Why Friendster sucks

The main link here is to a ZDnet article about a patent dispute involving the Friendster dating/personal networking site.  It's somewhat interesting, but linking to it is mostly an excuse for me to talk about my own impressions as a Friendster user.

Friendster is a site where you register, write something that is basically a personal ad, and attempt to convince your friends to register.  It keeps track of who is a friend of whom, and it allows you to search within your "personal network" - defined in graph-theoretic terms as those members who are connected to you by paths of at most four "friend" links.  So, the friends of friends of friends of your friends, or anyone more closely linked to you than that.  There are search and linking and email features designed to allow you to use it as a dating site, but also to encourage people to join even if they aren't interested in dating as such, so that they can act as introducers for others.

The trouble is, it sucks.  The first and most fundamental problem is that anyone more than two links away from you is basically a stranger - you don't have any friends in common with that person, so there isn't anyone who can act as a trusted introducer.  The rules of graph theory dictate that almost everyone in your personal network will be at the extrme boundary, so those in your "personal network" are basically just a random bunch of strangers, and in that case, the personal network concept really just serves as an artificial limit to how many people you can see.  Limiting the size of a dating site database is really bad.

The second problem is that because of that limiting effect, there's an incentive for you to form spurious "friend" links with strangers, because that will expand your network and allow you access to other strangers you might want to date.  That incentive goes both ways, so you and some random stranger have every reason to cooperate to form a spurious link.  Such links hurt the system for anyone who actually wanted to use the "friend" relation in a meaningful way, so the "friends" graph quickly loses whatever value it had to begin with.

Those two problems seem to be intrinsic to the system, and I don't know that there's anything that could be done about them.  Let's consider some things that could actually be done that might make it suck less.

First of all, it needs better searching capabilities.  At the moment there's a geographic search, but it only works in circle centred on your postal code.  I can't, for instance, say "Oh, I'm going back to Victoria for a school break, I'd like to know about people I might want to meet there." At best, I could temporarily change the postal code in my profile.  This would be a trivial search to add and it would be really useful.

Even the searches that are provided, don't work.  For instance, you're encouraged to list your favourite books, movies, TV shows, and other interests in comma-separated lists, and if you click on one of the items in your or anyone else's profile, the system searches your network for others who have listed the same item.  Trouble is, what it does is it splits the item into words, searches for each word separately, and then logical-ORs the lists.  So if I click on "youth rights", I get a list of people who have listed "youth", and/or "rights", not necessarily both and almost never as the single item "youth rights".  That's the default and only available behaviour for the search command, and it's the least useful behaviour they could reasonably implement.  Even the simplest searches are often flat-out incorrect - check the box on the form saying you only want to hear about single people, and you're likely to get a lot of married ones on the returned list anyway, because apparently activating some filters disables other filters although I still haven't been able to reverse-engineer the exact rules for that.

They lump "divorced" and "separated" people into one category.  To me, those words imply "unmarried" and "married" respectively, and that's a show-stopping difference for dating, as far as I'm concerned; the system does us a disservice by eliminating the distinction.

The rules regarding appropriate profile content are not enforced.  When I first joined, one of the very few people listed as a female in my area, was someone who had filled "her" profile with images of a model obviously stolen from a porn site, most of them sexually explicit.  The written portion of the profile was an explicit description of how much she enjoyed anal sex with multiple anonymous partners.  "She" had several hundred friends listed, and a long list of testimonials from people who made clear that they had never actually met "her" but just thought the profile was cool.  Now...  I don't wish to rain on anyone's parade, but I don't believe for a moment that the person described in the profile really exists; she's a construct created by someone who wanted to "game" the system.  It's conceivable that a gay man was trying to seduce credulous straight men with the fake profile, but I don't even think that's likely - I think it's probably just someone who thought it would be cool to try to form as many friend links in the system as possible.  Why is this a problem?  Because with several hundred links to her, it means that as soon as she moves one step in from the fringe of my network, all of a sudden there'll be hundreds of people in my network who are only there by virtue of the fact that they fell for the scam; and those are people I really don't want to meet or trace friends paths through.  There's no easy way to ignore them; I'm not told that my path to them goes through the fake profile, until I've already clicked all the way through the search to their profile.

