Bottom-up

It’s a few weeks old, but Matt Welch has a great example of the biases you find when mainstream journalists analyze the current trends in news gathering. He describes a conversation with Alex Jones, (who I critiqued here):

After Jones had gone on about how the “iron core” of news is shrinking, thereby imperiling our democracy, I pointed out that in a blog post I had written just that morning about a flurry of last-minute laws signed by California Gov. Arnold Schwarzenegger, the links I found (thanks to Google News) included MTV.com, a boating newsletter, and a hip-hop website. Instead of even entertaining the notion that the “iron core” had been increased on net by widespread sources, Jones was horrified. Why, how did I know that HipHopPress.com was a reliable source! Shame on the mainstream media for not covering each and every law!

In an argument like this, people like Welch are placed at a structural disadvantage because the emerging bottom-up structure of the news business is much harder to catalog than the top-down model it was replacing. Alex Jones can easily explain how the LA Times will cover California politics: they hire reporters and order them to Sacramento. He can say exactly how many reporters there will be, how much they’ll cost, how many stories they’re write, and so forth.

In contrast, Welch has to explain how a decentralized media ecosystem containing thousands of outlets like HipHopPress.com and MTV.com might, in the aggregate, provide more thorough coverage than the LA Times. This is quite a bit more difficult, because there’s no one offering a formal guarantee that any stories will be produced, and because the Internet news ecosystem is vastly more diverse, with many more contributors and many more types of contributors.

Yet the reality is that there are, in fact, a huge number of small sites covering California politics (and lots of other topics) that are the traditional domain of large newspapers. The fact that the blogosphere is hard to explain doesn’t mean it doesn’t work. After all, the Wikipedia process is much harder to explain than the Britannica editing process, yet it seems to work pretty well.

Yesterday I promised to post about why we shouldn’t view the decline of the newspaper as a blow to democratic accountability. Conor Friedersdorf offered what seems to be the mainstream view of the subject in the nation’s journalism schools: that one of the crucial roles of a newspaper is to put important stories in front of readers who may or may not be interested in reading them, and that we’re losing that capacity as the Internet replaces the newspaper.

This doesn’t seem quite right to me for two reasons. First, the I think it overestimates newspapers’ power to compel involvement in the political process by the otherwise apathetic. And second, I think the Internet is better at creating effective “soapboxes” than he gives it credit for.

Newspapers are fundamentally information networks, and part of the reporter’s job was to ration access to the network’s scarce and valuable bandwidth. There has always been vastly more civic information in a metropolitan area than there were column-inches, and this meant a lot of relevant information would wind up on the cutting-room floor. The reporter (and his editor) were responsible for ensuring that the most valuable information reached the readership.

As Clay Shirky has observed, traditional publishing involved a top-down “filter, then publish” model, in which some authority figure (in this case, a reporter) had to guess which information people would be interested in. This is the only way a newspaper can work, because it doesn’t have the bandwidth to publish everything. In contrast, the Internet operates on a “publish, then filter” model: everything goes up, and then we use after-the-fact tools—search engines, Twitter, blogs, Wikipedia—to find good stuff. No rationing is required. Not surprisingly, journalists are ambivalent about this because it removes them from their privileged gatekeeping position. But for readers, this is a feature, not a bug.

Now, a newspaper partisan might argue that reporters’ gatekeeper role was a good thing because, in essence, the reporter had a better understanding of what the reader needed than the reader himself did. I think this is not only paternalistic, but it also misunderstands the role of the front page in local politics. The reason putting something on the front page of the local newspaper could affect the political debate wasn’t that it mobilized otherwise-disengaged citizens. If you put a water board story on the front page, most ordinary citizens are just going to flip to the sports page. Rather, the reason getting on the front page mattered was mostly because before the Internet, there were a lot of people who were interested, but who—thanks to the limitations of the newspaper format—might not otherwise have seen this particular story.

It used to be pretty difficult for insiders to keep track of the issues they card about. Before search engines and RSS feeds, it took a lot of work to comb through newspapers looking for every story relating to a particular subject. This is why large organizations subscribed (and still do) to expensive newspaper-clipping services. The Internet has made life radically easier for the politically active. If I want to know what’s going on in a particular issue I care about—software patents or eminent domain reform, say—I can directly subscribe to RSS feeds related to those subjects. I can set up a Google News alert for my favorite topic, my favorite (or least favorite) politician, or the name of my home town. So the Internet is having two effects that are really opposite sides of a single coin: it’s making it easier for interested citizens to follow the issues they care about, and it’s making it easier for uninterested citizens to ignore the issues they don’t care about. Given that non-engaged citizens have always been little more than deadweight in the political process, I consider this a good thing.

