Human Computing and the Gamification of Surveillance Analysis

Recently unveiled surveillance blimp; courtesy of Raytheon, via Slate

Since the US invasion of Afghanistan in 2001, the American military has worked to create a system of virtually continual real-time drone surveillance of the entire country. The system is not entirely automatic, however: in 2010, Marine Corps General James E. Cartwright noted that at least 19 analysts were needed to process video feeds from a single Predator drone. Looking through thousands of hours of collected video and audio recordings is particularly difficult. Cartwright described the work of analysis as sitting for hours watching "Death TV," searching for single or valid targets, an activity he called "a waste of manpower [and] inefficient."

To combat this inefficiency, researchers have experimented with building smarter cameras, capable of recognizing and reporting on suspicious activity, but the development of information gathering technology continues to far outpace the ability of computers to make sense of what has been collected. As an alternative, organizations have experimented with crowd sourcing the work of analysis to online volunteers; the US Air Force even asked ESPN for help looking through the footage. But what happens when the work becomes play, and the people involved don't know they're working as surveillance analysts?

courtesy of NASA

In 2000, NASA began outsourcing the tedious job of identifying craters on the moon and Mars by encouraging pubic volunteers, nicknamed "clickworkers," to identify craters in photographs posted online. What would have taken a graduate student a year to accomplish was completed in only a week. In 2006, the state of Texas installed webcams along the Mexico border, streamed the feeds online, and encouraged the public to help monitor them for suspicious activities. One woman watching at 3:00 AM noticed someone signaling a pickup truck on the webcam and notified the police, which led to a high speed chase and the seizure of over 400 pounds of marijuana. Following the 2011 riots in London, police asked the public to look through thousands hours of CCTV footage and submit their own photos and videos to identify individuals who had participated in looting. Recently, a start-up in the UK began offering a service called "Internet Eyes," which connects the country's ubiquitous CCTVs to the Internet and offers the public rewards for identifying people committing crimes.

Important to note is that crowdsourced surveillance efforts don't necessarily lead to results: following the December 2012 shootings at Sandy Hook Elementary School, police asked for assistance from online crowds and were led to the wrong person. Following the April 2013 Boston Marathon bombings, a similar call for crowd assistance interfered with investigations and led to the wrongful accusation of several innocent people.

While many of these projects use crowdsourced volunteers to handle tasks computers are not able to do, the volunteers participating are typically aware of how their work is being used. These projects attract volunteers willing to give up a little of their time to help with a project they're interested in seeing succeed or help catch someone suspected of wrongdoing. In contrast, the next generation of surveillance analysis doesn't require volunteers to know who they're working for or even that they're working.