Galaxy Zoo turned a million galaxy images into a public classification game

The Sloan Digital Sky Survey gave astronomers nearly a million galaxy images, but morphology labels — spiral or elliptical — were too labor-intensive for a small research team. In 2007 Galaxy Zoo put the images online and asked the public to classify them.

Each galaxy was inspected by many volunteers, and aggregating their clicks produced classifications consistent with subsets classified by professional astronomers. The 2008 Monthly Notices paper reported more than 40 million classifications from roughly 100,000 participants.

The 2010 follow-up showed machine learning trained on the human labels could reproduce the crowd's classifications, turning volunteer effort into training data for automated astronomy. The project demonstrated that a well-framed public task can do work a research team cannot.

Crowdsource classification, keep the science

Many eyes per galaxy catch individual errors.

Public engagement made the work free and fast.

Expert validation kept the results scientifically credible.

Crowd labels became training data for machine learning.

A task too big for experts can be split into micro-decisions the public does well, as long as each item gets many independent judgments and expert samples validate the crowd.

Galaxy Zoo grew into the Zooniverse platform hosting dozens of citizen-science projects, and its morphological catalogue became a standard dataset for studying how galaxies form and evolve.