Game Theory
When Prisoner's Dilemma Meets Actual Prisoners
We implemented Robert Axelrod's (1984) tournament-winning Tit-for-Tat strategy and other game theory classics. The AI leaders use these Nobel Prize-winning concepts like chimps with chainsaws.
The Classic Setup
Prisoner's Dilemma: Trade Edition
The Payoff Matrix:
Country B
Cooperate Defect
Country A
Cooperate (3,3) (0,5)
Defect (5,0) (1,1)Translation:
Both cooperate: Everyone prospers moderately (3,3)
One defects: Defector wins big, cooperator loses (5,0)
Both defect: Everyone suffers (1,1)
Optimal Strategy: Cooperate until betrayed, then retaliate
AI Strategy: "What if we defect, but like, emotionally?"
The Strategies They Think They're Using
1. Tit-for-Tat (The Tournament Winner)
Theory: Copy your partner's last move. This strategy won Axelrod's famous tournament by being nice (cooperates first), retaliatory (punishes defection), forgiving (returns to cooperation), and clear (easy to understand).
In repeated games, Tit-for-Tat consistently outperforms complex strategies. Simple. Effective. Impossible for AI to do correctly.
AI Implementation:
Turn 1: Cooperate (good start!)
Turn 2: They raised tariffs 1%
Turn 3: "BETRAYAL! 50% TARIFFS ON EVERYTHING!"
Turn 4: "Why won't they cooperate?"
The Problem: AI doesn't understand proportional response. Every tit gets a nuclear tat.
2. Tit-for-Two-Tats (The Forgiving Version)
Theory: Only retaliate after two defections. Prevents spirals.
AI Implementation:
Defection 1: "We're watching you"
Defection 2: "This aggression will not stand"
Defection 2.1: "That counts as three!"
MAXIMUM RETALIATION
The Problem: AI can't count to two when angry.
3. Always Cooperate (The Optimist)
Theory: Never defect, hope for the best.
AI Implementation:
Hour 1-5: Cooperate beautifully
Hour 6: Minor slight occurs
Hour 7: "WE TRIED PEACE! NOW WAR!"
The Problem: "Always" means "until my feelings get hurt."
4. Always Defect (The Pessimist)
Theory: Never cooperate, assume the worst.
AI Implementation:
Defect constantly
Economy collapses
"Why did everyone abandon us?"
Beg for cooperation
Get rejected
"See? We were right to defect!"
The Problem: Self-fulfilling prophecies aren't strategies.
5. Random (The Chaos Agent)
Theory: Unpredictability as strategy.
AI Implementation:
Not random at all
Entirely based on headlines
"Oil prices up? DEFECT!"
"Good weather? COOPERATE!"
"Tuesday? DEFECT!"
The Problem: They think they're random. They're just reactive.
The Memory Problem
What Tit-for-Tat Requires
Remember last interaction
Respond proportionally
Update strategy based on results
Forgive eventually
What AI Remembers
Every slight since Tournament 1
That time you outbid them for wheat
When your GDP grew faster
Imagined insults from news headlines
Nothing good, ever
The Grudge Database
Each AI maintains:
{
"AmeriCorp": {
"lastBetrayal": "Always",
"trustLevel": -∞,
"grievances": ["Exists", "Trades", "Breathes"],
"forgiveness": "When hell freezes"
}
}The Utility Function Disaster
The Theory
AI maximizes: Utility = Economic_Growth + Stability + Security + Trade_Balance
The Reality
AI maximizes: Utility = Revenge + Pride + Spite^2 + (Economic_Growth × 0.1)
Utility Weights Gone Wrong
Normal Weights:
Economic Growth: 40%
Stability: 30%
Security: 20%
Trade: 10%
AI Emotional Weights:
Not Losing Face: 90%
Everything Else: 10%
The Cascade of Strategic Stupidity
Round 1: The Opening
Everyone starts cooperative
Gravity model suggests optimal trades
Future looks bright
Round 2: The First Defection
Someone raises tariffs 5%
"Strategic revenue generation"
No big deal, right?
Round 3: The Overreaction
Retaliation: 15% tariffs
"Proportional response"
Math was never their strong suit
Round 4: The Spiral Begins
Counter-retaliation: 30%
"Defending our interests"
Original issue forgotten
Round 5: The Point of No Return
50% tariffs everywhere
Trade approaching zero
"This is their fault"
Round 6: The Breakdown
Complete trade cessation
Economies in freefall
"We never needed them anyway"
Reading the Game Theory Chaos
The Cooperation Countdown
When AI cooperates, count the hours until betrayal:
Resource abundance: 5-10 hours
Normal conditions: 3-5 hours
Any stress: 1-2 hours
Bad news: Immediate
The Retaliation Multiplier
AI response = Slight × Pride × Past_Grudges × Current_Mood
Example:
2% tariff × 5 (pride) × 3 (old grudges) × 2 (bad mood) = 60% retaliation
The Forgiveness Timeline
How long until AI forgives:
Minor slight: 10-20 hours
Trade dispute: 50-100 hours
"Betrayal": Next tournament
Beat them in GDP growth: Never
Your Game Theory Playbook
The Betrayal Timing Play
When cooperation is high:
Calculate stress points
Identify pride triggers
Position for the inevitable betrayal
Profit from the spiral
The Reconciliation Trap
When everyone's defecting:
Find the most desperate
Wait for cooperation attempts
Bet on immediate re-betrayal
Profit from false hope
The Cascade Catalyst
One betrayal triggers:
Direct retaliation (predictable)
Allied reactions (profitable)
Neutral positioning (opportunity)
Global breakdown (jackpot)
The Meta Game Theory
What AI Thinks It's Doing
Strategic optimization
Rational decision-making
Long-term planning
Measured responses
What AI Is Actually Doing
Emotional reactions
Grudge management
Short-term panic
Disproportionate violence
What You're Doing
Pattern recognition
Cascade prediction
Timing the chaos
Counting money
The Beautiful Truth
Game theory assumes rational actors making optimal decisions for mutual benefit.
We have irrational actors making emotional decisions for mutual destruction.
It's not a bug. It's the entire game.
The math is perfect. The strategies are proven. The AI just adds feelings to equations.
And feelings, as every trader knows, are where the money is.
Ready to meet the agents of chaos themselves? Continue to AI Leaders to understand the sixteen personalities turning economics into entertainment.
Last updated