Crowd Sleuthing: Harnessing the Power of Crowds

Aug 24, 2017

Immediately after the nationalist rally in Charlottesville that turned violent, amateur sleuths went online to help identify participants.  Sometimes, though, the crowd gets it wrong.  A new project by computer scientists at Virginia Tech aims at separating truth from fiction in a time when misinformation is rampant.

Credit Idea Scale

It happened after Charlottesville, and after the Boston Marathon bombing in 2013. A social media savvy public, eager to help in the search for the culprits, identified the wrong person. This kind of Internet fueled attack is known as doxxing – for the online documents people mine for private information about someone else without consent. And it’s becoming a powerful force for quickly spreading information, whether it’s done properly or not.

“We see kind of an unorganized group of amateurs or crowd workers trying to investigate some sort of current event.  There’s a lot of risk there because these folks are not professionals.

The risks are that they could be investigating something that really shouldn’t be investigated in a public forum," says Kurt Luther, a computer science professor at Virginia Tech.

The National Science Foundation recently gave him an Early Career Development Award to study and improve the capabilities of crowdsourced investigations. And like a good academic, he began with a kind of control group, using crowd sourcing tools to identify people and places from the past.

“One of the reasons we’re starting with history is, if all the people we’re trying to identify have been dead for a long time, the risks of potentially identifying the wrong person might be a little bit lower.”

The approach is more than what we know as crowd sourcing.  Luther calls it “Crowd Sleuthing” and it works in tandem with its big brother in a related aspect of the project called “Ground Truth.”

“So, with 'Ground Truth, we’re trying to identify the locations where photos or videos were taken; Their locations somewhere on earth as specifically as possible.  And this is really useful when we’re trying to identify if images or video posted on social media, for example, are what people are or claiming they are. And if we can figure out where those photos actually came from, we can determine if those photos are, as they are presented.”

And that’s what makes this model of crowd sourcing different from the ad hoc hunts that spring up after sensational events.

“The key innovation is trying to bring the expert investigator and amateur crowd together.”

There are only so many experts and their time is limited.

“The crowd has a lot of person power and a lot of diverse knowledge and potentially a lot of time to contribute to the investigations, but they benefit from oversight and the knowledge of professional standards and good journalism practices.”

This combination of Ground Truthing and Crowd Sleuthing  would begin in a semi closed forum at first, to avoid the problems that can arise in over zealous crowds.

“Where, if different people are brought under suspicion, those judgments can be explored without potentially blaming the wrong people or having names leaked out before the investigation is concluded.”

Luther and his colleagues are still in the early stages of the project. Right now, they’re using paid crowd workers to test ideas and develop new iterations quickly as they finalize the platform. The plan is ultimately, to enlist volunteers such as journalism students, and members of other communities who like to engage in these kinds of investigations. They’re also looking for experts from all disciplines to join the group effort to support and enhance a social media born activity that’s clearly here to stay.