Try my new trust metric!

[ciphergoth]

I have implemented my new trust metric, and applied it (fairly crudely) to LiveJournal friends lists as an experiment. The result is a tool that tells you who is "closest" to your friends list who isn't actually on it, like "popwithfriends" but more sophisticated.

Who is closest to your friends list?

Let me know how plausible the results seem to you.

Update: Shit, I've broken the machine it was running on. Back later when I've made it check the load average before running the metric... Update: Back again now, with a rate limiter to keep the load average reasonable. Gosh, hadn't expected such popularity!

Comments

[mr-tom.livejournal.com]

Nice toy, although I question the utility. (although to be fair I have joined a community as a result...). Not a problem with the tool itself, but with the semantics of 'friend' lists, and then fact that a FOF on LJ is not necessarily such a thing. Something that would let one assign a non-boolean trust weight to one's own friends would be a neat tweak, and work around this a bit..

[ciphergoth.livejournal.com]

A version that allowed you to actually specify a "trusted" list in order would be far more powerful, but who would bother to define such a list? The bonus of running it against existing "friends" lists is that I have an existing graph of 1E6 nodes to play with.

[mr-tom.livejournal.com]

Well, if you had your own friends list in front of you with some tickyboxes or radio buttons before hitting 'go', you'd have the option. And as more people do this, the network would become more accurate. (Although you would need to store user's preferences somewhere...)

The big pool o' nodes is a definite bonus, but the effect of unweighted trust is that the system tends to recommend users that are totally obvious to anyone that's seen their friends' friends lists, but that are omitted from their own for social reasons. (Such as being a complete idiot, f'rinstance.) So the drop in elegance would give a great increase in utility...