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

[cryptodragon.livejournal.com]

It worked extremly well. Most of the people I at least knew of or I know why I should friend them.

[lnr]

keirf pointed me at it. Very very plausible, nearly everyone is people I know vaguely IRL or friends of friends.

[lnr]

And indeed on looking at the half dozen or so I didn't immediately recognise I was able to work out who they were or why they were connected to me almost immediately looking at their userinfo. Very very impressed indeed.

And you're number 33 :-)

[mirukux.livejournal.com]

very nifty tool there.

couple of suggestions; would it maybe be an idea to use a combination of both percentage of similar friends and number of similar friends, because i've seen people's results where the top linked friend is an empty journal with only one mutual friend friended.

also, how about a community 16x16 icon for communities, or is that info not included in the lj protocal (or whatever).

how about making a version which shows a list of the people who themselves have friended a large ammount of your friends. slightly less relevance, but it might be interesting to see the different social dynamics. just an idea, probably a bad on at that.

thanks for your time

[ciphergoth.livejournal.com]

My algorithm does not measure who has similar friends to you; it looks at who your friends list as friends, and who they list in turn, and so on.

I can prove this algorithm "attack resistant". I'm not sure I can do that for the sorts of proposals you're making.

The community thing is work; I'm not sure I can be bothered! Sorry!

How would you find such people? You can't check every LJ user. You'd have to rely on "friends-of" lists, but not everyone enables them. Also it's definitely not attack resustant.

No match, sorry ...

[kaunt.livejournal.com]

Terrific! Not a single person I know, no even you! Wow ..

Re: No match, sorry ...

[ciphergoth.livejournal.com]

It won't list anyone who's already on your friends list.

Looking at your friends list, the only people on it who list friends are me and thekonst - if you don't know the people on my friends list or his, you're out of luck!

[echo-echo.livejournal.com]

There was one person on my list I didn't know the name of. Everybody else has either been on my list at some time or other, is somebody who I am reasonbly aware of the name or in fact know reasonably well in person.

And thus we have a rough map of the entire UK goth scene. :)

Do the users tend to fall into clusters or are they fairly widespread? ie: big groups of people who know each other or is it a rather wide network with no real core?

[kate-r.livejournal.com]

Mainly an assembly of people I know who aren't on my freinds list. I wouldn't mind adding quite a few but always assume that people who I don't know that well don't really want me reading their stuff.

[red-phil.livejournal.com]

Pretty good.
Although your right about it picking people you may have left off intentionally.
I don't think there is much you can do about that.

The discussion on how people with multiple friend allocate droplets to one per cycle doesn't seam right.

How about starting with a size 1000 drop that splits between all child nodes at each stage until it is sub size 1 and there it evaporates?
This would decrease the effect of people having large friend lists a little and stop the search delving too far off into the depth of LJ space.

I don;t know much about these things , but from 1st principals it seems it may help. Just let me know If I'm being a dumb amateur.

[ciphergoth.livejournal.com]

Not sure I understand what problem you're trying to solve. Both our proposals mean that if you have twice as many people on your friends list, you pass half as much trust on to each of them.

If I was going to use continuous-size drops and actually split them, I'd write a version based on solving linear equations (actually finding eigenvectors, I've figured out).

[ishkhara.livejournal.com]

Very interesting! I have come across some of these people's journals, either through interaction with friends or just through browsing. But i'd say it gives a very accurate summary - spooky in fact! :)

[littlemsmortis.livejournal.com]

heh, some of these people used to be on my friends list but I had culled for whatever reason. I know some others and know of more of them. Not many totally unheard of ! cool :)

[mina-laury.livejournal.com]

Commenting as asked...

I found the tool through claycarnac and found the results fascinating - thanks!

[nim.livejournal.com]

you so clever!

[rainbowskye.livejournal.com]

Most interesting..I knew nearly everyone on the list, some have been on my list in the past, and interestingly there are two distinct groups of people I know from different places..using the other kind 'who is defined by your friends as a friend but not by you' tool I end up with almost exclusively uk_goffs!

What a great gadge!

[malildragonfly.livejournal.com]

I think it's great.

I'm always reading my friends' friends' journals! :)

Observations and snooping.

[grumpy-sysadmin.livejournal.com]

My own list includes almost exclusively people I don't know. But I think I'm an outlier, since most of my livejournal interaction is not physically local (or even people I'm likely to meet in real life). I fall into the age of computer users post-BBS but pre-local-interest web sites, I guess. I don't want to talk to real-world friends on IRC when I could just get together with them that evening.

In any case, one interesting observation: this breaks a certain security through obscurity. Certain people (I know of several) keep more than one livejournal account (one to actually be a journal and one to read and comment with or to participate in certain communities without association with a real name or real picture, whatever). On of those people's more public lj user was on my friends list, and I hadn't bothered to track down the other (though I knew it existed). It popped immediately up in the TrustFlow output (presumably because her friends and both of her accounts in their friends list).

Re: Observations and snooping.

[grumpy-sysadmin.livejournal.com]

Incidentally, you've haven't a clue who I am (and will probably never meet me). But go look at the commonalities in our list on interests. (I think you may be the only other person I could find who went to the effort to list both subsets of cryptography.)

[rozallin.livejournal.com]

"Calculation for rozallin is still in progress, please reload later. But not every three seconds, please! :-)"

Now that IS class! :D

[emmelinemay.livejournal.com]

interesting. very very interesting.

I probably know more than not, but there are a lot of names i don't recognise. but then, i'm crap with names!

found link from fuzzygoth

[sheepthief.livejournal.com]

I recognise every name but one, know the majority of them via other on-line means, and have met most of them in real life.

[sheepthief.livejournal.com]

Oh, back-tracked link to tya as many folks weren't saying where it originated! Don't know if you can change the output so that cut'n'paste of a link is more readily included.

[ladyserpentine.livejournal.com]

I am very impressed. Most of people listed I know of, but haven't met them IRL. Several I have met but don't know very well. I may add them all to my friends list and see what happens!
Found link through needler and fuzzygoth

[rozallin.livejournal.com]

I just got my results back:

2 of the users were people who were formerly on my friend's list.
2 were communities I used to be a member of, but left.
But most amazingly:
10 were users whom I was once quite close to on online chat three years ago, but whom I have lost contact with.

The rest were unknown to me.
Is it possible to disregard any journal that is a community?

[ippola.livejournal.com]

very interesting results for me too as I was aware of most people on my list (and even the ones I didn't know looked very interesting!)

[eggwhite.livejournal.com]

You asked people to post if they used this... well, after spotting it on 3 journals (Cryx, Hugggs & Blue_Cat) I've poked it myself out of curiosity...

[tk7602.livejournal.com]

i found this via a friendsfriends link.

for me, the results are :

1 community and 2 babies listed and discarded.

of the remaining 47, i know 42 of them in the real world, but none of them are in the core group of people that i see / talk to regularly. i would estimate that the most recently i've had any direct interaction (in person, phone, email, etc...) with any of the people listed was about 2 weeks.

[ailbhe]

I don't understand how this differs from http://www.livejournal.com/friends/popwithfriends.bml but it obviously does; my results are very different. How does it count?

[sinders.livejournal.com]

I know most of the people that came up on my list, and some are even on my friends list!

[redcountess]

First attempt failed - didn't have enough info about me cached. Just tried it again, and found it very amusing - list includes two former housemates.

[redcountess]

As you hypothesised, the people at the beginning of the list were people I have resisted putting on my friends list (or removed) for one reason or another.