The Neighbourhood
Every roof below is a household, slate blue or terracotta. Nobody here hates anybody. Each household asks one mild thing: that a few of its neighbours be like itself. If too few are, it lights a lantern, packs, and moves to any empty house going. Set them moving and watch what “mild” builds.
Before you play
A household’s neighbours are the eight houses around it, across the lanes as well. The comfort rule says how many of those must be like it. It starts at 30 per cent: fewer than one in three, a wish most people would call easy-going. Make a guess before you begin, then try it gentler, and stricter.
What you just watched
This is Thomas Schelling’s segregation model (1971), one of the founding demonstrations of agent-based social science, which he first worked out by hand, moving coins about a board. The unsettling part is the arithmetic: a preference as gentle as “I’d like three in ten of my neighbours to be like me” — a wish nobody would call extremism — reliably sorts the town into clear territories, where the average household ends up with about three-quarters of its neighbours like itself, more than twice what it asked for. Slide the rule down to 25 per cent and the town only gathers into small clumps; one notch up, at 30 per cent, the territories form. No household chose the outcome. Every household chose only its own comfort.
Watch who moves, too. At 30 per cent, about one household in five is unhappy on the first day, and the average household already has more neighbours like itself than it asks for. But every departure leaves a gap in someone else’s street, and the moves cascade: by the time the town settles, many more households have packed than were ever unhappy at the start, some of them several times. One small liberty with Schelling: his households moved to the nearest spot that suited them; these take any empty house going, and may find themselves unhappy there too.
The lesson generalises far beyond housing, and it is one of the load-bearing ideas in how I think about technology: systems produce outcomes that no participant intends, which means blame is often the wrong instrument and design is the right one. Feeds, defaults, recommender loops, agent swarms — each is a neighbourhood like this one, quietly compounding mild preferences into structures nobody voted for. If you want a different pattern, you rarely need different people. You need different rules of moving.
More play on the Playground · the essays on technology and society