With that said, Conor does have a good point about newspaper front pages: it’s a good thing that there was a tall soapbox that allowed newspaper editors to shine a “spotlight” on an issue that required broader public attention. But here, too, “publish, then filter” is an improvement, not a step backwards. As Clay Shirky pointed out in his discussion of the Catholic pedophilia scandal, different stories have different target audience. The audience for the Cardinal Law story wasn’t Bostonians (the readership of the Globe) but Catholics regardless of location. The Internet allowed the audience for the Globe‘s Cardinal Law stories to be much larger than its general readership.

Moreover, there are institutions with the “heft” to bring stories to the attention of large audiences. In the “publish, then filter” world, the sites with the most influence tend to be filters rather than publishers. And they frequently aren’t organized around geographical locations. Julian mentioned Daily Kos, but there are a ton of other examples: the Drudge Report, Slashdot, Digg, BoingBoing, and so forth. These sites build audiences that don’t necessarily break down along geographical lines, but rather (in this case) along ideological or professional lines. And of course the blogosphere and social news sites in the aggregate act as a news amplifier; stories that people find interesting get passed around rapidly, eventually reaching far more people than any one mainstream media outlet could have reached.

Which isn’t to say that the Internet doesn’t do local news. The reality is that a great many people still read newspapers, so it shouldn’t surprise us that the local metro daily is still one of the most prominent soapboxes in a given metropolitan area. But we’ve seen a proliferation of locally-focused blogs and news sites that help to disseminate news to civically-minded people in a metropolitan area. To choose a few DC-area blogs at random, you’ve got DCist, Greater Greater Washington, Beyond DC and Prince of Petworth. Obviously, none of these blogs individually are comparable to the Washington Post, and even in the aggregate they may not match the Post‘s readership yet. But as the Post declines, it’s not hard to imagine these sites, and others like them, picking up the slack.

Things look even better if we look at the hyper-local level. Virtually every neighborhood has a neighborhood mailing list for exchanging local community information. Here the reporter-middleman is cut out of the loop entirely: people who care about local politics communicate directly with one another. A neighborhood mailing list is going to have vastly more information than would be available in a local newspaper a generation ago. Not everyone will have the patience for that, but the ones who matter most will. And the rest will take their cues from their politically-engaged friends and neighbors.

Reporters no longer have a technologically-imposed monopoly on their readers’ attention. Readers are no longer stuck with whichever local newspaper happens to be in their home town, and so reporters have to work harder to keep the audiences that used to be theirs by default. Not surprisingly, established journalists don’t like this trend. They liked the privileged position they enjoyed in the pre-Internet age, and they built an elaborate self-justifying ideology that portrayed their privileged position as a benefit to readers. It’s a letdown for journalists to suddenly find themselves on a level playing field with hordes of amateurs. But frankly, that’s just the world works: if a bunch of amateurs can do your job as well as you can, then you should probably find a new job.

Which isn’t to say that journalism as a profession is going to go away. There are some categories of reporting that professionals can do better than amateurs; reporters will have to figure out which categories these are and how to earn a living from them. They’ll have to work hard to earn their readers’ attention. And that’s as it should be.

Two of my friends, Julian Sanchez and Conor Friedersdorf, did a BloggingHeads last week about the future of media. It’s a great show in which they make a number of solid points. However, I found myself disagreeing with the argument Conor makes here:

He says:

The very fact that a lot of people are seeing it on the front page because the editor decided to see it there because he thought as an insider that it’s important makes the people at the water board kind of sit up and take notice in a way that maybe wouldn’t happen if a blogger broke it. I saw Clay Shirky making the same argument about the pedophilia scandal in the Catholic Church. An independent paper in Boston broke all of these stories, but no one really took notice until it was on the front page of the Boston Globe. It isn’t just that the story get reported. It’s that there’s a form with enough institutional power that it gets seen by enough people that actors are forced to reform things.

Conor is talking about the story Shirky tells in Chapter 6 of Here Comes Everybody, about the pedophilia accusations against Father Geoghan. Shirky told the same story at this event. The point of Shirky’s story is not that coverage on the front page of the Boston Globe is the key to getting people to sit up and take notice. It was quite the opposite: that by itself, coverage by the Boston Globe didn’t mobilize the kind of strong reaction it takes to prevent future pedophilia cases. We know this because a virtually identical episode occurred in 1992:

There is, eerily, something vanishingly close to a two-slide comparison here. In 1992, a priest named Paul Shanley was pulled in for having raped or molested almost a hundred boys in the Archdiocese of Massachusetts. His bishop was also [Bernard] Cardinal Law, and the group covering it was also The Boston Globe. And they ran 50 stories that year on the priest abuse. And that story went nowhere. It shocked people, people were horrified, they were upset, and then it died out. And in the intervening decade, Geoghan kept after it.

