Galaxy Zoo turns galaxy sorting into an online game — 36M classifications in months

Astronomers needed to classify the shapes of a million galaxies from the Sloan Digital Sky Survey, a job humans do better than computers but that would take scientists years; Galaxy Zoo opened the task to anyone with a browser.

Give the public a simple online task scientists can't automate — classify galaxy shapes — and let a website game feed real astronomy data and discoveries.

The team launched Galaxy Zoo in July 2007 with a deliberately simple interface — users look at an image and click whether the galaxy is elliptical, spiral or something else. The site went viral after BBC coverage, hit 70,000 classifications an hour, and by November 2007 every galaxy had been classified by an average of 40 people; a forum let volunteers flag anomalies.

More than 36 million classifications within a few months, matching professional astronomers' results; volunteers discovered Hanny's Voorwerp — later shown to be a light echo from a long-dead quasar — and the green pea galaxies, which a 2016 Nature paper linked to the reionization of the early universe.

In July 2007 astronomers from Oxford and Portsmouth launched Galaxy Zoo: a website where anyone could look at an image of a galaxy from the Sloan Digital Sky Survey and click a button saying whether it was elliptical, spiral or something else. The task was one scientists needed done a million times over but that computers still could not reliably automate.

The design made participation effortless — one image, a few buttons, instant feedback — and a forum let volunteers discuss anything odd they saw. BBC coverage made the story one of the most e-mailed of its day and briefly crashed the server; traffic reached 70,000 classifications an hour, and by the end of November 2007 every galaxy in the sample had been classified by an average of 40 people, far exceeding the team's hopes.

The crowd did more than count shapes. Dutch schoolteacher Hanny van Arkel spotted an odd green blob, now called Hanny's Voorwerp, which follow-up observations showed to be a light echo from a quasar that had switched off more than 40,000 years ago; volunteers also noticed the compact 'green pea' galaxies, subject of a 2009 paper and a 2016 Nature study using Hubble that linked peas to the reionization of the early universe. Galaxy Zoo's model went on to power Zooniverse's hundreds of citizen-science projects.

It reduced a scientific bottleneck to one repeatable click, so a million classifications became a game anyone could play.

The forum made anomalies collectible — volunteers flagged what the workflow didn't ask about, which is how Hanny's Voorwerp and the green peas were found.

Feedback was instant and the site was social, so BBC coverage converted into 70,000 classifications an hour instead of a spike of curiosity.

Trusting the crowd's pattern recognition produced classifications that matched professional astronomers, at 40-fold redundancy.

The platform outlived its first dataset, becoming Zooniverse and exporting the model to dozens of other sciences.

Find the step of your research that needs human eyes, then design a website that makes that step fun: volunteers do the work, flag anomalies, and find what you didn't know to look for.

Galaxy Zoo expanded into new phases and surveys, and its technology grew into the Zooniverse platform, which now hosts hundreds of people-powered research projects across astronomy, biology and the humanities. The green peas discovered by early volunteers became the subject of a 2016 Nature paper, and the current incarnation is classifying early data from the Vera C. Rubin Observatory, according to the Zooniverse project page.