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

*applause*

[booklectica]

Incredibly accurate - I know virtually everyone on the list.

[purplerabbits.livejournal.com]

I could tell how accurate it was by the way so many of the names showed up as visited links!

[xenarachne.livejournal.com]

I don't recognise 34% (17 out of 50) of those names. Many of the others are ones that I know, but don't know enough to have include on LJ-friends.

Thanks, interesting!

[babysimon]

Is this supposed to tell me who I should trust?

[ciphergoth.livejournal.com]

No. Though the algorithm is intended as a trust metric, LJ "friend" links don't reflect trust, more aquaintance and interest. So your results should reflect who you are vaguely acquainted with or who you might be interested in.

[ergotia.livejournal.com]

Not bad, knew nearly everyone....I find this whole concept absolutely fascinating actually my love. If we can stop shagging and/or angsting for like, thirty minutes maybe we can chat about this on Thursday !

xxxxxxxxxxxx

[dennyd.livejournal.com]

Pretty good... glancing through the list, I'd estimate an even 3-way split between:
people I've previously had on my friends list and might do so again now that I've got more free time,
people that I know IRL or who are good friends of my friends,
people I don't know or know of, and communities that I probably wouldn't object to reading.

[ergotia.livejournal.com]

Ok, being more specific, there were only nine I did not recognise and eight of them were communities.

xxxxxxxxxxxxxx

[adjectivemarcus.livejournal.com]

"Not a valid LJ user name, sorry!"

Checked spelling twice...

[ciphergoth.livejournal.com]

Whoops, it thought the limit was 14 characters. It's 15. Fixed...

[skx.livejournal.com]

Looks neat, are you going to share the source at some point?

[ciphergoth.livejournal.com]

Sure will, need tidied though.

[sparklegoth.livejournal.com]

I don't understand, I really really don't. I tried putting my name in it but it didn't work.

[ciphergoth.livejournal.com]

Works fine when I tried it. I may have broken it briefly while tweaking it to protect against dodgy inputs.

http://www.gothboffs.co.uk/trustflow/trustflow.pl?user=sparklegoth

[neenaw.livejournal.com]

How does it work??

[neenaw.livejournal.com]

Oh, ignore me, read the journal about it. Neat idea!

[lilitufire.livejournal.com]

I was amused that two of the first three were people I'm not on speaking terms with, but yes, I think it does map associations reasonably well.

And provide potential people to stalk, I mean friend *grins*

Interesting!

[ciphergoth.livejournal.com]

Yes, that makes sense if you think about it; they're people close to your circle of friends, who aren't on your friends list "on purpose"!

[wechsler.livejournal.com]

Very effective as a "you're part of the crowd that these people are also part of", or alternatively a "you may not know these people have a journal" thingy. Probably worth looking at the few remaining unknowns for my own curiosity's sake.

[wechsler.livejournal.com]

And possibly more scarily, I actually followed most of the trust metrics discussion ;)

[ex-pipistre.livejournal.com]

worked well for me. most of them were either friends of friends or people who used to be on my friends list, but i removed for one reason or another, or who removed me.

does this work by suggesting people based on interests & common friends, or am i completely stupid?

[wechsler.livejournal.com]

1) What is it with you and buckets? ;) ISTR you were proposing a post-rate-limiting algorithm with much the same terminology a while back.

2) How deep does this algorithm go? I assume you're not pouring water into all 10^6 LJ user's buckets?

Rather, what I'm getting at is - what's the server load/complexity of the algorithm?

3) Know your load... ;)

[ciphergoth.livejournal.com]

1) Maybe I have a bucket fetish I don't know about?

2) As deep as it needs to to answer the question. That's the beauty of this algorithm - it fetches parts of the graph as it goes along, it doesn't need the whole graph from the start.

3) Cute!

[http://users.livejournal.com/_nicolai_/]

Update: Shit, I've broken the machine it was running on. Back later when I've make it check the load average before running the metric...

Haha.

Got a machine to run it on? Otherwise I'm sure people (including me) can give you an account on a reasonably studly machine.

[selectnone.livejournal.com]

Hmm, various people I know slightly, communities I've heard of, alternative-name journals... I'll have to number them later for you.

Have you seen the LJ Graph Browser Java thing?

[wechsler.livejournal.com]

Be more useful if they packaged it for non-windows platforms... any idea how you make it run under unix-a-likes? (OSX)

Directed here by Claycarnac

[miss_squiddy]

Oh dear. My top ten includes a girl I have always disliked immensely and the close friends of a lad who dumped me.

BUT it did find the LJ of a mate I haven't seen for over a year and didn't know was online, which is cool.

*good experiment*

[ephemera.livejournal.com]

very plausiable, in fact - nifty toy.

[lapis-lazuli.livejournal.com]

Even stranger, it shows where worlds collide. I'm finding people who are connected to several (to me) completely separate and distinct groups of friends!

Ack! World too small... Brain exploding! :)

[lapis-lazuli.livejournal.com]

Looking at it more carefully, it also includes pretty much the entire friends list of someone I only share one other friend with.

Maybe she (and therefore her friends list) are connected to me in multiple other ways that I haven't figured out yet?

[tempaccount99]

hmmm... majority for me are people who were once on my friends list, or aren't specifically because i don't want them on, not because i don't know them - it's kindof more a list of people i don't trust... which is kinda kooky, i suppose. but it makes sense, considering what the system is doing.

[ex-stephanie345.livejournal.com]

I found your post above via a friends of friends page.

Quite accurate - not all were people I'd like on my FL but I can see the logic of why they're there.

[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.

[thekumquat.livejournal.com]

Very plausible - I could have sworn that a third of them were on my friends list already but apparently not. Only 1 person I don't know - nemy, who hereby earns my respect for surviving growing up in Egham.

[nils.livejournal.com]

Attempt to fetch info for nils failed: Cannot calculate trust for someone with no friends listed (eg a community, nonexistant user), died at /home/www/docs/trustflow/trustflow.pl line 155.

Oops. Not entirely surprising, I guess...

[ciphergoth.livejournal.com]

heh, I've been fixing the error reporting for a while, it used to just pretend it was still calculating forever when it got an error, I've finally stopped that, and removed some stale files, so you must have been reloading pretty often!