SANTA CRUZ — If you could contribute to research every time you unlocked your smartphone, would you?
A team of students and computer scientists at UC Santa Cruz and Stanford University is hoping you will. They’ve created an Android application called Twitch that replaces the gesture that smartphone users perform all the time — swiping a finger across the bottom of the screen to wake it up — with snappy tasks and questions that provide useful information to researchers.
The app is designed to pioneer a new kind of “crowdsourcing,” a concept that involves asking a bunch of people to contribute to a particular task or project by tapping their collective intelligence. Wikipedia’s dominance as a publicly contributed trove of online information is one example of crowdsourcing’s power.
But existing crowdsourcing technologies take time, said Rajan Vaish, a soft-spoken computer science doctoral student at UC Santa Cruz who is leading the project. “You have to go to your computer or application, browse to a website, log in and wait for it to load,” he said. And non-Internet forms of crowdsourcing, such as knocking on doors or catching people in a good mood for a phone survey, take even longer.
Enter Twitch.
“We want people to be able to contribute to crowdsourcing in one- to two-second bursts — like when you’re waiting for the elevator or waiting for a meeting to start,” said Michael Bernstein, an assistant professor of computer science at Stanford University and a leading crowdsourcing researcher who is advising the team.
While Vaish says the app is still in its early development, he said he hopes Twitch will eventually give researchers deeper insights into issues such as pollution and mental illness. If the app can distinguish whether someone is traveling by train, car, bike or foot, it can assess that person’s carbon footprint. If many users tell the app that they feel sad around the holidays, researchers can further understand seasonal mood disorders. That data can then be aggregated into a Google Map and viewed by the public in real time.
“We’re trying to capture the pulse of human activity,” Bernstein said.
Each time a Twitch user unlocks a smartphone, he or she is asked questions from different categories of tasks.
One category, called Census, asks users about their surroundings and their mood. Questions such as: How many people are around you? What are they wearing? What are you doing? How are you feeling?
“You can see how people feel during exam time on the UCSC campus, for example,” Vaish said. “This data can be used by the administration to decrease the stress level of students.” Nap pods can be installed in libraries, Vaish said, or student counselors and advisers could be told to spend more time with students.
Information source
The data can also provide users with information that may affect their commute, social activities and jobs. For instance, if a woman waiting in line for an Americano at her favorite cafe reaches for her phone to pass the time, the app may ask her how many people are nearby. If she selects the box suggesting that it’s crowded, that information can be uploaded to the app’s Google Maps page that can tell someone else whether that location is a good spot to grab a quick cup of coffee.
Another category, called Photo Ranking, presents a group of images relating to a theme — nature panoramas, for example — and asks the user to select the most appealing image. The selected images are filed into larger databases relating to that theme, which helps computer recognition software classify them. If a computer needs to recognize a photo of flowers, for instance, the application weeds out the best images of flowers for the computer to use.
Last spring, Bernstein had approached Vaish with his idea for the new app. They, along with Stanford student Brandon Cheung, massaged the concept into an actual application. Shortly after, Keith Wyngarden, a computer science undergraduate at Stanford, linked and compiled all of the data the app was producing and polished the final version of the app.
But first, they needed to ensure that users could perform meaningful tasks without becoming distracted. They also need to confirm that the tasks took no more mental energy and time than it takes to perform a normal “slide-to-unlock” gesture on an Android phone.
In the summer, the team asked a group of people to use their phones while walking. Users were shown a sequence of letters on the screen that were randomly interrupted by Twitch tasks. After completing the task, participants were asked to recall previous letters.
It takes just seconds
It turned out that users remembered the letters correctly, which was surprising to Vaish. They also found that most could complete Twitch tasks within one to three seconds, consistent with the time it normally takes to unlock an Android phone. The application was no more distracting and time-consuming than a normal “slide-to-unlock” gesture.
Asked by a reporter to test the application, a small group of Android owners had mixed reactions.
Jeffrey P. Bigham, an associate professor at the Human-Computer Interaction Institute at Carnegie Mellon University, said he was excited about the app’s utility: “Twitch effortlessly crowdsources a particular kind of expertise that has been difficult to get — that is, expertise from people in a particular place.”
Cassie Davenport, a teacher in Watsonville in the Pajaro Valley Unified School District, said she found the application easy to use and not distracting. But the app’s unclear research goals made her uneasy: “If I knew what the purpose was, I would be more inclined to use it. But because I don’t fully understand why I am answering these questions, I don’t know if I would keep (the application installed) on my phone.”
And therein lies the conflict: In order for researchers to gain information about people’s lives, they have to ask them. And some people aren’t willing to blindly give up that information without knowing exactly how it will be used.
As of now, the app is still young. The developers are working on the second version. Eventually, they hope Twitch will contribute to larger goals, such as helping visually impaired users by labeling images of their surroundings.
The app is currently available only on Android smartphones, and can be accessed via Google Play. The team eventually hopes it can offer the app on the iPhone, but for now Apple’s tight controls on its unlocking screen are preventing the team from doing that.Still, the researchers hope the app has the potential to make big impacts — with the simple swipe of a finger.