Crowdsourcing and Citizen Science

A single lab studying a million galaxy photos would take years. Split that pile into small batches and hand them to a hundred thousand volunteers online, and each person only classifies a handful. The task finishes fast, and because so many independent people look at each image, mistakes get caught and averaged out. This is the core trick of crowdsourcing: divide one huge job into tiny identical pieces, let anyone with time and interest take a piece, then combine all the small answers back into one result.

Citizen science applies this to collecting real-world data rather than just labeling existing data. Picture a bird-counting app where anyone can log the species and location of a bird they see. Each sighting is tiny on its own, but thousands of people submitting sightings every day builds a live, constantly updating map of bird populations across a whole region, something no single research team could gather by walking around themselves.

The catch is who shows up to participate. If almost all volunteers live in one city, the resulting map looks detailed there and blank everywhere else, no matter how many total sightings come in. A large volunteer count does not fix this: it is concentrated, not spread out, so conclusions drawn from it about places with few or no volunteers are unreliable. This is the same idea as distributed computing, where one large computation is split across many machines working in parallel. Crowdsourcing is that same strategy applied to human attention and judgment instead of processor cycles, with the extra wrinkle that people, unlike machines, are not evenly distributed or equally available.