The Geekdom of Crowds

The Obama administration experiments with data-driven democracy.

My favorite bar in Washington is the Raven Grill, a shoebox-shaped (and -sized) dive in the city’s Mount Pleasant neighborhood. The second half of the bar’s official name is a total lie, unless you count the overpriced beef jerky on the shelf next to the liquor bottles. But the beer is sold at prices otherwise unheard of in D.C., the north wall of the men’s bathroom used to feature some very nice anatomical artwork, and the jukebox has been strategically stocked to thwart even the worst taste (the only real hazard is Bob Marley’s “Legend”). Most crucially, the Raven is an easily walkable twelve blocks from my row house in neighboring Columbia Heights. The only real problem is that in the past year, those twelve blocks have been the site of seven robberies, eight assaults, and two homicides.

I know this because of a free, online computer application called Stumble Safely. The program plots recent criminal activity—what happened where, when—on a map of the neighborhoods surrounding popular nightspots in Northwest D.C., in a spattering of gray and red dots. Heading to or from the bar, I can find it on the map, select the relevant time of day, and look for the clusters of dots to see which corners might be worth avoiding. I can click on locations and get the police report of what exactly happened there: whether it was, say, a guy in a mask robbing a convenience store on 14th Street with a sawed-off shotgun (as happened around midnight on a Tuesday in January), or just someone’s cell phone getting snatched up the block (11 p.m., August 28, 14th and Columbia Road). In short, any D.C. resident with an iPhone and a signal can now carry more crime data in his or her pocket than beat cops in even well-wired police departments had in their squad cars fifteen years ago.

Stumble Safely was designed by Eric Gundersen, the president of a small D.C.-based tech consulting firm called Development Seed—and, as it happens, a former bartender at the Raven, where I met him for a beer on an evening in early June. Gundersen is a tech guy in the freewheeling Silicon Valley mold, a bearish twenty-nine-year-old with a scruffy beard who dresses like a bike messenger—when I met him he was locking up an immaculate new creamsicle-colored fixed-gear bike in front of the bar—and lives on a houseboat on the Potomac River, tied up at D.C.’s southwest waterfront. As we settled into a booth by the jukebox, Gundersen explained that Stumble Safely was more a demonstration project than anything else—he wanted to show people the power of well-arranged government data. Development Seed, which Gundersen had cofounded five years ago while he was working on a master’s degree in international development in Peru, had designed software and Web sites for the World Bank, the World Food Program, and the U.S. Agency for International Development, among others, building tools to map food aid deliveries in sub-Saharan Africa and flu pandemics in the developing world. “Maps forever have been about telling stories,” Gundersen said. “A lot of the stories we work with are very complex—information from thousands of school districts, or all the bird flu data in Southeast Asia. But we wanted to tell stories that people would understand.”

In October, Gundersen and his Development Seed colleagues entered a contest hosted by the district government, called Apps for Democracy. The contest was the brainchild of Vivek Kundra, then the district’s chief technology officer, and it was conceived in order to try out a theory that Kundra, a thirty-four-year-old veteran of the Northern Virginia IT community, had: government data were useful in the government’s own hands, but radically more so when people outside the bureaucracy had access to them. Advances in computing power, Internet connectivity, and open-source software development over recent years meant that more people were able to do more things with data; applications like Google and devices like the iPhone had increased the appetite among even the technologically semiliterate for the stuff.

Kundra figured there were probably plenty of civic-minded geeks in the district with the ability to provide information-based government services better than the government itself could, or come up with new ones the city bureaucrats had never even thought of. Why not open up all the information the city could to them and see what happened?

Kundra’s office made feeds of raw data—records of everything from trash collection to government employee purchases, provided in easily downloadable formats—available on the district’s Web site, and offered a prize to the programmer who made the most enterprising use of them. Forty-seven entries arrived by the November deadline; Development Seed’s Stumble Safely placed second, behind an online guide to historical walking tours. Other more substantive entries allowed users to track and discuss city procurement contracts and building permits. The district says that in all, the Apps for Democracy project alone saved the government some $2.3 million in software development.

