Infovark

Nate Nash at e2.oh has an interesting blog post about folksonomy, Facebook and metadata extraction. He wonders whether automatic indexing tools might a better way to get accurate productivity information about employees, rather than relying on performance reviews. Nate touches on an idea that Gordon and I have been kicking around for a while.

In psychology, it’s called the self-reporting bias. Most people like to think well of themselves. It’s such a common tendency that people who think badly of themselves are usually thought of as depressed or maladjusted. Having low self-esteem has been blamed for all sorts of social ills.

Our brains are wired to err on the positive side. If you asked people, they’d tell you that they donated more to charity than they actually did, or worked out at the gym more, or engaged in more positive activities and less negative activities than was actually the case. In psychology experiments, you have to plan carefully to eliminate or minimize the self-reporting bias if you want to get at the motivations and causes of people’s behavior.

The classic business-school example of the self-reporting bias is the Nielsen Ratings for television. The early Nielsen Ratings were based solely on surveys, and showed that Americans liked high-minded fare such as Masterpiece Theater. Later, Nielsen incorporated set-top boxes into the surveys, to report the station and the time whenever the television was turned on. Unsurprisingly, given human nature, sports, sitcoms and shows like American Idol began appearing in the top ten.

The self-reporting bias is a core part of what it means to be human. It plagues social software like Facebook and LinkedIn. It shows up in the business world as a tendency to inflate one’s own resume or to indulge in marketing hype. To counteract the self-reporting bias, you’ve got to develop independent measures — or independent measurers. You’ve got to watch what people do and not listen to what they say they do.

(If you’d like to take a good objective look at your own cognitive biases, wikipedia has a fascinating list on the subject.)

April Fools Day has come and gone, and the Internet has largely returned to normal.

Yeah, the gags are pretty annoying. But they’re also kind of fun, right?

Yesterday’s antics led Dean to point out that April Fools Day would freak out the semantic web and emerging services based on it, like Calais and Twine.

As an example, out of no other reason than pure mischief, I mailed some of my friends a link to a video which I claimed was a leaked internal video analysis of HP’s recent acquisition of our former employer, TOWER Software. In my email, I harped on at length about some completely made up rubbish about strategy and future direction. Of course, the accompanying link was, in fact, to this video.

I know, I’m terribly original and downright hilarious, but back to the point:

Without the April Fool’s Day context, a careful semantic analysis of my emailed rickroll might permanently associate HP, TOWER Software, Strategy and Bad 80’s pop music. Or it might indicate that I was related to HP in some way (which I’m not). Regardless of how effective or capable any semantic engine is, any meaning that could possibly be extracted from my joke would be largely false, with the possible exception of the close — but now perhaps a bit strained — relationships between me and my friends.

No amount of metadata, microformats or markup can save the computers from human exaggeration, humor, or outright lies. And we humans have institutionalized a day where that’s all we do.