In the fall of 2002, the Bush administration announced Total Information Awareness, a massive effort to build a counterterrorism database that can help track the activities of "people loosely organized in shadowy networks." The name was changed to Terrorism Information Awareness after critics charged that gathering gigabytes of data on private citizens is a threat to individual freedoms. But meanwhile the domestic surveillance program continues. So MIT graduate student Ryan McKinley decided to do something—beat the government at its own game by watching the watchers.
On this July 4, after six months of programming, McKinley went live with a Web site that inverts the idea of Total Information Awareness. It's called Government Information Awareness (www.opengov.us). The premise is a simple but powerful one: If our elected officials use the latest information technologies to follow our every move and detect unusual patterns in our behavior, perhaps we should subject them to the same intense scrutiny.
The Internet is already teeming with personal sites known as blogs that explore every nook and cranny of political news and opinion, ranging from insightful analysis to conspiracy theories. Blogs helped stir up controversies that recently led to the fall of two of the most powerful figures in America: Trent Lott stepped down as U.S. Senate majority leader because of his indiscreet pro-segregation remarks, and Howell Raines was forced to resign as executive editor of The New York Times because of his mishandling of a plagiarism scandal. But there isn't one overarching site that pulls it all together, detecting interesting patterns or connections in the political data stream. There's a vibrant marketplace of political ideas circulating through the Web, but there's no Dow Jones or Nasdaq to help you see the big picture.
Illustration by Leo Espinosa
McKinley's project is designed to fill that void. It takes publicly available information pulled from a variety of official sources—government Web sites, telephone directories, C-SPAN—and integrates it into a single database. At first glance, the site's interface has an almost stark simplicity: four icons representing the three branches of government, plus industry. Click on the legislative branch—an image of the Capitol building—and you're taken to a directory that lists the major congressional institutions, including the House, the Senate, the Congressional Budget Office, the General Accounting Office, the Government Printing Office, and the Library of Congress. Nothing much yet.
Click on "U.S. Senate" and you see a list of all 100 senators. Still boring. But follow the trail to a page on Hillary Clinton and the data start to get interesting. You'll find everything from her religion (United Methodist) to a list of her top corporate campaign donors (for example, Goldman Sachs). Click on the former and you'll see a list of all active politicians self-identified as Methodist; click on the latter and you'll see a list of candidates supported by Goldman Sachs, ranked by dollar amount.
There's a rich ecosystem of political connection lying in McKinley's database. As you move through the space, you're bombarded by factoids at nearly every turn. For example, former Republican leader Lott received more money from the insurance industry than from any other sector, but the insurance industry itself gave more than four times as much money to its favorite politician, the liberal Senator Charles Schumer. That particular infobite took only five clicks of the mouse to uncover. Most of that information probably exists online in other formats, but before Government Information Awareness came along to connect the dots, the data were buried in obscure disclosure documents or scattered across the Web on dozens of unconnected sites. McKinley's site makes exploring the tangle of relationships that constitute our government as easy as flipping through the Yellow Pages—and much more productive.
McKinley is developing special tools for tracking information as it's added to the database—not unlike the stock-tracking services offered by some financial sites. Eventually, you'll be able to tag anyone or anything—Tom DeLay, the Congressional Budget Office, the Air Force's secret shuttle program—and you'll get an e-mail alert if any new information becomes attached to the item. Government Information Awareness will also track activity on C-SPAN broadcasts and offer an alert if one of the officials you're monitoring appears.
The surge of traffic to Government Information Awareness was so great after its launch that features were disabled for a week, and the entire site went dark for a few days as McKinley rejiggered the code to handle all the visitors. Like a government official himself, he puts a positive spin on the blackout: "It's nice to finally be working on something that people are actually interested in."
And over time, user contributions will expand what kind of information the site can track. Everything in the database is either an "entity" or a "link," in McKinley's words. Hillary Clinton is an entity; the contribution from Goldman Sachs is an entity; and Goldman Sachs itself is an entity. Each has been linked to the others in the database. Expanding that portrait is as simple as creating a new entity and linking it to an existing one. Chris Csikszentmihályi, head of the Computing Culture group at MIT and McKinley's thesis adviser, describes the site's architecture as "a backbone with lots of nerves going off in all directions." Many of those nerves will take you outside the government itself. "While we have the three-branches structure, it's the little alleys and eddies that you can get to that will be particularly interesting," Csikszentmihályi says. "That's where a lot of decision making and authority happens."
Government Information Awareness uses image and text recognition software to follow activity on C-SPAN. A citizen can make a standing request—"tell me when Donald Rumsfeld appears" or "tell me when the word 'abortion' is spoken"—and receive an automated e-mail alert.
For the initial launch, McKinley hand-seeded the site with data culled from public databases. But Government Information Awareness has been designed to accept contributions from ordinary users, including Crossfire junkies, activists, and armchair chiefs of staff. These users are collectively trying to solve an interface problem that the government itself has generally been colossally bad at: figuring out a way to make all this publicly available information useful to ordinary citizens. The site has already been flooded with contributions, ranging from telephone directories for obscure subdepartments in the District of Columbia bureaucracy to elaborate charts documenting the connections between the Mormon church and the judiciary, though reprogramming the site has thus far kept McKinley from adding submissions. "Both useful and useless information is flowing in," Csikszentmihályi says with a smile.
Separating the useful from the useless is the tricky part of maintaining a user-authored site. McKinley specifically modeled the Government Information Awareness system after eBay's decentralized approach to commerce, in which users police themselves by rating each other's reliability each time they conduct a transaction with another user. Instead of giving people high marks for sending a package on time and low marks for bouncing a check, Government Information Awareness users will rate contributions for the quality of the information provided.
"Every piece of information is annotated with information about who entered it," McKinley explains. "Everything can be ranked for veracity and interest. The hope is with that you can come up with pretty simple rules to keep out the things that would make this kind of database unusable. I honestly believe that there's enough genuine interest out there for people to put time into this, and so the amount of garbage that goes in will be pretty small." If the approach works, the garbage will be quickly spotted and demoted by the users of the site, while quality information will rise to the surface. High-rated contributors might rise up as well.
The self-policing of the site makes for an arresting sociological experiment. Most of the time, the trust ratings of eBay manage to block an individual's financial incentive to misrepresent information online. Will a similar approach work when the incentive is ideological instead of financial? Systems that self-regulate by following predictable rules can be gamed, of course. A group of users could agree to rate their own contributions highly and downgrade everyone else's. If they reached a critical mass, they could significantly alter the perspective of the database. It would take a lot of work—and a lot spare time—but ordinary people have done a lot of extraordinary things on the Web in their spare time.
The ultimate hope with a site that accepts user contributions is that the self-regulation will create a reliable body of information, and not a partisan free-for-all where every crackpot theory gets as much attention as genuine facts. But McKinley knows that at a certain point he won't be in control anymore. If a political faction attempts a hostile takeover of the Government Information Awareness site, he says he'll enjoy watching: "I actually find that just as interesting."
GIA was inspired by the official Terrorist Information Awareness Web site. Learn about the program and its objectives at www.darpa.mil/iao/TIASystems.htm.