That figure is pretty questionable—would the district really have paid someone to create Stumble Safely?—and Gundersen is the first to point out that the program Development Seed designed has its limits. “Most of these apps are bullshit,” he told me at the bar. “Including mine—it’s a bar-crawl site!” The point, he said, was to demonstrate the potential that was there: the opportunity for technology to transform the way citizens interact with their government. A simple application like Stumble Safely might allow a normally disengaged group of people—say, those of us who had chosen to be at the Raven on a weeknight instead of staying home to watch city council meetings on cable access—to visualize failures of public safety, and to ask questions about them. “The real benefits are going to come when community organizations get ahold of this [technology],” Gundersen said. “Can you imagine how cool this is going to be? Think of how politicians represent themselves and what they’ve done for their constituents.” He set down his bottle of High Life and gestured emphatically with his fist: “Fuck that. They’ll have to contend with apps that let people actually see exactlywhat they’ve done.”

This idea is often called Government 2.0: the notion that information technology can not only make it easier to get the information government already produces, but can be turned back onto the government itself to make the bureaucracy more responsive to its constituents, a generally faster and smarter beast. The idea has gained prominent champions, not least President Obama, who ran a tech-savvy campaign last year (his was the first in which IT staff sat in on strategy meetings) that also promised transparency—a pledge on which he has delivered in some, if not all, cases since taking office. But no one within the government is now more closely associated with the concept than Kundra, who Obama tapped as his chief information officer in March. In May, Kundra launched, a federal version of the online data clearinghouse he pioneered at the district, which is eventually supposed to catalog thousands of individual feeds of government information. The Sunlight Foundation, a nonprofit pushing the idea of using technology to foster transparency, is sponsoring Apps for America, a national version of the district-wide contest that produced Stumble Safely.

But a funny thing has happened since Eric Gundersen launched his application: Stumble Safely has become less useful, rather than more so. When you click on the gray and red crime-indicating dots that have appeared on the map in the past few months, you don’t get much information about what exactly happened—all you get is a terse, one-word description of the category of the incident (“assault,” or “theft”) and a time, with no details of whether it was a shootout or just a couple of kids punching each other in an alley.

This isn’t Gundersen’s fault—it’s the cops’. Because while Kundra and the open-data community were fans of opening up the city’s books, it turned out that the Metropolitan Police Department was not. Earlier this year, as apps like Stumble Safely grew in number and quality, the police stopped releasing the detailed incident reports—investigating officers’ write-ups of what happened—into the city’s data feed. The official reason for the change is concern over victims’ and suspects’ privacy. But considering that before the clampdown the reports were already being released with names and addresses redacted, it’s hard to believe that’s the real one. More likely, the idea of information traveling more or less unedited from cops’ keyboards to citizens’ computer screens made the brass skittish, and the department reacted the way bureaucracies usually do: it made public information harder to get. The imperatives of Government 2.0 were thwarted by the instincts of Government 1.0.

Whether it will always be thus is the question hanging over the whole open-data gambit: whether the past decade’s worth of technological advances are enough to overcome habits that are as old as government itself. “Vivek Kundra has his work cut out for him,” the Sunlight Foundation’s Clay Johnson says. “I have nothing but respect for what he’s trying to do. But it’s a hard job, and it’s going to take some time for this to actually happen right. I mean years.”

Talking with an IT manager who is bent on rewiring the government is a strange experience. When I spoke with Kundra on the phone recently about his plans for the Obama administration, he used the same vocabulary as the office tech-support guy who patiently guides me through toolbar submenus when I call about a problem with my desktop. The difference was that Kundra was walking me through how to debug a 233-year-old democracy, not a five-year-old Dell. “What we’re trying to accomplish is to fundamentally change the default setting in the public sector when it comes to information and transparency in general,” he explained. “We want to by default assume that data is going to be public.”

The decision to do this selectively in the past, Kundra argued, had paid dividends. When the National Institutes of Health’s Human Genome Project decided to make publicly available the sequences of human DNA it had decoded, it spawned a wealth of innovation in pharmaceuticals. “We’ve had an explosion of discovery, and a pipeline of drugs moving through the FDA, as a result of that decision,” Kundra said. In the 1980s, the Pentagon declassified a large portion of the data from its military satellites—”That gave birth to the GPS industry,” he said.

The ultimate goal of all this was to “change this old way of thinking about government and institutions, to recognize that we don’t have a monopoly on the best ideas,” Kundra continued. “The best way to do it is to say, ‘Look, here’s what the government is doing, here’s the data we have,’ and challenge the American people to help us deal with some of the toughest problems we face.”