We can’t say that if the web had been in wide circulation in ‘92, that the Stanley case would have created the reaction to Geoghan case. But what we can say is that many of the good effects in limiting the Catholic Church’s ability to continue doing this were a result of the public reuse of the documents in ways that were simply not possible in 1992 and had become not just available, but trivial by 2000.

The key difference was that in 1992, rank and file Catholics had no efficient way to organize themselves outside of the church hierarchy. This gave the church a key strategic advantage, and it allowed them to contain the outrage over any given pedophilia scandal to a single parish or (at worst) metropolitan area. By 2002, widespread access to the Internet meant that it was vastly easier for the laity to quickly create an independent organization with sufficient numbers to pose a real challenge to the church leadership. It tooks just weeks for Voice of the Faithful to assemble online and make themselves into a force the Catholic hierarchy had to reckon with.

It’s understandable that Conor would have missed the point of this particular Clay Shirky story. But I think the reporter-centric view of the political process Conor is articulating here is flawed in ways that go beyond this particular example. To be sure, newspapers are an important part of the political process, and will likely continue to be so for many years to come. But as I’ll explain in my next post, it’s a mistake to view the decline of the newspaper’s top-down reporting model as a blow to democratic accountability. The reality, I think, is something close to the opposite: the Internet is removing a key information bottleneck and making it much easier for ordinary citizens to keep their elected officials accountable.

I’ve done a few posts about the problems with Apple’s App Store, and its requirement that developers use it to get software onto the iPhone. Paul Graham has a more thorough and better-written critique:

The way Apple runs the App Store has harmed their reputation with programmers more than anything else they’ve ever done. Their reputation with programmers used to be great. It used to be the most common complaint you heard about Apple was that their fans admired them too uncritically. The App Store has changed that. Now a lot of programmers have started to see Apple as evil…

Software development has evolved from a small number of big releases to a constant stream of small ones. But Apple doesn’t understand that either. Their model of product development derives from hardware. They work on something till they think it’s finished, then they release it. You have to do that with hardware, but because software is so easy to change, its design can benefit from evolution. The standard way to develop applications now is to launch fast and iterate. Which means it’s a disaster to have long, random delays each time you release a new version.

Apparently Apple’s attitude is that developers should be more careful when they submit a new version to the App Store. They would say that. But powerful as they are, they’re not powerful enough to turn back the evolution of technology. Programmers don’t use launch-fast-and-iterate out of laziness. They use it because it yields the best results. By obstructing that process, Apple is making them do bad work, and programmers hate that as much as Apple would.

How would Apple like it if when they discovered a serious bug in OS X, instead of releasing a software update immediately, they had to submit their code to an intermediary who sat on it for a month and then rejected it because it contained an icon they didn’t like?

Read the whole thing.

For years, the Associated Press has been complaining about websites that quote what it regards as excessive quantities of AP stories (although most of these quotes are likely legal under copyright’s fair use doctrine). In July, the Associated Press made an announcement that utterly mystified the world’s computer geeks. It consisted of a press release and a ridiculous graphic supposedly explaining the technology:

As my advisor Ed Felten explained at the time, no plausible technology could do the things the graphic claims the technology can do. Plain text is an open format, and every browser on the market enables effortless cutting and pasting of text content. No consumer is going to agree to run a browser that disables cutting and pasting. It turns out that the technology AP was actually releasing was a microformat called “hNews” that allows news stories to be tagged with metadata that describes the ownership and permissions associated with content. The hNews format can encode certain ownership information, but it doesn’t include any enforcement mechanism. The technology does not provide any “protection” from unauthorized copying, nor is there a “tracking beacon” that “sends signals” to anyone. As Ed puts it, “if there’s a story here, it’s in the mismatch between the modest and reasonable underlying technology, and AP’s grandiose claims for it.”

The interesting question is: how did this happen? No competent engineer would have created the graphic above. Indeed, few competent engineers would even been able to keep a straight face if shown that graphic. And it’s a safe assumption that the AP has at least a few competent engineers on staff. So how did this graphic get released?

I think this is another example of the problems I’ve been talking about the last few days: poorly chosen abstractions and a managerial structure that stovepipes decision-making and insulates senior decision-makers from bad news. I don’t have any inside information about the AP’s decision making process, but I’m willing to bet that what happened was that senior management asked their engineers to come up with a technology to help “protect” their news. The engineers realized that what management wanted wasn’t possible, but they didn’t want to tell their bosses that. So they gave their bosses hNews, which doesn’t do all the magical things described in the graphic above, but is a small step in that general direction. At this point I imagine management handed the project off to some marketing flacks who were tasked with taking the engineers’ dry technical specification and making it sexy. And since the marketing flacks didn’t know that the technology management wanted was impossible, they created the clip-art-and-bullet-point monstrosity you see above.

