Link exchange and interestingness in the blog space. Synchronicity's funny, idn't it? No sooner do I write a post referencing the power-law distribution in blogging, than Brad DeLong and Henry Farrell (at Crooked Timber) take note of the problem of blog spam at Technorati, and write in worried tones about its implications for what Brad calls the "Web-structure information commons." (Henry juxtaposes with Brad's discussion a note about the problem of "flogrolling," based on a report that has blogger Jason Kottke selling links.)

Here's Henry:

The underlying value of the blogosphere is that it is a system that more or less efficiently conveys the decisions of readers that a web page is worth viewing. It does so through links. As Rebecca Blood observed a couple of years ago, the best way for a blog to get attention or readership is to get links from other blogs - especially well known ones. Links are the currency of the blogosphere - and they’re valuable because they have real informational content - they tell us about the blogs, or posts, that another blogger considers to be worth reading.

This has led to the creation of a sort of informal economy of link exchange, with norms regarding due credit, reciprocity and so on. It’s by no means perfect, but it does a pretty good job in ensuring that good posts and good blogs get attention. Not a perfect job - network effects, path dependence, link cartels and so on all have a distorting impact - but, as stated, a pretty good job. ...

The problem is that the political economy of link exchange on which this rests may have some inherent fragilities. Links are valuable currency because they refer to the ‘interestingness’  ... of a specific post or blog. Interestingness is a subjective concept, of course, but the more people (especially people who share your tastes), that find a post or blog interesting, the more likely it is that you yourself will find it interesting. If the relationship between links and underlying interestingness is broken, than many of the advantages of the blogosphere as a means of sifting through views and highlighting the interesting ones will evaporate.

I see some very interesting theoretical questions lurking here, and though I'm neither an economist nor a statistician nor a network theorist, I've been thinking about these questions for a while now, so here goes.

I think there's a problem with Henry's term "interestingness" that somewhat vitiates his discussion. The problem is measured by the difference between "interestingness" and "relevance"—which is also, I think, a parallel difference between what Technorati does and what Google does.

The conceptual leap made by Google was to recognize, in the structure of the Web link (a marriage of resource-pointing behavior with keywordish description), an opportunity to analyze, across the information space, the existing association of resources with intelligence about them. That's an analysis of relevance: simplistically, the more pages associate a given resource with a given bit of description, the likelier it is that the resource is relevant to the bit of description, and vice versa.

To treat links as bloggers and blog-mapping tools like Technorati do, as an attentional currency—ideally, a measure of interestingness—follows a different conceptual model from Google's, though, even if both can be made subject to similar attacks meant to distort the information space. "Reciprocity" is the chief difference: the link, in the Google model, isn't reciprocal; it ends at the pointed-to resource. Where links are treated as currency, though, exchange—rather than pointing-to—is the fundamental gesture. "Interestingness" may or may not be part of the real-world impetus to any given link-exchange, but it's not part of the structure of the exchange. At best it's a social form that exerts pressure on the character of the exchange, that biases people within a given sphere (say, the political blogosphere) toward valuing "interestingness" highly among the whole set of motivations that produce a link exchange.

The notion that "the relationship between links and interestingness" might be in danger of breaking isn't persuasive to me, in other words, because I think the relationship was always-already broken as soon as links became currency. (More accurately, I don't think the relationship can have been broken, because it was never present in the structure of link exchange.) And while the blogosphere does, at least so far in the game, reliably indicate the pressure toward "interestingness" as a social aspect of link exchange, link exchange makes a wildly unreliable measure of interestingness within the blog space. (Assuming for the moment that such a thing could really be measured: but one does have a sense of the difference between an ideal map of information worth in the blog space and how that map is constructed in reality.) In fact, the power-law effect in social networks pretty much guarantees that, as it expands, the link structure of the blogosphere will become less and less reliable as an index of good/interesting stuff: necessarily, more and more such stuff is going to languish in the tail of the distribution. Flogrollers, in that sense, are already (as it were) behind the curve: they can't distort the picture more than the network effect already has, and will continue to.

Addendum: After I made a version of this point in a comment at Crooked Timber, Henry noted that he and Dan Drezner have done research to indicate that, at least for political blogging, the network effect produces not a power-law distribution but a (less drastically skewed) lognormal one. I haven't had time yet to read their paper, but I certainly will.

posted by michael  1:42:00 PM  
tell me about it []