-data (Pure History):

data/events/
+-- historical_events.json           # PURE historical data
    {
      "event_id": "openai_gpt4_2023",
      "title": "OpenAI Releases GPT-4",
      "date": "2023-03-14",
      "description": "OpenAI announced GPT-4, a large multimodal model...",
      "historical_significance": "Major milestone in AI capability development",
      "probability_impact_analysis": {
        "methodology": "Expert survey of AI researchers",
        "p_doom_change_estimate": "+5% (median estimate)",
        "confidence_interval": "2% to 12%",
        "sample_size": 47,
        "survey_date": "2023-04-15"
      },
      "sources": [
        "https://openai.com/research/gpt-4",
        "https://arxiv.org/abs/2303.08774"
      ],
      "research_notes": "Represents significant capability jump...",
      "related_events": ["openai_chatgpt_2022"],
      "verification_status": "peer_reviewed"
    }

pdoom1-data (Game Adaptations):

data/game_events/
+-- gameplay_events.json             # GAME-ADAPTED versions
    {
      "event_id": "openai_gpt4_2023",
      "title": "GPT-4 Released!",
      "player_description": "A breakthrough AI model shocks the world...",
      "game_impacts": [
        {"variable": "ai_capability", "change": 45},
        {"variable": "public_awareness", "change": 30},
        {"variable": "research_funding", "change": 25}
      ],
      "rarity": "rare",
      "gameplay_weight": 0.8,
      "unlock_conditions": ["year >= 2023", "research_level >= 3"],
      "player_choices": [
        {"text": "Accelerate our research", "effects": {...}},
        {"text": "Focus on safety", "effects": {...}}
      ],
      "based_on_historical_event": "openai_gpt4_2023",
      "adaptation_notes": "Simplified for gameplay, added player agency"
    }

BENEFITS OF THIS ARCHITECTURE

For Researchers

  • Pure historical accuracy in pdoom-data
  • Citable source for academic work
  • Unbiased probability estimates
  • Complete source documentation

For Game Development

  • Gameplay-optimized content in pdoom1
  • Player-friendly language and concepts
  • Balanced mechanics for fun gameplay
  • Flexible adaptation without corrupting source data

For Maintenance

  • Clear separation of concerns
  • Source of truth remains pure
  • Game updates don't affect historical accuracy
  • Multiple games could use same historical base

MIGRATION PLAN

Phase 1: Clean Current Data

# Remove game-specific elements from historical events
python scripts/purify_historical_data.py

# This removes:
# - game_impacts arrays  
# - rarity classifications
# - game-specific descriptions

Phase 2: Enhanced Historical Documentation

# Add scholarly elements to historical events
python scripts/enhance_historical_accuracy.py

# This adds:
# - Detailed source citations
# - Probability analysis methodology
# - Research confidence levels
# - Peer review status

Phase 3: Create Game Adapter

# In pdoom1 repository
class PDoomDataAdapter:
    def sync_from_historical_repo(self):
        # Fetch latest historical events
        
    def generate_game_versions(self):
        # Transform for gameplay
        
    def maintain_traceability(self):
        # Link game events to historical sources

VALIDATION STANDARDS

pdoom-data Quality Standards:

  • All events must have primary sources
  • Probability estimates must include methodology
  • Historical accuracy verified by multiple sources
  • Suitable for citation in academic work

pdoom1-data Quality Standards:

  • Balanced for engaging gameplay
  • Clear player-facing language
  • Consistent game mechanics
  • Maintains connection to historical truth

This architecture ensures pdoom-data becomes a valuable resource for AI safety researchers while still enabling rich game content through thoughtful adaptation.