The AP is not the first incumbent content organization to be suckered into betting on copy protection technology that doesn’t work. Indeed, this is a pattern that has been repeated over and over in content industries. The music industry tried to create a DRM-based music store with PressPlay and later with the “new” Napster and Microsoft’s comically-named PlaysForSure format. It also introduced DRM on CDs that wound up behaving like spyware on users’ computers. Hollywood, for its part, has had a string of hopeless DRM-based movie services, including MovieLink and CinemaNow. Even the “successful” DRM schemes on DVDs and Blu-Ray discs are only successful in the sense that they didn’t destroy the viability of the underlying format. No one seriously believes these formats prevent piracy.

The fundamental problem, as computer scientists have understood for at least a decade, is that the basic assumptions of digital rights management are wrong. Steve Jobs put it well back in 2003:

When we first went to talk to these record companies — you know, it was a while ago. It took us 18 months. And at first we said: None of this technology that you’re talking about’s gonna work. We have Ph.D.’s here, that know the stuff cold, and we don’t believe it’s possible to protect digital content. What’s new is this amazingly efficient distribution system for stolen property called the Internet — and no one’s gonna shut down the Internet. And it only takes one stolen copy to be on the Internet. And the way we expressed it to them is: Pick one lock — open every door. It only takes one person to pick a lock. Worst case: Somebody just takes the analog outputs of their CD player and rerecords it — puts it on the Internet. You’ll never stop that. So what you have to do is compete with it.

People tend to forget that the iPod predated the iTunes store by about 18 months. Apple reportedly refused to include digital rights management technology in the first iPod because they didn’t want the resulting compatibility problems to hurt iPod sales. Ultimately, though, the labels persisted and Jobs relented in order to get the iTunes Store off the ground. The iTunes DRM not only failed to stop piracy, it also dramatically increased Apple’s power in the music industry.

The problems I’ve been writing about in the last few posts—leaky abstractions and hierarchies that suppress bad news—are exacerbated when the people at the top of the hierarchy are known to prefer a particular outcome. In late 2002 and early 2003, everyone knew that George Bush and Dick Cheney wanted to find evidence of Weapons of Mass Destruction in Iraq. This had a pernicious effect on the entire intelligence community, because every intelligence officer knew it would be better for their careers if they found evidence of WMD than if they didn’t. The recording industry faced a similar situation earlier this decade. Label executives wanted very badly to believe that Internet piracy could be stopped. So people who offered plausible-sounding anti-piracy solutions within the organization got more traction within the organization than people pointing out that these “solutions” were snake oil.

The graphic above has the same basic defect as the PowerPoint slide I featured in my previous post: it’s wishful thinking unmoored from the “facts on the ground.” Insulated from reality by layer upon layer of bureaucracy, the key decision makers marched confidently off a cliff in the fashion of Wile E. Coyote. Only months or years later did they look down and realize that they were no longer standing on solid ground.

If you want to study the deficiencies of large, hierarchical organizations, a good case study is the most powerful hierarchical organization on Earth: the United States military. Our military has long exemplified the dysfunctions of large, hierarchical organization (see, for example, the phrase SNAFU, the American GI’s most eloquent expression of bottom-up thinking). Journalist Thomas Ricks wrote an excellent book about one of our military’s more serious mistakes: the invasion of Iraq. Not surprisngly, PowerPoint played a role in this disaster too:

[Army Lt. General David] McKiernan had another, smaller but nagging issue: He couldn’t get Franks to issue clear orders that stated explicitly what he wanted done, how he wanted to do it, and why. Rather, Franks passed along PowerPoint briefing slides that he had shown to Rumsfeld: “It’s quite frustrating the way this works, but the way we do things nowadays is combatant commanders brief their products in PowerPoint up in Washington to OSD and Secretary of Defense…In lieu of an order, or a frag [fragmentary order], or plan, you get a bunch of PowerPoint slides…[T]hat is frustrating, because nobody wants to plan against PowerPoint slides.”

That reliance on slides rather than formal written orders seemed to some military professionals to capture the essence of Rumsfeld’s amateurish approach to war planning. “Here may be the clearest manifestation of OSD’s contempt for the accumulated wisdom of the military profession and of the assumption among forward thinkers that technology—above all information technology—has rendered obsolete the conventions traditionally governing the preparation and conduct of war,” commented retired Army Col. Andrew Bacevich, a former commander of an armored cavalry regiment. “To imagine that PowerPoint slides can substitute for such means is really the height of recklessness.” It was like telling an automobile mechanic to use a manufacturer’s glossy sales brochure to figure out how to repair an engine.