Kundra seems a better fit for a big-thinking presidential administration than D.C. City Hall, which in general is less concerned with creating the government of the future than with simply putting space between Washington and Detroit in nationwide quality-of-life measures. The district was actually ahead of the curve on internal use of technology before he arrived in 2007, but Kundra’s decision to use the district’s IT office as a lab for data-driven civic participation led to some arched eyebrows—in a city where barely half the residents have home broadband connections, Kundra played unapologetically to the elite digerati, rather than the far larger group of people for whom XML sounds like a confusing shirt size. Most of the high-profile programs that Kundra instated, such as a wiki (a collectively editable Web page) designed to make city contracting more transparent, have been up and running for less than a year, so it’s too soon to judge their success. But the city council oversight staffers who spoke with me on background about the technology office—with whom the council has had a fractious relationship—weren’t holding their breath.

“It may be like a nightclub that’s popular—you get past the facade and it’s just not a big deal,” one of them says of Kundra’s efforts. “I’m not saying the guy doesn’t have some oomph—he may have some great ideas. But how much of it is fire and how much of it is smoke?” While Kundra was awaiting confirmation as Obama’s tech czar this spring, two employees in his former district agency—including his top deputy for security—were arrested by federal agents for an alleged city-contracting bribery scheme. Kundra hasn’t been implicated in the investigation, but the fact that it happened on his watch, while he was talking up plans to use technology to make contracting more transparent, suggests that maybe the district’s tech office needed a hard-nosed taskmaster more than a starry-eyed visionary.

The federal government, however, could use a starry-eyed visionary. For an organization that has always been largely concerned with gathering information, it is often shockingly bad at actually using it. Part of the problem is technical—only in recent years has the technology been available to inexpensively make use of vast stores of data. Rigid organizational structures also keep information siloed in discrete agencies, making it difficult or impossible to compare figures across bureaucratic boundaries. A more fundamental barrier has been the reflexive aversion everyone has to being personally held accountable for their actions—something which, in the world of government bureaucracies, often translates into functionaries protecting their turf by hoarding information about it. “It’s a basic human desire,” says Jerry Mechling, a professor at Harvard’s Kennedy School of Government who studies the intersection of government and technology. “It’s the instinct of a teenager: ‘Where are you going?’ ‘Out.’ ‘When will you be back?’ ‘Later.’ “

In the private sector, these tendencies are curbed by the need to compete—when the player with the best intelligence wins, you look at your numbers pretty closely. But governments don’t face the same kind of outside pressure, and thus have remained data-squandering dinosaurs, slowly masticating and barely digesting the information they collect. Incredibly, it was only in the early ’90s that the New York Police Department, weary of flailing blindly against ever-rising crime rates, began using crime statistics in a scientific fashion—a now widely emulated approach called computer statistics, or “compstat”—to deploy its resources more effectively. Washington’s police department didn’t adopt the strategy until five years ago.

The federal government has unique problems, too—there are whole categories of information it simply doesn’t have. Consider education, on which the federal government spends $67 billion each year—we’d want to keep track of that money, right? In theory, yes, but in practice, not so much: the Department of Education knows how much each state gets, but figures for individual school districts—that is, the information that could actually help you figure out if the money is going where it’s supposed to be going—are scattered across an array of state record-keeping systems that range from the excellent to (more frequently) the appalling. “That’s where the real breakdown happens,” says Jennifer Cohen, a policy analyst with the New America Foundation’s Education Policy Program. “A lot of times, I’ll get copies of dot-matrix printouts.”

There are plenty of people who would like to keep things this way, not only bureaucrats seeking to avoid scrutiny but also politicians, for whom the opacity of the system makes it easy to discreetly reward constituents and campaign contributors. But the issue isn’t only one of accountability—it’s also a matter of the government simply being able to do the things it’s supposed to do. When the Central Intelligence Agency famously failed to predict the fall of the Soviet Union, it was in part because the agency had kept classified basic agricultural and economic data it had collected, information that would have rendered the USSR’s incipient collapse potentially foreseeable to economists had they had access to the numbers.

But something unprecedented has happened in the past decade: greater computing power, better software tools, and the ever-extending reach of the Internet have all democratized the once-rarified field of data use. Making sense of huge piles of raw information used to require a degree in computer science, a university lab mainframe’s worth of circuits, and an awful lot of time. Now all it takes is an Internet connection and the ability to type in “Google.” Once-expensive tools are basically free, and often better than their pricey predecessors. It used to be that if you wanted financial intelligence, you had to pay for the services of a ratings agency like Moody’s, where analysts made sense of the data tapes gathered in person from the Securities and Exchange Commission. Now you can get a comparable analysis at, a site launched by a pair of amateur programmers after last fall’s financial meltdown (which Moody’s failed to anticipate), where open-source software digests the same—now easily downloadable—SEC data.

