The Evolution of Trust
← Back
Game Theory & Social Trust — Interactive Experience
The Evolution of Trust
Nicky Case's interactive guide to the game theory of why and how we trust each other — based on Robert Axelrod's groundbreaking research on the mathematics of cooperation.
"Our problem today isn't just that people are losing trust — it's that our environment acts against the evolution of trust."— Nicky Case, The Evolution of Trust
Origins & significance
Background
The Evolution of Trust is a free interactive web experience created by Nicky Case in 2017. Based on Robert Axelrod's 1984 book The Evolution of Cooperation — which documented famous computer tournaments where game theorists submitted strategies to compete in the Iterated Prisoner's Dilemma — the experience opens with the 1914 Christmas Truce, when WWI soldiers spontaneously stopped fighting and shared gifts across no man's land, as proof that cooperation emerges even between enemies. Case also drew from Axelrod's 1997 sequel and Robert Putnam's Bowling Alone (2000) on declining social capital.
Created
2017
Nicky Case
Based on
Axelrod (1984)
The Evolution of Cooperation
Framework
Iterated Prisoner's Dilemma
Game theory's most studied model
Play time
~30 minutes
Free, browser-based
The game — how it works
Two players face a choice each round: Cooperate (put a coin in a machine) or Cheat (don't). If the other player cooperated, you receive 3 coins. If you cooperated, they receive 3 coins. Each coin you put in costs you 1. In a single round, cheating is always the "rational" move. But when you play repeated rounds against the same person, everything changes.
Both cooperate
+2 / +2
Win-win — the foundation of trust
You cheat, they cooperate
+3 / −1
Exploitation — tempting but unsustainable
Both cheat
0 / 0
Deadlock — mutual suspicion, zero value created
The characters — eight archetypes of human strategy
Each character represents a famous game theory strategy. They are not just game pieces — they are models of how real people approach trust.
Copycat
Tit for Tat — Anatol Rapoport
Start with cooperate. Then copy whatever the other player did last round. Cooperate with cooperators, retaliate against cheaters, forgive when they return to cooperation.
Tournament winner
Always Cheat
All-D (All Defect)
Cheat every round. No exceptions. Exploits the naive but scores 0/0 against other cheaters and gets punished by retaliatory players.
Hostile · Self-defeating long-term
Always Cooperate
All-C (Golden Rule)
Cooperate every round. No matter what. Gets the best mutual outcomes with cooperators but is endlessly exploited by cheaters.
Nice · Naive · No retaliation
Grudger
Friedman strategy
Start cooperating. Keep cooperating. But if cheated even once, cheat back forever. No second chances. Ever.
Retaliatory · Unforgiving · Brittle
Detective
Adaptive prober
Test sequence: Cooperate, Cheat, Cooperate, Cooperate. If you retaliate → play fair. If you never retaliate → switch to exploit mode.
Analytical · Exploitative when possible
Copykitten
Tit for Two Tats
Like Copycat but more forgiving. Only retaliate after two consecutive cheats. A single cheat might be a mistake.
Forgiving · Wins in noisy environments
Simpleton
Pavlov / Win-Stay-Lose-Shift
Start cooperating. If rewarded, repeat last move. If punished, do the opposite. Simple reinforcement learning.
Adaptive · Simple logic
Random
Chaos agent
50/50 chance of cooperate or cheat each round. No strategy. No memory. No learning. Pure noise.
Unpredictable · Baseline
What the simulations reveal — four scenarios
1
One-on-one: you play each character
You play 10 rounds against each archetype. Against Always Cooperate, you can exploit them. Against Copycat, mutual cooperation is optimal. Against Always Cheat, you learn that trust without verification is self-destruction. There is no single "best move" — the optimal strategy depends entirely on who you're facing.
Insight: Context determines everything. The same move is brilliant against one opponent and catastrophic against another.
2
Round-robin tournament: all characters face each other
Copycat wins the tournament — not by beating anyone outright, but by cooperating with cooperators (+2/+2) and refusing to be exploited by cheaters. Always Cheat does well against Always Cooperate but poorly everywhere else. The winner never dominates a single match — they do consistently well against the widest range of opponents.
Insight: The winning strategy isn't the most aggressive or the most generous — it's the one with the best average across all possible opponents.
3
Evolutionary tournament: populations compete over generations
Winners reproduce, losers shrink. Starting with 15 Always Cooperate, 5 Always Cheat, and 5 Copycat — the cheaters devour the cooperators first and multiply. It looks like the cheaters win. But then, with fewer victims to exploit, Always Cheat starts scoring 0/0 against other cheaters and gets outperformed by Copycat. Over many generations, Copycat takes over the entire population. Always Cheat goes extinct.
Insight: Exploitative strategies destroy their own food supply. The predator that kills all the prey starves. Short-term winners who depend on victims create the conditions for their own extinction.
4
The twist: when mistakes happen
Each round now has a 5% chance of accidental miscommunication. Between two Copycats, a single accident triggers an endless revenge cycle — one accidentally cheats, the other retaliates, the first retaliates back, spiraling over a mistake neither intended. Copykitten (Tit for Two Tats) emerges as the winner — it absorbs the occasional accident without triggering a spiral, while still punishing deliberate exploitation. But when errors exceed ~10%, even Copykitten fails and hostile strategies prevail.
