Interesting Ideas Afloat

...by people thinking about how to better understand the blogosphere, through topic clouds and weighting of bloggers. People are asking for things they are interested in seeing, expressing concerns and making great observations about the overall problem. This continues a discussion I started there, and continued there, there, there and there.

The del.icio.us tag for this meme is linklove.

Ericka Menchen Trevino/technology and the social on Blog Rankings:

It’s not surprising that there is no single way to determine ‘the best’ blogs, and any attempt to sort the mass of blogs will be threatened by SPAM and the preconceptions of the rank creators. Alternative metrics do need to be developed - the more the merrier, but what will be more informative is studying how real people find the blogs they read.

So should we have different metrics for different communities? Or is that too much? Maybe we think about including ratios of different metrics of participation, so that many kinds of social gestures can be included depending on styles of participation. A blog would still be weighted for some participation in a community? In other words, if someone comments a lot, and posts now and then, they might be as conversational as someone who comments little and posts a lot.

Ed Vielmetti in Yi-Tan: Jerry Michalski and Mary Hodder on "Link Love" (a post based on today's Yi-Tan call where I was a guest host talking about the problem of links as ranking mechanisms, and how we might find topic communities and weight things we value, like conversationalness):

The observation is that all of the sites that keep track of "top 100" lists or "top hits" scores for searches or weblogs have opaque algorithms for determining same, and what results as the lists are not very accurate once you exit the mass public Internet and get into more specialized fields or subfields.

...

...one really interesting salient point about the metrics used to compute influence and activity within the sadly named "blogosphere". (ugh) Most of the blog ranking tools use links as their proxy for love. In my experience, however, it's the good and useful comments and discussions that are a lot better reflection of whether someone really cares enough to click through and make a difference...

Comment discussions are important and if we can solve the comment spam issue, might be a great metric for folding into measures of conversation.

Ed also suggests we look at energy as a measure:

...people are differently influential in a network by the amount of energy they bring to others in their sphere. Should we be recasting our measures and metrics as "link energy" rather than "link love"?

J. LeRoy on Are You Really Atrios?

So the issue here is how do we invent metrics to judge the relative social worth of one blog post over another. The analysis needs to take into account the fact that gamers of this sytem are flooding the internet with content-free blog posts and web pages that contain key words and copious links. These sites are primarily aimed to get eyeballs to make money for the people putting up these sites. They are web spammers. Web spammers are very good at gaming rankings like these and defeating their purpose.

Community is measured differently by different people. Some measure it by the amount of participation in discussions. Some measure it by linking. Some measure it by blogrolling. But all these are, at best, indirect measures.

Direct measures may be no better. If our solution is merely to tag links by their relative importance, those tags are easy to spot and easy to fabricate for web spammers.

Peter Kaminski talks about the Network Map vs. Cocktail Party

This reminds me of Flickr's new interestingness feature, which can rank large sets of pictures by, well, interestingness, as demonstrated by more or less social cues. Social cues are key, because they illuminate what real people find interesting, instead of just statisticians (or marketers).

...

...there is lots of interesting stuff in any long tail. So while you may find a (biased) ranking scheme you like, and be happy with the top 100 most of the time, you (or the system you're using) also need to make sure you randomly see some of the long tail once in a while, to discover the interesting things your favored ranking system hasn't.

...

...a network graph that shows nodes and maybe does fancy node sizing based on rank and all that is cool, and fun to play with if it's interactive (in the same way sodaplay is fun to play with), but it doesn't tell me anything. It's hard for my monkey mind to parse the map.

A better metaphor, I think, is a cocktail party. I want to see faces, who's clustering with whom, and how the clusters evolve. Make me an animated picture of a cocktail party (or a tribal gathering, or a barn-raising), and that I'll get -- I've been evolved over thousnads of years to intuitively grok those kinds of social situations.

That's a very interesting idea.. what that would look like I don't know, and many bloggers are pseudominious and don't really want their pictures all online (not all of us are exhibitionists, in fact few of us are across 14m blogs).