This classic Arms and Influence post points to the PowerPoint slide above, which is reproduced in Ricks’s books from a military presentation on reconstruction planning. It’s a particularly egregious example of PowerPoint thinking. This slide seems to gesture in the general direction of a strategy, but it’s obviously not a coherent plan for reconstruction. One hopes that the slide made more sense to people who heard the accompanying talk, but even then it seems unlikely that it would have had the kind of rigor you’d want for an effective military campaign.

One reaction to a story like this is to just say that Franks and Rumsfeld were bad managers. And they probably were. But Franks and Rumsfeld are hardly the only managers to give their subordinates vague or non-sensical orders.

There’s a parallel between the job of managers and the job of computer programmers. Managers, like programmers, are supposed to control systems (of people rather than silicon) that are far too complicated to understand in their full detail. And they use the same basic strategy that programmers do: abstraction. Just as we give our computers high level instructions (“fetch this website”, “open this file”) and let our software take care of all the messy details, so do managers give their subordinates high-level instructions (“capture this village,” “build these airplanes”) and let their subordinates take care of the details.

And like programmers, managers are often brought low by the problem of leaky abstractions. A strategy that seems plausible when illustrated on a high-level PowerPoint slide might be revealed as tragically inadequate when it comes into contact with reality. Programmers have the crucial advantage that they usually have access to the complete system their abstractions represent. This means that they can take advantage of the tight feedback loops provided by fail-fast design: you run your program, and if it crashes you know you still have work to do. It’s much harder for managers to do this kind of rigorous testing. This is why businesses build prototypes and release products in test markets. It’s why the military spends so much time war-gaming.

These strategies are far from perfect. Even the most elaborate wargame is itself a leaky abstraction that will fail to capture all of the nuances of a real war. And managers, unlike programmers, have a serious problem of biased reporting. If a programmer gives a computer instructions that don’t make sense, the computer isn’t shy about pointing that out. In contrast, if a manager gives instructions that don’t make sense (“Find evidence of WMD in Iraq”), the people under him will often bend over backwards to avoid telling their boss unpleasant truths.

Good managers understand the danger of their subordinates telling them telling them what they want to hear. Joel Spolsky has a good story from his days at Microsoft when he had a meeting with Bill Gates. “Bill doesn’t really want to review your spec, he just wants to make sure you’ve got it under control. His standard M.O. is to ask harder and harder questions until you admit that you don’t know, and then he can yell at you for being unprepared.” It’s a little bit cruel, perhaps, but it’s the kind of thing you need to do to successfully manage a large organization.

And of course most managers aren’t as good as Bill Gates. Typical managers are prone to deal in high-level abstractions that have only a tenuous connection to reality. And sometimes the result is like a cartoon character who careens off a cliff but hasn’t yet noticed the 1000-foot drop below him: everything seems fine until the situation gets so bad that your subordinates can no longer sugar-coat it.

The larger your organization is, the more gut-wrenching the subsequent fall will be. If you’ve only got a couple of layers of management below you, it’s relatively easy to stay in touch with the “facts on the ground” and catch problems early. If you’ve got a dozen layers of bureaucracy below you, like Donald Rumsfeld, it requires a Herculean effort to cut through layer upon layer of half-truths to find out what’s really going on at the bottom of the hierarchy.

Joel Spolsky is the author of Joel on Software, a popular blog about computer programming. He wrote a great post in 2002 about what he called “leaky abstractions.” Programmers think about abstraction a lot because it’s indispensable for managing the complexity of modern computers. Abstraction is what allows us to think about “the Internet” rather than a collection of millions of routers, switches, and servers and “files” rather than strings of ones and zeros etched onto the magnetic platters on our hard drives. Abstraction allows both users and programmers to ignore the mind-boggling complexity under the hoods of our computers, and instead deal with a sparse, clean, intuitive interfaces.

The problem, Spolsky says, is that abstractions are leaky. he uses the TCP half of the Internet’s basic TCP/IP protocols, to illustrate the problem:

Earlier for the sake of simplicity I told a little fib, and some of you have steam coming out of your ears by now because this fib is driving you crazy. I said that TCP guarantees that your message will arrive. It doesn’t, actually. If your pet snake has chewed through the network cable leading to your computer, and no IP packets can get through, then TCP can’t do anything about it and your message doesn’t arrive. If you were curt with the system administrators in your company and they punished you by plugging you into an overloaded hub, only some of your IP packets will get through, and TCP will work, but everything will be really slow.

