Vetting Videos of War

In an emotional appeal last week, President Obama asked that every American view a video of Syrian children lying on a lacquered gray floor in Damascus. The clip shows more than 400 children in East Ghouta, arranged side by side as they convulse, choke, and eventually die. Videos like these, which the White House and its allies cited as evidence to support their argument for intervention in Syria, were not plucked at random from YouTube. They were vetted by a group called Storyful.

Truth, of course, is the first casualty of war—especially a two-year civil war, fought by guerrillas beyond the reach of most independent journalists. Amateur videos (purportedly captured by locals with smartphones) have poured into the information vacuum claiming to show atrocities or victories on both sides. To deal with the problem, Storyful relies on a network of unpaid fixers and experts on the ground in Syria to verify what's real. Then its staff editors and reporters—including old hands from the BBC and Reuters—compile and share the hand-curated clips with media and others.

In April 2010, Mark Little, a former Irish television presenter, launched Storyful with the help of venture-capital start-up money. He saw that as news organizations slashed budgets and laid off their staff, newsrooms increasingly lacked the ability to check on the massive amounts of social-media information. Storyful—in particular, its new crowdsourcing platform, Open Newsroom—would help the press sift through it all. Already, its clients include ABC, The New York Times, France 24, and YouTube, as well as more than 35 other newsrooms, according to news editor Malachy Browne. Unlike some other independent verification specialists, Storyful charges up to $25,000 for a year's access to its journalists and experts, who deliver information via a private Twitter account that has nearly 7,000 followers.

A relatively small number of moderators curate Open Newsroom's content. They include Browne; Iain Martin, a former editor at The Wall Street Journal Europe; and David Clinch, a former international editor of CNN, among others. Google Hangout connects seven offices from Hong Kong to San Francisco to the Dublin headquarters as editors review photos and videos from war zones like Syria, where both the rebels and the regime of Bashar al-Assad have incentives to tell their own versions of the story.

To check veracity, Storyful compares where a video was shot to where it was uploaded; surveys the past social Web behavior of the uploader; and looks for his or her official membership with an established organization. One of the biggest problems, explains Browne, is "scraping." Scraped videos have been copied several times over. The more a video is copied, the harder it is to find out who originally uploaded it, and where. So Storyful uses satellite imagery or Google maps to identify landmarks in a video—looking for a particular minaret or intersection, for example. A video taken in Saudi Arabia and uploaded in Syria, for example, is not considered reliable. 

Eliot Higgins, one of the more prolific independent verifiers—who contributes, gratis, to Open Newsroom—points to the BBC as an example of what can happen when social information isn't fact-checked. In May 2012, the network used a photograph (of a boy jumping over a row of bodies) provided by a Syrian activist who claimed it was from the massacre in Houla. In fact, it was taken in 2003 in a desert south of Baghdad. Twitter users began to take note, and the photo was removed after 90 minutes.

Storyful guards against errors like that by storing information about uploaders: where they live, which militia controls their town, their track record. It runs a large database of pre-vetted sources that it—and its clients—can trust.

One morning in late August, Browne's dashboard was full of massive streams of information: raw video footage from the Syrian massacre in Ghouta, anti-Morsi protests in Egypt, and tape from car bombs in Lebanon. Browne toggles between Facebook, Twitter, and YouTube, creating lists of previously vetted outlets and uploaders from different countries that give him an idea of what is happening in real-time.

On Aug. 22, a user called TheSyrianrev2011 uploaded a video to YouTube titled "[944]Syria, Ghouta: Child's Hysterical symptoms due to exposure to Gases 21/8/2103 [Multilingual]." It was one of the first claiming to document the chemical attacks in East Ghouta, and it quickly began to appear on social media. Browne suspected the video was a scrape. The first clue was that the video was described in both English and Arabic; normally only Arabic is used, and Storyful's multilingual staff translate with the assistance of Google. Browne then looked through the uploader's video history and established that most of the videos were branded with different logos and had varying quality, further discrediting the account. Finally, many of the uploader's videos were drawn from several locations on the same day. "An uploader of user-generated content," says Browne, "couldn't be in disparate locations on the same day."

The attack in East Ghouta, however, was a different kind of success for Storyful. Its techniques validated the video; its editors liaised with ABC News, and the verified clip played on Good Morning America. Ultimately, it became part of the president's argument.

CORRECTION: An earlier version of this story said Storyful shares information directly with governments; it does not. It primarily works with newsrooms.