Insight: In imperfect environments, a bit of forgiveness is mathematically optimal. But too much forgiveness enables exploitation. The calibration between forgiveness and retaliation is the core skill.
Strategic principles — what the math proves
01
Be nice first
Winning strategies always start by cooperating — never by cheating. The player who opens with hostility forfeits the possibility of mutual gain before the game begins. In any new relationship, lead with trust. Not naive trust, but deliberate, eyes-open trust that establishes the possibility of cooperation.
02
Retaliate when exploited
Niceness without retaliation is not a strategy — it is an invitation to exploitation. Always Cooperate scores worse than Copycat because it never pushes back. Trust without boundaries is not generosity — it is self-destruction. The person who never retaliates trains others to exploit them.
03
Forgive after retaliating
Grudger retaliates forever after a single betrayal — locking itself into permanent conflict over one incident. Copycat returns to cooperation the moment the other player does. Permanent grudges are strategically inferior to calibrated forgiveness. The cost of maintaining a grudge almost always exceeds the satisfaction of perpetual revenge.
04
Be clear
Copycat's pattern is immediately readable: cooperate with cooperators, retaliate against cheaters, return to cooperation when they do. Predictable fairness attracts cooperators and deters cheaters. Strategic clarity is power. Consistent, legible values attract better partners and repel manipulators.
05
Forgive mistakes, not patterns
The Copykitten refinement: a single cheat might be an accident. Two in a row is a policy. The first offense deserves the benefit of the doubt. The second consecutive offense deserves a response. Retaliating against every slight creates permanent conflict. Never retaliating against anything creates permanent exploitation.
06
Exploitation destroys its own food supply
Always Cheat wins early by devouring Always Cooperate. But once the cooperators are gone, cheaters face each other — and score 0/0 forever. Any strategy that depends on the existence of victims is self-limiting. The company that exploits all its customers has no customers. The person who burns all their bridges stands on an island.
07
The game shapes the players
Whether people cooperate or cheat depends less on their character and more on the structure of the game. Change the payoffs, the repetition, or the communication quality — and you change the behavior. If you want to change how people behave, change the game they're playing. Designing environments that reward cooperation is more effective than hoping for better people.
The three conditions for trust to evolve
1
Repeat interactions
Trust requires the knowledge of future encounters. If you'll never see someone again, cheating is mathematically rational. When you know you'll interact again, cooperation becomes strategic — today's cooperation earns tomorrow's in return.
Erosion: Shorter jobs, more mobility, digital anonymity reduce repeat interactions
2
Possible win-wins
Trust evolves only where mutual cooperation beats mutual cheating. Both cooperating yields +2/+2; both cheating yields 0/0. If the game is zero-sum, there is no incentive to cooperate. Trust requires "we can both be better off."
Erosion: Winner-take-all economics and us-vs-them framing eliminate win-wins
3
Low miscommunication
If the error rate exceeds ~10%, noise overwhelms signal and cooperative strategies collapse. Accidental cheating triggers spirals that destroy trust faster than forgiveness repairs it.
Erosion: Social media, polarization, and filter bubbles maximize miscommunication
In business
Start every new relationship with cooperation — customers, vendors, partners, employees. Retaliate immediately when exploited — renegotiate terms, enforce contracts, end partnerships that violate trust. Forgive when the behavior changes — don't let a single breach permanently destroy a valuable relationship. Be clear and consistent — make your values and boundaries legible so cooperators seek you out and exploiters avoid you. And the structural lesson: design your business environment to satisfy the three conditions. Create repeat interactions (long-term contracts, loyalty programs). Ensure win-win dynamics (aligned compensation, mutual-benefit partnerships). Minimize miscommunication (transparent metrics, clear expectations). The companies that build these conditions don't need to hope for trustworthy partners — they create the environment where trustworthiness is the winning strategy.
In life
Don't be Always Cooperate — the person who trusts everyone unconditionally and gets exploited repeatedly. Don't be Always Cheat — the person who trusts no one and lives in zero-value interactions. Don't be Grudger — cutting people off forever over a single transgression. Be Copykitten — lead with trust, forgive honest mistakes, retaliate against deliberate exploitation, and return to cooperation when the other person does. And when trust feels impossible, don't blame the people — examine the game. Are there repeat interactions? Is win-win possible? Is communication clear? If the answer to any is no, the game itself is broken — and no amount of personal virtue can fix what the structure prevents.
Bottom line
The Evolution of Trust proves that cooperation is not naivety — it is the mathematically optimal strategy in any repeated, non-zero-sum interaction with tolerable noise. The winning formula: be nice, be retaliatory, be forgiving, be clear. But the deepest lesson transcends individual strategy: trust is not a character trait — it is a product of the environment. Build repeat interactions, create win-win possibilities, and reduce miscommunication — and trust will evolve. Destroy these conditions, and no amount of moral exhortation will save it. The question is never "why don't people trust each other?" It is: "what about the game makes trust the losing strategy?"
