Finding the Collective Intelligence in Social Bookmarking systems

After laying out a few ideas yesterday about how I see a rudimentary collective intelligence emerging out of social bookmarking services I've decided to do more regular tests on the theory. Which is - the stronger the consensus around any tag the more broadly it applies. And the less the consensus the more subjective or specific its application. For example 4 different people might describe Bill Gates in 4 different way -

1. Very rich man
2. Extremely wealthy man
3. Rather wealthly man
4. Incredibly rich man

Take the word count for the above, or the Common Tags in del.icio.us lingo  -

4 man
2 wealthy
2 rich
1 incredibly
1 rather
1 extremely
1 very

Everyone can  agree that Bill's a man! Two people think he's some degree of wealthy. Another two consider him some degree of rich. And each applies a different subjective qualification to that degree. So we can see that the more obviously a description applies to an item the broader the consensus around that description. Its like an inverted funnel with little agreement about the descriptors at the bottom, or the open mouth of the funnel, and a tighter definition emerging as we move towards the top.

Back to the del.icio.us popular page then and scan down to the first item without a giveaway title. Hmmm - Feedwhip - bookmarked by 93 other people (45) recently. Clicking on that link I'm taken to the bookmark information page and shown the Common Tags on the right hand side -

  37 - rss
  21 - web
  13 - email
  9 - tools
  8 - monitor
  8 - web2.0
  5 - website
  5 - useful
  4 - change
  4 - monitoring
  4 - feeds
  4 - service
  4 - tracking
  4 - free
  3 - online
  3 - aggregator
  3 - feedwhip
  3 - feed
  3 - news
  2 - tool
  2 - scraper
  2 - utility
  2 - internet
  2 - tech
  2 - web2
  2 - webservices

Now to remind ourselves that the rules for generating a sentence are -

1. You must start at the bottom and work upwards.
2. Any two tags with the same frequency are interchangeable.
3. You can combine two consecutive tags
4. You can insert conjunctions and plurals (in brackets) to move from a pidgin to a creole.

So here's one way to interpret the tags above. I think they're telling me that Feedwhip is -

[An] internet scraper tool [and] tracking service [for] monitoring change [to] website(s) [via] email [and] rss

Now, clicking through to the FeedWhip website we learn they "track all your favorite websites so that you don't have to. Every time one of your website changes, we'll send you an email!"

Hmm, so the prediction was very close but incorrect, apparently, in saying that the monitoring could be done via RSS. I wonder how many people realize that though? Why is it called FeedWhip then? Shouldn't it be called EmailWhip instead? In fact they explain - "Feedwhip creates an RSS feed based on the changes that we automatically detect,        whether or not the website creators have taken the time to provide one for you." Ok, its reasonable to take this as something of an anomoly then.

Its rather illuminating to carry out a further analysis of individual tagger behaviour. Taking the first 20 taggers I found that the average number of tags they applied was only 2.3 each. That was certainly a surprise as I'm rather verbose myself when tagging. Indeed many users applied no tags at all (but I didn't include any of those when averaging). The interesting thing to note is that no single del.icio.us user tagged as accurately, completely and descriptively as the collective.

Technorati Tags: social bookmarking, collective intelligence

Comments

[Dermod Moore]

I'm learning so much about the semantic web through these articles, thank you so much! I am especially taken with the notion that the collective paints a more accurate picture than an individual.

I've always been aware that people perceive things and people in very different ways, but it is exciting to think that those perspectives can be woven together, through social bookmarking, into a useful and truthful "cloth".

Keem 'em coming!