How to Grow Your Own Army of Citizen Scientists

Eli is a contributing correspondent for Science magazine.

In 2007, astronomer Chris Lintott and colleagues were drowning under a data deluge—1 million images of galaxies to characterize and only one graduate student to do it. His student characterized 50,000 in a week "before telling me where to stick the other 950,000 galaxies," Lintott explained during a presentation last weekend at the annual meeting of the American Association for the Advancement of Science. So Lintott, who works at the Adler Planetarium in Chicago, Illinois, and his collaborators set up a Web site in hopes of attracting some volunteers.

The goal was a few thousand helpers. But after some favorable news coverage, the site they dubbed Galaxy Zoo "went viral," says Lintott. The current count is 200 million classifications by 375,000 people working from the comfort of their own homes. More than 20 astronomical papers have come out of the project, including the discovery of two astronomical phenomena. (The data come from this project.)

Now Lintott's team has expanded its citizen-science efforts to other projects, including studies of the moon and an analysis of old ship logs for climate data. Here are some of his suggestions to scientists for successful partnerships with armchair scientists:

Build a bulletproof interface so that you don't "waste people's time." Repeated testing to make Galaxy Zoo both useful and intuitive were crucial, said Lintott, since "you only get one chance" to impress the public.

People can sense when they are being given busywork, he adds, but the interface must be straightforward. Galaxy Zoo, for example, offers volunteers a few simple options per galaxy on type or orientation.

Although the site has attracted both astronomy buffs and students completing a homework assignment, Lintott said, "The motivation for most people [is] contributing to research." As one user wrote, "Astronomy, the oldest science, is one of the few that still has ample room for citizen scientists. We can't all build particle accelerators or experiment with gene splicing, but anyone can participate here."

Provide tools and data for advanced users. Galaxy Zoo's most passionate citizen astronomers can delve deep into the data. And this can lead to real discoveries. In 2007, as the astronomers behind Galaxy Zoo were still dealing with the deluge of users, a small group of Zooites, as they call themselves, puzzled over galaxy images that looked like little green spheres. Calling themselves the Peas Corps, the team downloaded spectral and other data on the objects, set up a special Web site to analyze them, hired a programmer, and presented a full analysis of the objects to the pros. "All we had to do was write the conclusion," says Lintott. Several were included on the paper which announced the discovery of the objects, which turn out to be some of the most active star-producing galaxies astronomers have ever seen.

Ranking users can backfire. The database calculates which users are doing a better job at classifying galaxies by comparing their classifications. Early on, the system gave users a score to encourage better performance. But under that system, the users ranked as the best classifiers ended up stopping their work after achieving that distinction. "They felt they had won the video game," said Lintott.

Computers and humans need each other. Multiple citizen scientists do a better job than one exhausted graduate student, and computer classifications have an error rate of 20% to 30%. Human classifications help train computers to reduce that error. With better telescopes finding millions more galaxies, the number of required classifications will simply overwhelm the human army. And with so much information available, scientists will weed out the objects easiest to classify first. That leaves the toughest challenges for the volunteers.

But machines aren't curious. In 2007, Dutch school teacher and Galaxy Zoo hobbyist Hanny van Arkel wondered about a greenish unidentified object appearing in images adjacent to galaxies. The object became known as "Hanny's Voorwerp" and astronomers took up the hunt to identify the new phenomenon. "People are now chasing it with some of the best telescopes on the planet," said Lintott. "Computers will slowly get better at classifying galaxies, but looking at an image and asking, 'What's that odd thing?' remains uniquely human," the Galaxy Zoo site explains.

Posted in Technology, Space