Data enthusiasts like Cohen and Gundersen see in this changed information landscape the potential to do something big: leverage the interest of a newly enlightened public with ever-improving technological tools at its disposal to shine a giant spotlight on the inner workings of the government. Eventually we could see more clearly not just what the government is doing, but also how well what it’s doing is actually working—for instance, tracking education reforms through a matrix of crime and employment data to gauge the rippling effects of policy decisions in something like real time, and adjusting them accordingly. In April, Kundra explained the concept in a Senate subcommittee hearing, comparing the government to a precision-guided missile: “One of the reasons those missiles actually hit their targets is because you get a constant feedback, a loop-back mechanism that lets you know how you are performing in relation to where you are,” he told the panel.

Kundra’s efforts to put these ambitions into practice, however, have thus far mostly thrown into sharp relief just how far the federal government has to go to meet them., the stimulus-money-tracking Web site that Kundra helped conceive as a member of Obama’s transition team, was supposed to herald a new era in accountability. Instead, its attempts to track the biggest burst of government spending in history have been slow moving and incomplete, and hamstrung by classic Government 1.0 problems. Some agencies have been reluctant to cooperate, and the fact that the federal government doesn’t keep tabs on states’ spending on things like education and highway improvements means that it simply doesn’t have much of the information it needs to make the site useful. “You end up with an almost intractable problem once those dollars leave the Beltway,” says Eric Gillespie, senior vice president of the Seattle-based data research firm Onvia, which employs a staff of hundreds to pore over local government records and newspaper notices for that kind of information. “If you’re counting on some sort of persistent or permanent source of data, it just doesn’t exist.” (To prove this point, Onvia has launched a site called, which uses its own store of hard-won data to provide a lot of the information has promised but failed to deliver.)

The initial launch of, Kundra’s bigger bid for bureaucratic transformation, in May was similarly underwhelming. When the site debuted, it featured forty-seven data feeds. But they included very little of the data that anyone interested in transparency would really want—half of the feeds were from the U.S. Geological Survey. “The top data source is on the world’s copper smelters,” says the Sunlight Foundation’s Johnson, “which isn’t going to tell us very much about what’s going on inside of our government.” Once again, as if to prove Kundra’s point that citizens are often more inventive with data than governments, when the official site debuted it fell short of a similar amateur site,, a government data catalog that a programmer in New Jersey named Robert Loftin built in his spare time by e-mailing and calling government agencies one by one and asking for data feeds. ( has since expanded its offerings modestly, and when this issue went to press Kundra’s office was expecting to add significantly more resources to the site in coming weeks.)

While open government advocates and data geeks are critical of the performance of these projects so far, they give Kundra’s goals almost universally high marks. “I think what they’re trying to do is quite admirable,” Loftin says. “Unless you’re a fan of waste, fraud, and abuse, it’s absolutely the right thing to do,” Gillespie says. And these efforts tend to have rocky rollouts, for the simple reason that you often don’t realize what information you’re missing until you start asking for it. In 1986, after a pair of high-profile fatal chemical plant mishaps in India and the United States, Congress ordered the Environmental Protection Agency to create a database that kept track of toxic chemicals emitted by heavy industrial operations in the U.S. Chemical companies predictably balked, and at first the data were shoddy and full of holes. But as the system expanded over time, so did the quality of the companies’ monitoring. In the end, the database became valuable to industry itself—companies could see how they stacked up against their competitors, and promote their efficiency—and spawned innovations in pollution control. “It was a mobilization by the public that led to better data,” Mechling says. “A little bit of data leads to better data, and then you’re onto a positive spiral.”

In other words, the government and its citizens have to learn—as the Obama campaign did—to see information technology less as a service to be provided, someone who fixes BlackBerries and fine-tunes spam filters, and more as a process, a sort of techno-civic game of Marco Polo. “Look at the Gallup polls—the number of people who believe the government is doing the right thing most of the time has historically fallen from about 70 percent to about 15 percent,” Mechling says. “There has been a huge erosion of trust. The idea that says, ‘Government has all this data, let’s just make that available, and show them we have nothing to hide,’ that’s the big thing. [It] is potentially a game changer. The game has yet to be played.”