So it needs to have an "ignore" feature; I'd like to be able to put some kind of flag in my account saying "don't show me anyone who is only connected to me through the buttsex girl".  Furthermore, they need to listen to their complaints and have real consequences for people who "game" the system - after multiple complaints about this person from me and (I hope) others, I eventually saw that the profile had been changed.  The obviously pornographic pictures were removed but the remaining photos were still obviously posed professional photos of the same model, almost certainly stolen from the same picture series on the same porn site that had yielded the nudes, and the most explicit text in the profile was toned down, but the hundreds of spurious friends links remained.

Even without sex scams, the few extremely-high-degree vertices still raise a problem.  I'm connected to something called AsianFOCUS which doesn't even claim to be a real person (its photo is an organization logo) but has 351 friends, and to someone named Jason in Boston who has 76, and through those two I'm connected to what appears to be most of the Hong Kong immigrants in Toronto, half of whom are high school girls who've lied about their ages to get into the system at all (DAMN AGE RESTRICTIONS! The limit doesn't keep anyone out, it just forces them to lie, which is the worst possible solution.) and then registered under phony names in order to be unique and stand out on the first-name list just like everyone else.  I wouldn't go to far as to advocate a limit to how many friends one account can have, but allowing a per-user killfile (so I can say "don't bother tracing paths from me through Jason or AsianFOCUS"), and actual enforcement of rules like "don't register under an obviously phony name" and "don't use your profile to advertise a business or organization" would go pretty far towards making the system usable.

I'd like to be able to ban everyone who puts a photo of someone other than themselves, or a pretty flower, or themselves as a baby, or their pet, etc., into their profile, but that might be going a bit too far as an overall policy.  I don't want to poop the parties of other people who want to use the system in a stupid way, I just want them not to interfere with my own use of the system.  Letting me flag individuals to not appear or be used as links for my own searches, would be good enough for this kind of thing.

What if it allowed arbitrary-length friends paths, instead of blocking them at four hops?  Then there would be plenty of people in your network you have essentially no connection to, but that's pretty much already the case - ten hops isn't really any worse than four hops.  It would eliminate most of the reason for people to form spurious links.  Some people would still want to collect as many links as possible just as a goal they set for themselves, but there would be less drive to do so in order to expand one's own dating pool because the pool would only expand when you formed a link that joined two connected components, and that would be pretty rare.  Such graphs are almost always found to contain one huge component that contains almost everyonel; additional links after that happens just shorten the paths without expanding the component.  So I think the system would work better if they eliminated the four-hop limit, although that would require them to admit that the "everyone in your network is close to you by friends links" thing is a sham and it's really just another dating site.

Now, Livejournal also has a "friends" system, but it doesn't seem to get as much abuse.  Why is that?  I think Livejournal has a limit on how many friends you can have (although that may be different depending on your account class).  More importantly, with Livejournal there are lots of other ways to find and contact other users, and Livejournal doesn't encourage you to explore the friends graph out to more than two or three links.  So with Livejournal there's much less incentive to try to collect distant fake "friends" in order to expand your network.

Finally, Livejournal uses a directed graph unlike Friendster's undirected graph - I can list you as a friend of me, without needing your consent to do so, but that doesn't create a link in the opposite direction (me as a friend of you).  I don't know exactly what the effects of that difference are, but given that Livejournal seems to work and Friendster doesn't, I wonder if the difference may be important in same way.  The bottom line is that if Friendster wants to survive, and especially if it wants me to pay for an account when it eventually becomes a pay site, it has to come up with some solutions to its problems.

[Why Friendster sucks]

Comments

Copyright © 2003, 2007 Matthew Skala
Updates to this entire site:
Updates to this category (reviews) only: