As I’ve mentioned before – Brand Association and Your Twitter Followers – in some of my more mercenary moments I see the network of people who follow me on Twitter as a collection that has value, which is one reason why I spend a few minutes every two or three days blocking new subscribers that don’t seem to add any value to the current state of my network (see also Chris Lott/Rumninate: Twitter Pruning Principles).
Whilst pruning new subscribers today, I got a bit fed up of it, and started wondering about whether it would be useful to have an ‘all followers’ category, which might include spammers and pr0nstars, and an ‘acknowledged followers’ list, which is more like my current followers list.
Then I started wondering about the phrase ‘circles of trust’, and what sorts of properties a clustering algorithm would linarly order followers based on how ‘valuable’ (whatever that means) or ‘trustworthy’ they might be.
(Of course this is a getting a little PageRank and HITS like, so I must be nearly ready to start reading that literature with feeling… but there’s still a little way to go on my personal voyage of discovery yet, I think…!)
And it struck me that one of the measures I mentioned in Finding New People to Follow in a Hashtag Community might be appropriate – specifically, the number of my friends who follow each of my followers.
So the first pass of the ranking/ordering algorithm would look something like this:
for each of my friends, fr: - get the list of people they follow (i.e. the list of their friends); for each of my followers, fo: - count how many of my friends follow them, N_fr(fo); order the list of my followers according to N_fr(fo)
This requires (N_fr + 2) calls to the twitter API: 1 call each for a list of my friends and my followers, N_fr calls (the number of people on my friends list) to grab the follower list of each of my friends.