This is what I call a leaky abstraction. TCP attempts to provide a complete abstraction of an underlying unreliable network, but sometimes, the network leaks through the abstraction and you feel the things that the abstraction can’t quite protect you from.

One of the most important differences between a novice programmer and an experienced one is that the seasoned programmers understand the leakiness of the abstractions they use. They understand what’s going on “under the hood,” and this gives them a sense for how to use the tools they’re given effectively—and how to recover gracefully if they break. Novice programmers are like novice drivers in a stickshift car: because they don’t understand how the transmission works, they grind the gears. And novice programmers are blindsided when the abstraction breaks and they’re suddenly exposed to the messy underpinnings of the software stack.

One of the ways programmers deal with leaky abstraction is with fail-fast design. Fail-fast design says that when a component in a complex system encounters an error, it should immediately report the error and shut down, rather than trying to recover from the error and continuing. Fail-fast design helps users manage leaky abstractions, because it gives the user clear and immediate feedback that they’re using the abstraction wrong. This should prompt the programmer to read the documentation more thoroughly, change the way he’s using the abstraction, choose a different abstraction more suitable to the task, etc. It’s frustrating to develop software for a system that’s not built in a fail-fast fashion because a programmer might sink a ton of time into a program built on a poorly-chosen abstraction before problems crop up.

Fail-fast is a useful strategy in fields far removed from software development. Here is an essay about how important it is for entrepreneurs to fail fast. And, of course, failing fast is closely related to failing cheaply, a concept I’ve discussed here in the past.

In 2003, the world watched, shocked, as the Space Shuttle Columbia exploded as it was re-entering the Earth’s atmosphere. NASA investigated and concluded that Columbia exploded because during take-off, a bit of foam fell off the shuttle and hit the wing. The shuttle made it to orbit safely, but on re-entry the damage from the foam was enough to destroy the vehicle.

The astonishing thing about the Columbia tragedy is that NASA engineers not only noticed the foam in the take-off video, they even realized that it could cause problems on re-entry. Indeed, while the space shuttle was still in orbit, NASA hired a Boeing team to conduct an analysis of the potential damage. Tragically, after hearing the Boeing team’s report, NASA decided to go ahead with the re-entry.

Edward Tufte is a legendary statistician and graphic designer whose research has focused on the challenges of clearly and accurately conveying complex empirical information. In an analysis of the Columbia disaster, he blames PowerPoint for the shuttle’s explosion. Specifically, he argues that when the Boeing team presented its findings to senior NASA officials, the limitations of the PowerPoint format became an impediment to clear communication:

In the reports, every single text-slide uses bullet-outlines with 4 to 6 levels of hierarchy. Then another multi-level list, another bureaucracy of bullets, starts afresh for a new slide. How is it that each elaborate architecture of thought always fits exactly on one slide? The rigid slide-by-slide hierarchies, indifferent to content, slice and dice the evidence into arbitrary compartments, producing an anti-narrative with choppy continuity. Medieval in its preoccupation with hierarchical distinctions, the PowerPoint format signals every bullet’s status in 4 or 5 different simultaneous ways: by the order in sequence, extent of indent, size of bullet, style of bullet, and size of type associated with various bullets. This is a lot of insecure format for a simple engineering problem.

The format reflects a common conceptual error in analytic design: information architectures mimic the hierarchical structure of large bureaucracies pitching the information. Conway’s Law again. In their report, the Columbia Accident Investigation Board found that the distinctive cognitive style of PowerPoint reinforced the hierarchical filtering and biases of the NASA bureacracy during the curicla period when the Columbia was damaged but still functioning.

Tufte then quotes from the CAIB report:

The Mission Management Team Chair’s position in the hierarchy governed what information she would or would not receive. Information was lost as it traveled up the hierarchy. A demoralized Debris Assessment Team did not include a slide about the need for better imagery in their presentation to the Mission Evaluation Room. The presentation included the Crater analysis, which they reported as incomplete and uncertain. However, the Mission Evaluation Room manager perceived the Boeing analysis as rigorous and quantitative. The choice of headings, arrangement of information, and size of bullets on the key chart served to highlight what management already believed. The uncertainties and assumptions that signaled danger dropped out of the information chain when the Mission Evaluation Room manager condensed the Debris Assessment Team’s formal presentation to an informal verbal brief at the Mission Management Team meeting.

While Tufte blames PowerPoint, it’s clear that what he’s talking about is fundamentally a management problem. It’s true that PowerPoint bullets are a lousy way to communicate complex technical information. But it’s also not a coincidence that middle managers love them so much. A big part of the job of a middle manager is act as a kind of information funnel: to gather information from dozens or hundreds of people below him in the hierarchy and communicate it in condensed form to the people above. The beauty of bullet point is that it makes it possible to present complex information in arbitrarily truncated form while glossing over details that the presenter may not fully understand.

