Sugarscape

Growing Artificial Societies — Epstein & Axtell (1996) / Beinhocker Ch. 5

Where Does Inequality Come From?

Start with a landscape of "sugar" — a renewable resource distributed in two mountain peaks. Drop hundreds of agents onto this world, each with slightly different abilities: some can see farther, some burn less energy. Give them one simple rule: look around, move to the richest empty spot you can see, eat. That's it. No exploitation, no institutions, no cunning. Yet within a few dozen ticks, the egalitarian utopia collapses into staggering inequality. A Gini coefficient rivaling the most unequal nations on Earth emerges from the simplest possible rules. This is the profound lesson of Sugarscape: inequality is not a bug — it is an emergent property of heterogeneous agents on an uneven landscape.

Grid Size

Num Agents

Vision Range

Metabolism Range

Regrowth Rate

Reproduction

Speed 50ms

Preset Experiments

Tick

Population

0.00

Gini

Mean Wealth

Median Wealth

Deaths

Mean Vision

Mean Metab

What's Happening

Press Start to watch agents forage on the sugar landscape. Each dot is an agent; brighter cells have more sugar. Watch inequality emerge from simple rules.

Sugar Landscape

Gold/yellow intensity shows sugar levels. Colored dots are agents — green for wealthy, red for poor. Agents gravitate toward the two sugar mountains.

Low Sugar

Med Sugar

High Sugar

Rich Agent

Poor Agent

Population Over Time

As agents with high metabolism and low vision starve, the population drops rapidly at first, then stabilizes. Epstein and Axtell called this the "carrying capacity" of the landscape.

Gini Coefficient Over Time

The Gini coefficient measures inequality from 0 (perfect equality) to 1 (one agent has everything). Watch how quickly it rises even though no agent is "exploiting" another.

Wealth Distribution

The wealth distribution is highly right-skewed — a long tail of wealthy agents and a large mass of poor ones. This closely resembles real-world wealth distributions.

Wealth Over Time (Mean & Median)

The growing gap between mean and median wealth is a signature of skewed distribution. When the mean far exceeds the median, a few agents hold most of the sugar.

Beinhocker's Key Insight

In The Origin of Wealth, Beinhocker uses Sugarscape to demolish a common misconception: that inequality requires exploitation, corruption, or unfair institutions. Even in a world of perfectly simple, identical rules — where no agent can cheat, steal, or collude — heterogeneity alone produces inequality. Agents with slightly better vision find sugar faster; agents with lower metabolism keep more of what they find. These tiny differences compound over time, producing a wealth distribution that looks startlingly like the real world. The lesson: inequality is the default emergent state of any system with heterogeneous agents and uneven resources. Reducing it requires deliberate, sustained intervention — it will not happen on its own.

Fun Facts About Sugarscape

> Published in 1996, Sugarscape was one of the first large-scale agent-based models. Epstein and Axtell ran it on hardware with less computing power than a modern smartwatch.
> The "two mountains" landscape was inspired by geographic resource distribution. The agents' migration toward the peaks mirrors real human settlement patterns around fertile land and water.
> Adding just two rules — trade and reproduction — produces emergent phenomena including market prices, cultural transmission, warfare between "tribes," disease epidemics, and economic cycles.
> The Gini coefficient in a standard Sugarscape run typically stabilizes between 0.4 and 0.6, comparable to many real-world economies. The US Gini was ~0.49 in 2023.
> Epstein later coined the term "generative social science" — the idea that if you haven't grown it from simple rules, you haven't fully explained it. Sugarscape was the manifesto.
> Natural selection in Sugarscape favors high-vision, low-metabolism agents. Over time, the surviving population's average traits shift — evolution without any genetic algorithm.