Thought Toys · Emergence · Exhibit 88
Two hundred fifty dots, each one only ever doing one thing: turn to match the handful of neighbours within its own small radius, then wobble a little on its own. No one sees the whole sky, no one is in charge, no one is told to form up. Turn the wobble down and the entire field falls into a single moving flock anyway.
—
order φ — · noise η — · sky L —
Every dot on the field is a bird with a direction of travel and nothing else — no map, no destination, no idea where the flock as a whole is headed. Each one only looks at the handful of birds within its own small neighbourhood and turns to match their average heading. Then, because no turn is perfect, it adds a little random static of its own before moving forward. That is the entire rule. There is no leader bird, no shared signal, and no bird that can see more than its own tiny patch of sky.
At high static, the rule barely matters — each bird's own wobble overwhelms whatever its neighbours are doing, and the field is a directionless scatter: as unaligned as birds facing uniformly random directions would be. Turn the static down and, for a while, still nothing changes much. Then, past a threshold, small pockets of agreement start to catch: a few birds nudge each other into alignment, that little patch is now a stronger "vote" for its neighbours to match, and the patch grows. Watch it happen and there's no moment where anyone announces it — the field simply stops being a scatter and becomes a current.
The sky dial matters too, and for a reason worth sitting with: at the exact same amount of static, a packed sky holds a flock together that a roomy sky cannot. More neighbours in reach means each bird's turn is averaged over a bigger, steadier vote, which is more resistant to any one bird's wobble. Flocking isn't only about how careful each turn is — it's also about how many neighbours there are to be careful about.
improve/verify/88-vicsek.js, 12
checks): at the default density, low noise gives φ > 0.95 and full noise (η = 2π)
lands within 0.03 of the exact finite-size random baseline √(π/4N); a re-run across an
eleven-point noise sweep is non-increasing everywhere within tolerance; and at fixed noise, a packed sky
(L = 6) sustains materially more order than a roomy one (L = 16). Negative
controls: switch off neighbour-averaging (every bird reacts only to itself) and, at the SAME low
noise that gave φ > 0.95 with the rule on, φ collapses back to the random baseline —
proving the flock needs the alignment rule itself, not merely a quiet sky; the same holds at the SAME packed
density that helped order survive high noise with the rule on. Robust across five independent random seeds.
Also in Emergence: Phantom Traffic Jams →
← the cabinet · Thought Toys — a cabinet of explorable explanations. Exhibit 88.