Semantic Werks

I attended an interesting talk last night by David Eaves, expert on social media and government (Government 2.0).

In general I thought he made some excellent points, but I wanted to push back on the long tail portion of his presentation. His point was that the old model positioned the government official as the only source of knowledge on a topic. His example was health policy and Health Canada. The individual bureaucrat knows a lot (head of the power-law curve), but that knowledge is at best equal to the remainder of the curve. In other words, collectively Canadians know as much as the expert does.

I’m not so sure. I think there are two reasons. One is immediacy and contextual knowledge, the other is tied to expertise.

Firstly, as David himself acknowledges, it’s one thing to have that knowledge distributed among Canadians. But it is another thing to make that knowledge accessible. And if the knowledge isn’t available when needed, then it might as well not exist. Admittedly, this can be improved with cognitive support tools like wikis.

My bigger problem is the dismissal of expertise. I’ll reference Malcolm Gladwell’s book Outliers, in which he mentions the popular notion that to be good at something requires 10,000 hours of practice. More importantly, 10,000 people with 1 hour of practice does not equal 1 person with 10,000 hours of practice. But the former is exactly what crowd-sourcing suggests we do.

For some problems, I agree, you don’t need that much expertise. So we can get by asking the retired fisherman what he thinks of the weather recently, or the person who worked on NAFTA what a EU-Canada free trade deal should be. But for most problems, I don’t see expertise as additive.

With the climate change issue, I think it is clear that a majority of people hold opinions which are at odds with the best understanding we have of the issue. So I also don’t think the Long Tail is continuous. It is entirely possible that the experts know the reality, while the ‘great unwashed’ are hoodwinked by ad campaigns, incompetent reporting, or poor communication from the experts. The flip side is also possible, of course – that the experts are horribly uninformed (although not in the case of climate science!).

I also am unconvinced by the Mozilla example. Popular models of open-source software development frequently mention the bazaar approach, and I’m sure this is what David means. But studies on large-scale open-source software (Eclipse, Linux, Apache) show that often most progress is made by the 10-12 people at the heart of the project. My own research bears this out. There is a lot of irrelevant and distracting noise for these projects.

Tangentially, I’m not a believer in the “Given enough eyeballs, all bugs are shallow” mantra, either. First of all, because it is a maxim, and not at all tested empirically. More importantly, it depends on what kind of eyeballs, doesn’t it? Most users of Firefox, for example, have no expertise in diagnosing errors in Javascript, let alone C++ internals. Their bug reports tend to be on the most obvious bugs. It takes experts to find the source of the more obscure (more important?) bugs.

David also illustrated his point about crowd-sourcing by mentioning the alternative to the space shuttle, DIRECT. His point was that these people were able to design a large spacecraft without any institutional assistance. Totally agree about the design, but two points. One, these people were most certainly experts. In fact, they were all NASA employees. Secondly, there is a vast difference between building the spacecraft and designing it. Spacecraft take millions, if not billions, of dollars to build. That tools have reduced the cost for entry into the design phase does not make it any cheaper to produce. I still see that as a public good which no company has incentive to assume.

This all sounds a bit cynical, but I do agree with David that there is a lot of expertise that ought to be leveraged (by government) that isn’t. And I think he is right to focus (at least at first) on improving government’s internal knowledge management processes.

Via Kottke’s year-end list:

• a fascinating article on language and culture. The subject of the article is the Pirahã, an Amazonian tribe who apparently have no words for past events, don’t use recursive sentence structures (the man with the red bicycle went to school), and no system for counting numerically. Their language seems to contradict the Chomskyian universal grammar, which argues that all human beings have an innate capacity for recursive sentences. Two issues stood out for me: one, how can we know what a language is, if the only teachers might not want us to understand?  Secondly, what role did the disease epidemics of the post-Contact world play? For example, perhaps this culture was quite successful, but horribly vulnerable to smallpox. The disease wipes out most of the elders and culture of the group, leaving only young people and an aversion to recalling the past. There are scholars who believe upwards of 90% of New World people were killed. Wouldn’t this type of group trauma have a profound impact on culture and language? The only thing I was disappointed by in the article was the lack of insight into the Pirahã as a people independent of the people studying them. The journalist writing the piece seemed more interested in the academic questions of language than the people themselves.
• Malcolm Gladwell on race and DNA. He reviews a book which argues that IQ scores have been steadily increasing since the test was first introduced. This by way of establishing that the notion that IQ scores compared across generations or cultural groups are pretty meaningless. For example, the test that Chinese immigrants took that showed their average IQ to be 10% higher than white Americans was easier than the comparison group.  I don’t think any of this was a terrible surprise for me; I’ve never understood what IQ tests were designed to measure. Things like charisma, social intelligence, emotional intelligence, numeracy, problem solving skills and so on always seemed difficult to measure with what seems to amount to a Tetris skill test. I think most of the foofarah around IQ tests comes from the old “correlation is not causation” error.