Zooming into the history of Zooniverse
If an adult, teenager, or child wants to get involved in a citizen science project, a good place to start looking for one is Zooniverse, an online platform that provides support to over 80 active civic science projects.
But how did Zooniverse get started? Well, it didn’t start as an online site where people could learn about or access dozens of citizen science projects. Rather, some scientists just needed help sorting through the millions of images they received.
In 2007, a small group of astronomers from the Adler Planetarium in Chicago and UK’s Oxford University had a dataset consisting of images of millions of galaxies as part of the Sloan Digital Sky Survey.
“We knew that with this set of a million galaxy images, we’d be able to answer some really interesting questions about how galaxies evolve over time, their structure, their morphology — all those ideas,” said Laura Trouille, principal investigator for Zooniverse. Trouille is also vice president of science engagement at the Adler Planetarium.
But for these astronomers and graduate students to delve into their studies on the galaxies, they first needed to classify them by the type of galaxy: spiral versus elliptical versus emerging.
And this was a big task. One researcher classified 50,000 galaxies in one month, but even that was still only a “tiny drop in the bucket of this whole million galaxy data set,” Trouille said.
“This is part of what characterizes modern science: large data sets and how efficiently and fully we can really take advantage of these data sets to then do the research of interest.” The driver back then and now is to unlock data, she said.
Around the time of Zooniverse’s founding, the game Angry Birds and other online games were becoming really popular, so researchers wondered if they could harness the power of the crowd to unlock their data: “These beautiful galaxy images were something that a kindergartener could say, that’s a spiral or an elliptical, or that one has this kind of structure versus that kind of structure. It’s just pattern matching, which human eyes are very good at,” Trouille said.
So, the researchers put the data online. In the first year of the online project, later named Galaxy Zoo, more than 100,000 people from around the world participated, contributing millions of classifications and providing a proof of concept showing that it is possible to get quality data from a crowd — and that people can also find satisfaction from it.
According to Trouille, “Part of the magic of Zooniverse [is] to have that partnership between a public-facing, public-serving institution [such as Adler] that has expertise in how to engage the public in science in meaningful ways and a research institution [such as Oxford] with all the reputational and expertise in research and data and publishing.”
The partnership between these two institutions reflects the dual purposes of Zooniverse, which is to value what the public gets out of the research experience and to ensure that the research is quality, real and authentic, she said.
Following the success of Galaxy Zoo, another study involving similar citizen science parameters also needed volunteers. That project, led by researchers at the University of Minnesota, had colllected a million camera trap images across the Serengeti desert in Tanzania. Researchers needed the public to look for animals in the images, so they put the images online.
That project processed a backlog of 18 months of data, garnering three million classifications over its first weekend.
“In each of these instances, it’s helpful to know the model. For Zooniverse, many eyes are on a single image or audio clip or video file, meaning it’s a consensus result from a crowd that we then use in the research,” Trouille said. About 97% of the time, the researchers found that the public could reach a consensus. But in the 3% where there was disagreement amongst the public, even the experts disagreed with what animal they were seeing.
Zooniverse’s evolution into a citizen science platform
When Zooniverse first started, researchers chose roughly two or three projects per year to build. But when Zooniverse launched in 2015 as a project builder platform, that allowed researchers to upload project details via a browser-based interface. As a result, the number of projects accessible via Zooniverse exploded.
There are now between 80 to 90 active projects on the Zooniverse platform, ranging from astronomy and ecology to biomedicine and the transcription of historical documents.
Since its 2007 inception, over 450 projects have found their home in Zooniverse, with 2.7 million participants from around the world registered on the platform. About half of the public registered on the site is based in the U.S., 15% are in the UK, and 35% are from the rest of the world, with high representation from India, Canada, Australia, and Germany. Zooniverse vets the projects and ensures that the projects are providing quality data to the public.
While Zooniverse’s staff come from three host institutions — Adler, Oxford and the University of Minnesota — there are more than 120 research institutions and government agencies such as zoos and museums that use Zooniverse’s tools for their research and public engagement.
“There’s been really neat examples across the different disciplines of these serendipitous discoveries because you have so many eyes on the data,” Trouille said. This happens in citizen science all the time, she said.
For example, the research team with the Snapshot Serengeti project pre-identified 60 animal species that the researchers wanted the public to mark and tag.
But “a neat, unexpected discovery in Snapshot Serengeti was that in a couple of images, some participants saw this little skunk-like creature that was not on the list. And it turns out it was a zorilla, which is an amazing name for an animal. It’s a little black-and-white furry creature,” Trouille continued. People working at the park thought that maybe they might have seen a couple of zorillas out of the corner of their eye, but the zorillas were never documented as living in the park.
“Zorillas have a particularly impactful role in predator-prey relationships, so this unexpected discovery of the zorilla through the Snapchat Serengeti project meant that that added a new element to the research and the models and a new aspect in the research that nobody had anticipated beforehand,” Trouille said.
The ability of the public to connect with researchers and each other through discussion forums has been beneficial in the Zooniverse projects. Trouille, who is also an astronomer involved with the Zooniverse Planet Hunters project says a community has arisen from the discovery of over 5,000 planets located around distant stars.
That human element complements any automated algorithms or machine learning algorithms that might also accompany the research. About a quarter of Zooniverse projects have machine learning integrated into the research.
“Some of our transcription projects are still growing in their sophistication and quality. We have a couple of projects where our participants transcribe a whole lot of documents, and then a team uses those documents to train a model,” Trouille said. Another transcription project has a machine model processing documents and then participants can edit the transcriptions.
“It’s a cutting-edge area of online citizen science or online participatory sciences platforms like Zooniverse: how do we best utilize the efficiency of machines while designing for good human experiences, keeping in mind the motivations and values of the human participants and making sure we’re not undercutting those by integrating machine learning alongside their efforts,” Trouille said.
In summary, online citizen science projects like the ones on Zooniverse not only benefit the general public who volunteers to work with scientists. Scientists can also find tremendous value in having fresh eyes look at the data and potentially find additional areas to research and observe.
—
What did the insights above inspire you to do? Let us know here: https://shorturl.at/cQi4E
—
Search for more insights on the CivicSciTV Network Terminal https://civicscitv.org
