Every scout that inspects a candidate site flies home and "dances" for a duration proportional to that site's quality, recruiting idle nestmates to go check it out. Better sites get longer, more vigorous dances, so they recruit faster — a positive feedback loop:
P(recruit → site i) = (support_i · quality_i) / Σⱼ (support_j · quality_j)
The colony commits the moment one site's accumulated support reaches the
quorum threshold Q:
support_i ≥ Q → swarm commits to site i
No single bee ever inspects every option or decides for the group — this decentralised cross-checking is how real honeybee swarms pick a new nest site in the wild.
A honeybee swarm has to choose a new home together, with no single bee in charge. This simulation models the real mechanism — scouts inspect candidate sites, dance in proportion to how good each one is, and the colony commits the instant enough scouts agree.
Scouts fly out from the swarm cluster to one of four candidate nest sites, evaluate it, then fly home and "dance" for longer at higher-quality sites — recruiting more nestmates and accelerating support for the best options through positive feedback, until one site reaches quorum and the whole swarm commits.
Lower the quorum threshold for faster decisions, raise the exploration bias to see more scouts checking sites at random instead of following dances, and adjust dispatch rate and speed to slow the process down or speed it up. Use "New candidate sites" to reroll site quality at any time.
Real swarms of scout bees can evaluate dozens of potential nest cavities spread over several square kilometres, cross-checking each other's dances, and typically converge on the single best option without any individual bee ever having visited every candidate itself.