This “information funnel” function is essential to the role Paul Graham describes managers playing in hierarchical organizations: to make a big group of people appear to upper management as if it were a single person. A given manager might be an MBA who’s never computed a rocket trajectory in his life, but when there’s a meeting of the organization’s senior staff, he’s a stand-in for the 100 aerospace engineers who report to him. Not only is it unreasonable to expect him to accurately represent the knowledge and concerns of all 100 people, it’s not even reasonable to expect him to understand the views of all those people. Yet it would be awkward for the manager to admit that he has only a vague idea of what a lot of the people under him do. So he’s going to gravitate toward a presentation style that allows him to sound authoritative while summarizing information he may or may not understand in any detail.

A related phenomenon is a tendency toward excessive optimism within large bureaucracies. People dislike giving their bosses bad news. Given that managers have to summarize and abbreviate the information their employees give them anyway, there’s going to be a natural tendency to pay less attention to negative results than positive ones. And this tendency gets amplified when there are multiple layers of reporting: At each layer, bad news gets pruned more aggressively than the good news, and so the glasses get more and more rose-tinted. The information that bubbles up to the top of the hierarchy can be massively skewed and incomplete.

Unfortunately, reality doesn’t respect organizational hierarchies, so when the people under you think your space shuttle is going to explode, it’s pretty important that you hear about it. There are a number of worthwhile strategies for mitigating this kind of problem. Tufte advocates that technical information be communicated in the form of written reports rather than PowerPoint bullets. But the more fundamental point is that we need to recognize that these kinds of information-flow problems are inherent to hierarchical organizations. Even in the best-run large organizations, the people at the top levels of management are going to have a partial and distorted view of what their employees know and what the organization is doing.

What Matt said:

I sort of wish we called our November 11 observance Armistice Day like they do in other countries.

Something that I think is missing from American political culture is the thing that in Europe is taken to be the lesson of World War One, namely that a war can be bad for reasons other than it being lost. France and Britain were ultimately victorious in the war, but it was ruinous nonetheless. What was needed from the political leadership of the time was a way to avoid the war, not a way to win it. In America, though, evaluation of military endeavors is ruthlessly governed by considerations of efficacy. To lose a war, like in Vietnam, is a bad thing. But there seems to be a growing conventional wisdom that the surge has somehow redeemed Iraq and that the only thing we’re allowed to talk about with regard to Afghanistan is whether we can or will “win.”

This is a tricky business because the qualities of leadership needed to win a war are almost diametrically opposed to the qualities needed to make good foreign policy decisions. On the one hand, winning wars requires convincing soldiers and citizens that they’re in a titanic struggle between good and evil. Ambivalent soldiers are never going to be effective war-fighters. And so militaries inculcate cultures of obedience and discipline, in which not only specific orders but also beliefs and attitudes flow down the chain of command. If the president of the United States decides that the mission of the military is to stop Naziism in Europe, communism in Asia, or terrorism in the Middle East, then rank-and-file soldiers are expected to wholeheartedly support the shared objective. And this bleeds over into civilian society. Millions of Americans put “support our troops” ribbons on their cars, and politicians accuse one another of failing to sufficiently support our troops.

On the other hand, it’s tremendously important that we not get ourselves involved in any more unnecessary or counterproductive wars like the ones the United States entered in 1917, 1964, and 2003. And all the habits of thought that promote effective war-fighting make it harder to avoid getting into unnecessary wars in the first place. Pursuing a sane foreign policy requires precisely the kind of nuance and ambivalence that militaries try so hard to discourage within their ranks. It’s important that our policymakers be capable of drawing fine distinctions and to re-think old decisions in light of new evidence. They need to understand (for example) that Al Qaeda, Hamas, Hezbollah, the Baath party, the Iranian Mullahs, and the Taliban are independent entities with different and sometimes contradictory goals. It requires us to trade off different objectives, to realize that (for example) that we need to give up on fighting opium production in Afghanistan if we want to win the broader battle for the hearts and minds of the Afghan people. And above all, it requires us to remember that all wars, even “successful” ones, are tremendously costly. Yet once the build-up to war begins, pundits and politicians alike face tremendous pressures not to dwell on these kinds of problems for fear of demoralizing our troops and undermining the war effort.

In my view, the second objective of avoiding unnecessary is vastly more urgent for the contemporary United States. As the most powerful nation on earth, there are no serious threats to our territorial integrity. But we’ve squandered trillions of dollars, gotten thousands of Americans killed, and upended the lives of millions of foreigners in wars whose connection to national defense is tenuous at best.

