TrustFlow frequently asked questions

[ciphergoth]

TrustFlow for LiveJournal

Frequently asked questions

What do the results mean?
TrustFlow is making a guess at who is "near" your friends list; who might be on it, but isn't. It does this by looking at your friends list, and the friends list of your friends, and so on.

Is this based on who reads my journal, or interests, or what?
No.

Comments

[immorak]

What does it really have to do with trust?
Doesn't it only list the friends of your friends?

[ciphergoth]

It's designed to be a trust metric; see the description in the community. If LJ friending indicates trust then TrustFlow finds users who you might consider conferring small amounts of trust to. It doesn't quite, of course, which skews the results.

Under some circumstances it will look beyond the friends of your friends to their friends and so on.

[tommytesto]

Maybe it is about trust as much as the LJ 'friends' list is like a real friend>

[brokenbeast]

LJ friends are closer to being trusted than foaf:knows, at least, and people used to get away with papers on the latter as trust networks...

[rosefox]

Does it break for me because my flist is too big? *)

[ciphergoth]

What error do you get? I just tried as you and it seems to work so far...

[rosefox]

Something like "Oops, that didn't work, try again".

[cest_la_vie]

It doesn't seem to be working for me.

[anonymous]

fast flying thing whipping across the screen ...

[ciphergoth]

A better animated graphic would be quite welcome!

Re: progress indicator

[sewcute]

was it on a mac? was it black and white?

I used to have one like that!

[dennyd]

There seems to be a strong bias towards listing people who are (friends of (my friends with small friends lists)) high up in the results, which doesn't seem right to me... are you postulating that the trust of those without many friends is more valuable? If so, do you think this factor should outweigh the (more distributed/diluted) trust of many friends to the extent that it currently seems to do?

[ciphergoth]

Yes, that's right. It's hard to know exactly how to modify the algorithm to reduce that bias; I can't remove it altogether because that destroys the attack resistance, so it's a question of introducing a fudge factor that cuts in when your friends list is short. The tricky think is making sure the fudge factor doesn't introduce a different weird behaviour.

Thinking about it just now, I think I have an idea (extending dead end detection), but it would take a little while to implement it...

[amberdulen]

I wonder if you could factor in how long a person has had their journal? It seems a reasonable guess that newer users have smaller friends lists, rather than assuming that they extend trust less easily.

[ciphergoth]

Sadly, that would make it even harder!

[figg]

Is the referer information used for anything?

[ciphergoth]

To pre-fill the form with a guessed name; that's it.

[figg]

Sorry, I was talking about the links posted to livejournal having the journal name in them: I.e. http://trustflow.lshift.net/?src=figg

I know this isn't taken account into the trustflow computation, but I was wondering what you were planning to do with this information?

[ciphergoth]

I'm watching how the meme propogates. I was using it for things like massaging the way the results are presented and seeing what made people want to post it to their journal.

Actually it's only just occurred to me how much information that is - I was mainly curious to know how many entries were resulting in new traffic, but you've just pointed out to me that it also tells me who is recommending who. It's an unexpected treasure trove of meme propogation information, but one which is going to be diabolically difficult to exploit because I didn't think to give anyone any cookies so I could tie the referrer to the request cleanly. Pissing about with IP addresses and timing is much more annoying.