The initial planet candidates are found by an algorithm that scans TESS’ images. Then, astronomers enlist a second, highly effective algorithm to confirm the results of the first program. But even these cutting-edge computer codes get tripped up sometimes, leaving huge numbers of potential exoplanets that need to be confirmed manually.
“With TESS, the number of targets to look at now is in the millions, and even after you run the vetting software, the number is still in the millions,” says NASA astronomer and citizen science officer Marc Kuchner.
Most of those millions of targets are not exoplanets at all. But NASA still needs to do a basic analysis to find out if it's noise, a binary star, or something more interesting. Astronomers can’t handle such a large job on their own, so citizen scientists have stepped in to take up the task.
Citizen science in astronomy
Turning to citizen science volunteers for this kind of help isn’t unusual in astronomy these days. Modern astronomical surveys commonly use computer algorithms to discover new objects in their vast archives of images. But astronomers know that computers can’t catch everything. And the human brain is often still better at identifying new planets than a computer.
“The datasets we’re getting for astronomy are just so huge right now,” says Kuchner, who also works on Planet Patrol. “Citizen scientists have already discovered half of the known comets and most of the long-period exoplanets. They’re making discoveries left and right. I think that astronomy projects are realizing that they need to work with the citizen science community to make the most of their data science.”
Citizen scientists have already helped discover hundreds of new exoplanets, thousands of brown dwarfs, plus other strange objects, like Boyajian’s star.
Volunteers with Planet Hunters TESS have made over 250,000 classifications of astronomical objects. And so far, some 5,000 Planet Patrol volunteers have made more than 400,000 image classifications. In fact, they’ve already analyzed all the project’s uploaded data, so the researchers are working on uploading new images for analysis.
In Planet Patrol, users are asked to look at an image and apply some basic categorization, like whether there’s a bright spot in the center of the image, more than one bright spot, or no well-defined bright spot.
“Our automated algorithms are correct around 90 percent of the time,” says Planet Patrol’s project leader, Veselin Kostov, who works as a researcher at NASA’s Goddard Space Flight Center and the SETI Institute. “[But] sometimes, they struggle with weak signals.”
According to Kostov, the algorithms can get tripped up by false-positives, which look like the signals from exoplanets but are actually other natural phenomena. Common sources of confusion include stellar eclipses and occultations, where one star passes in front of another, as well as instrument artifacts.
