In a power law distribution, the highest-rated and lowest-rated members enjoy an inverse-power relationship: e.g., if one site gets 1 million hits a day, then there are 1 million sites that receive one hit.

Most of the time, people ('bloggers anyway), are interested in how you become the site with the 1 million hits, and the answer varies, but probably has a lot to do with a concept called preferential attachment.

I've been interested in the other end of the curve, the 1 million sites that get one hit apiece. Conventional wisdom would hold they must not be very interesting - preferential 'disattachment' - but that's not necessarily the case.

One million bloggers is sufficiently large a 'universe' that there must be someone writing (or drawing or photographing) something interesting. In fact, there could be what I'm calling a 'lone genius' in a set that large - someone whose ideas and insight are way beyond the normal range.

But how would the lone genius be discovered? Presumably, once such a resource were located, preferential attachment would quickly drive the site's traffic to some other place on the curve. But, lacking promotion, mass linking from other sites or some other discovery method, lone genius might languish, unread.

The typical discovery - search engine indexing - wouldn't work if there were no existing links to lone genius. Indeed, many low-flow Weblogs (and other Web sites) are indexed in lots of search engines, but receive very little or no human traffic because they don't show up in the first couple pages of hits.

Which raises the question of linkage between the quality and usefulness of a site and Web search engine rankings. Word frequency algorithms are likely to be a poor judge of genius, which is why many search engines also look at the number of inbound links. A number of observers have pointed out the value of the trick of co-opting human judgment to assess quality - things like the MIT spin-out Firefly and Amazon's 'suggestions' feature use this mechanism.

But it's also instructive to look at realms where preferential attachment drives to the middle, not the ends, of a theoretical 'genius curve'. Take music: mass appeal tends toward choices like Britney Spears, who, for all her popularity among people of a certain age and economic status, would not be described as a musical genius by many serious musicians.

So, even if lone genius gets indexed on Google, she's not particularly likely to be 'discovered' and promoted onto the other end of the curve. So a 'genius discovery engine' might do better by trolling sequential IP addresses, rather than looking for massed inbound links. Alternately, if there were a 'certified genius index', presumably a Yahoo-like human-moderated construct, then the genius engine could look at the outbound links of that group. And the sequential troller might be the input for the moderators... Rainy morning thought...
Comments 10:58:39 AM    

Chris Gulker's view from Silicon Valley - in words and pictures