The game theory of Google Knol - rich writers, poor content.

When I first looked at Google Knol, I felt a pay-the-writer-via-Ad-clicks was the wrong game theory / biz model for an information service. At the time I reckoned the following would happen with the system as it currently is:

- people will write about topics that are highly trafficked regardless of cultural value
- they will write content that aims to drive traffic - ie it will be sensationalist, not factual
- because anyone can put up a soapbox, it will be far harder to discern fact from fiction, opinion from reality.
- voting and other feedback will be largely irrelevant as people vote for what they like, not what is correct or valuable

Add to this is another one which in hindsight is obvious - given that the money comes from traffic driven advertising, why bother to keep content current and accurate at all, in fact why bother with your own content when you can just lift it from Wikipedia and make money yourself?

As the Technologiser blog notes when they looked at the copied, pasted and non-current data on Knol article they surveyed:

Much of the power of Wikipedia, of course, comes from its collaborative nature. And within moments of news breaking such as McCain picking Palin, you can be sure that someone will add it to the appropriate Wikipedia entries*. When someone makes a mistake in a Wikipedia piece–and it happens all the time–there’s a good chance someone else will come along and fix it.

Knol is fundamentally different: It’s designed to hold entries written by individuals. “[No] one else can edit your knol (unless you permit it) or mandate how you write about a topic,” states the Knol entry about Knol. Which means that information that’s inaccurate or stale may stay so forever–you gotta think that if the person who added the Knol on Sarah Palin hasn’t gone back to update it by now, there’s a strong chance that he’s lost interest.

The article goes on to catalogue a series of poor performance practices on Knol, showing examples of all we mention above plus wholesale lifting of (out of date) Wikipedia articles. Tom Lehrer would be proud

Plagiarize,
Let no one else's work evade your eyes,
Remember why the good Lord made your eyes,
So don't shade your eyes,
But plagiarize, plagiarize, plagiarize -
Only be sure always to call it please 'research'.

Not only that, but the rating mechanism for one of the articles (surprise surprise) had 5 stars. No idea of how many votes, of course.

This system must have just about the worst set of "game" rules to deliver useful content that one could imagine devising, in fact its sort of hard to imagine how one could consciously design better rules to design cr*p content. The thing is, Googel is full of very smart people so you either have to believe that (i) it really is meant to be this way, (ii) they are not as smart as we imagined, or (iii) - and my fron-runner - is that internal politics drove all sorts of compromises from various internal power holders, as Wenka Booij noted in discussing KPN's web redesign (see our previous post about that here)

So how to fix it?

The thing about Wikipedia is reward for writing (in karma) is positively linked to content quality, and reward is at least neutral in terms of topic chosen. Knol however rewards with ads based on hit volume (assuming a % of hits = a % of clicks), and volume is driven by "pop" content value and volume, not quality. Rewarding people based on "user rating" would just lead to gaming of the system, so thats a non-starter.

The only way to make it work is to treat it the same as all other webpages (there are suspicions that Google favours it in searches) and Google also use it's considerable "spam analysis" arsenal to at least warn writers when they are spam-blogging that they will be removed. And then do it. That gives a carrot and a stick to at least write quality material. Getting people to write on less popular items is still an issue, maybe a declining % rate of payout as traffic rises will reward more obscure topics?

How to make them keep their stuff current is another matter - maybe a "last updated" signal on the page or something like that?). Maybe make writers authenticate themselves via having an "about me page" so that they cannot hide behind clearly anonymised names like "Jean Jacques Frapsauce"

*Or before even - see here.