We are still in the immediate wake of the Boston Marathon bombings, but it is already clear that the investigation into these attacks is taking a very different shape than the investigation into the 9/11 attacks. One of the big reasons, naturally, is the explosion of smartphone use in recent years, providing a wealth of user-captured photo and video for investigators to pore over.
Two days later, the FBI and Boston police do not seem to have many solid leads. We are only just now learning of some basic details about the explosive device that was used at the finish line – a pressure cooker-type bomb which is, apparently, a very common improvised explosive used around the world – and the container it might have been transported in – a black nylon bag or backpack. Investigators are urging the public to submit any and all information they may have from the marathon, leading to already a few terabytes of visual data that the FBI will begin to process.
But the 4chan community, which is not always known to be pro-social or classy, has organized an investigation of its own. Together, the 4chan members are combing through the images that have made their way onto social networking sites, annotating images with possible clues, mostly noting people wearing large black backpacks. It is a self-organized game of “Where’s Waldo?” and it shows that you can never underestimate the power of Linus’ Law: “given enough eyeballs, all bugs are shallow.”
By charging citizens with the task of finding and submitting all visual evidence from the marathon scene, investigators have effectively launched a crowdsourcing venture. Specifically, this kind of crowdsourcing falls into the “knowledge discovery and management” (KDM) type of crowdsourcing that I outline in my book.
Meanwhile, the 4chan community appears to have organized their own crowdsourcing venture on a “4chan Think Tank,” poring over the tons of visual clues available online and noting suspicious people and objects in photos. But this type of crowdsourcing is the “distributed human intelligence tasking” (DHIT) approach, a different kind of crowdsourcing process. Both the KDM and DHIT crowdsourcing types are suited for solving information management problems, while two other crowdsourcing types – the broadcast search approach and the peer-vetted creative production approach – are suited for solving ideation problems.
Here’s a modest proposal: stitch these two crowdsourcing processes together. As Spencer Ackerman noted in Wired yesterday, “the data used in the investigation will be crowdsourced. The investigation will not be.” But why not? Why can’t we blend the two processes – KDM by soliciting images from the public combined with a DHIT process of analyzing them? Ackerman notes that this could quickly turn into a witch hunt, with vigilantes jumping to conclusions based on loose leads turned up in the image annotation. He’s possibly right. But I think there is a way to avoid this, and that way involves investigators proactively taking charge of the kind of process 4chan is taking on themselves.
I am no security expert, nor do I have a deep knowledge of detective work, but as an outsider watching this unfold, I would propose that organizations like the FBI build dark site crowdsourcing platforms for crisis events, ready to spring into action as soon as attacks happen. This site would be a place to upload images, videos, and tips, which, perhaps with some moderation, would be posted publicly on the site. Volunteers would also have the opportunity to comb through the images and videos, tagging them with clues and notes, all of which would be viewable by investigators in real time.
It is a perfect crowdsourcing situation, following the best practices of what makes crowdsourcing work. You have a clearly defined problem (find out who may have placed the bombs). You have a clear way for the crowd to provide solutions (tag images and videos looking for large backpacks and people snapping photos on the ground that might clearly capture the faces of people carrying those backpacks). You have a highly motivated crowd (basically, the whole world wants to help) that has the expertise to do the job (anyone who can see can play a classic image-hunt game).
But what about the vigilante/witch hunt problem? Well, if an organization like the FBI was taking charge of the platform, it could partner with media to caution the public against using possible leads on the crowdsourcing site to act out in vigilante justice. It is no guarantee that people would not act out in this way, but, I would argue, it is better to have some leads, quickly question these leads (get to the people of interest before the crowd does), and caution the public from reacting than it is to have a complete absence of leads and have people’s ignorance and fear take over and start treating, say, all dark-skinned Arabic-speaking people as suspects or circulating other conspiracy theories and misinformation. We don’t need a post-9/11 xenophobic, racist reaction, and that is what will fill the void if a series of leads don’t come forward, I’m afraid.
So why not? Why could we not have a flexible platform ready to spring into action, tapping into a public that is eager to help in some way? The government should look to the power of online communities like 4chan and consider ways to press them into service in times of terror. There are some obvious concerns here about a surveillance state, privacy, and other issues, but, really, these things are already happening, and if we’re going to be on camera all the time, let’s at least use it for good.