The problem is that it’s hard to simultaneously honor the sacrifices of our veterans and have an honest discussion about whether those sacrifices were actually necessary. I have an uncle who served in Iraq, and I certainly wouldn’t want to tell him that the world would be better off had he not gone over there. But if it’s true, and I think it might be, we shouldn’t avoid saying so simply because it might hurt some soldiers’ feelings. Saving future soldiers’ lives is vastly more important than sparing past soldiers’ feelings.

More good stuff from John Quiggin and Jacob Levy.

On Monday, the Supreme Court heard oral arguments in the Bilski case, which I’ve written about before. My write-up for Ars Technica is here. At issue in the case is which inventions are eligible for patent protection. The case is nominally about a method for hedging against risk, but the ruling is expected to have broad applicability for “business method” and possibly software patents. Software patents played a surprisingly large role in the argument. If you’ll forgive me for quoting myself:

Stewart [representing the government] argued that applying the machine-or-transformation test [for patentability] would not have changed the outcome of State Street, the 1998 Federal Circuit decision that opened the floodgates to both software and business method patents. This is surprising because State Street was the decision that introduced the Federal Circuit’s now discredited “useful, concrete and tangible result” test. But Stewart contended that the patent in that case, which involved using a software algorithm to manage mutual funds, would be approved under the new “machine or transformation” test the government is now defending.

That provoked a strong response from Justice Stevens, the court’s senior associate justice and its most fervent software patent critic. He noted that the patent in the State Street case merely took a standard computer and installed new software on it. “You can’t say that’s a new machine,” he said. Stewart insisted that installing a sufficiently innovative program on a computer could transform it into a new machine for purposes of patent law.

But Justice Breyer objected that this reasoning would undermine the government’s goal of limiting business method patents. “All you do is you get somebody who knows computers, and you turn every business patent into a setting of switches on the machine because there are no businesses that don’t use those machines.”

Justice Roberts asked about another hypothetical software case: whether you could patent the process of using a calculator to compute “the historical averages of oil consumption over a certain period and divide it by 2.” Stewart responded by drawing a distinction between a calculator with “preexisting functionality” to “crunch numbers” and a computer that “will be programmed with new software” and “given functionality it didn’t have before.” We’ll let readers judge for themselves whether this distinction makes any sense.

One of the things I love about writing for Ars is they let us mix a bit of editorializing in with our straight news coverage. But but I’d like to expand on that last comment. The Supreme Court said 37 years ago that mathematical algorithms are not eligible for patent protection. Ever since then, patent lawyers have bent themselves into pretzels trying to describe software in ways that evades this rule. They have been helped tremendously by the fact that most lawyers and judges have only the vaguest idea of how software works. So they can rely on distinctions that make intuitive sense to legal professionals without worrying too much if these distinctions make any sense in the real world.

In this case, Stewart seems to be suggesting that there are “mathematical” algorithms, and then there is software that’s something more than mathematics. I’ve never seen anyone give a coherent explanation of this distinction, but I think it relies on a basic sense that Newton’s method (say) can’t possibly be in the same category as Microsoft Word. The former is clearly a mathematical algorithm, while the latter is… something else.

This distinction simply doesn’t stand up to close scrutiny. The legendary computer scientist Donald Knuth explains the point well:

I am told that the courts are trying to make a distinction between mathematical algorithms and nonmathematical algorithms. To a computer scientist, this makes no sense, because every algorithm is as mathematical as anything could be. An algorithm is an abstract concept unrelated to physical laws of the universe.

If you examine Microsoft Word carefully, what you’ll find is that it’s nothing more than pile of mathematical algorithm encoded in a machine-readable format. To be sure, it’s an extraordinarily large pile of algorithms, and it performs some extremely useful functions. Moreover, Microsoft has cleverly engineered things so that the process doesn’t look mathematical to the average user. But the fact that the mathematics are “under the hood” doesn’t mean they aren’t there.

Stewart claimed that loading software on a computer transforms it into a new machine by giving a computer “functionality it didn’t have before.” This argument doesn’t withstand close scrutiny. Obviously, it’s true that loading software on a computer gives it functionality it didn’t have before. But this is little different than saying that setting my alarm clock causes it to perform a function—waking me up at a particular time—that it wouldn’t have done otherwise. We don’t say a set alarm clock is a different machine than an unset alarm clock. It’s the same machine with different settings. Programming a computer is exactly like setting an alarm clock except that the computer can handle dramatically more complex instructions. If setting your alarm clock doesn’t create a new machine, then neither does installing Microsoft Word on your computer.