What do Tomek from Poland, Stephanie from Boston, David from Los Angeles, and Nathan, also from Boston, have in common?

Facebook thinks I should be friends with all of them.

And I have no idea who any of them are.

I am, of course, talking about Facebook’s Suggestions — people I should consider being friends with.   And the suggestions are awful.  But they don’t have to be.  Facebook has a ton of data about me — how old I am, where I live, where I’ve lived, who I’m married to, where I work, have worked, went to school, who I consider my friends, which friends I interact with most often (on Facebook, at least), etc.

And one would assume that Facebook would want to make the Suggestions tool as close to perfect as possible.  Hey, it fits with their mission, over there on the right in big bold letters: “Facebook helps you connect and share with the people in your life.”   If Facebook could actually help me find those people, it’d be a huge first step toward connecting and, subsequently, sharing.

Further, Facebook certainly makes suggestions a focal point of the site.  They appear on the front, News Feed page, right near where the only ad on the page lies.  On other pages, they’ll often take the place of ads.  That’s dedication.

So Facebook should be able to get Suggestions right, or, at least close.  But do they?  Hardly.  The criteria Facebook uses in making its suggestions seems to be simple and binary: Do you and your suggested new friend have a friend in common?  And the results are ridiculous.

Let’s look at Tomek,  Stephanie, David, and Nathan once more, and see what information Facebook has.   As you will see, Facebook has the ability to draw an impressive graph of my social network, and with a high degree of accuracy, should be able to predict whether I actually know someone.   That they fail here is stunning.

Tomek

I can’t figure out why Tomek and I should connect.  How can Facebook?

  • Tomek and I have one mutual friend, a former co-worker of mine.  That co-worker and I have eleven friends in common; Tomek is not friends with any of them.
  • Tomek lives in Poland.   I haven’t worked there.  I did not go to school there.   There is no reason to believe I have ever been there for anything more than a few days.

The mutual friend and I share eleven friends; that’s a social circle of 13 people.  Tomek is connected to only one of those thirteen people.  Facebook should take this as evidence that he’s not connected to the other 12; unfortunately, they do the opposite.

Stephanie

The interesting angle here: I may have actually met Stephanie before — in fact, it’s incredibly likely — and I don’t think either would recognize the other on the street.   If Facebook can differentiate a  person I may have once met from a person that I actually know, well, that would be great.  They should be able to.

  • Stephanie and I have two mutual friends.  They are siblings, which Facebook knows.
  • The siblings and I have, in total, three other mutual contacts; that is, there are three people who are friends with me and both of the siblings.  One of those three people is my wife; another is one of the sibling’s spouses.

My guess is that Stephanie is a relative of the siblings, and I’m pretty sure Facebook could do some interesting math to figure that out.     Facebook also has Stephanie’s connection data, and that would buttress my guess.   If I’m right — that this is a question of cousins, then she’s be a poor suggestion, albeit not a terrible one.

David

This one is obviously wrong.

  • David is originally from Los Angeles but now lives outside D.C.  I, again, have never lived or worked in either place.
  • David and I have one mutual friend.  That mutual friend lives in D.C. (and did not grow up in LA).
  • The mutual friend and I went to college together (we’re in the same college network) and our four mutual friends all also went to that same college at roughly the same time.

So: David and our mutual friend know each other from their current lives, while I know the mutual from college.   David is not part of that social circle.  This suggestion is clearly incorrect.

Nathan

This is another case where the social circles are easily drawn, and Nathan and I are not in the same ones.

  • We have two mutual friends — who are married.
  • The couple and I have three other mutual friends, who are not friends with Nathan.
  • The couple and I went to the same college.  Nathan went elsewhere.

Married couples have joint friends; that is, friends of the couple as much as any individual.  But they also have individual friends — friends who are legitimately friends with both individuals independent of the other.   I don’t know if Nathan is in the former group, but there is reason to believe he is.  Meanwhile, I am most likely in the latter group, having going to college with both husband and wife.  In any event, the three other mutual friends make it clear that Nathan is not within that circle.

Facebook can easily draw my “social graph” — they, quite literally, coined the term.  They have all the info they need.  Why aren’t the using it?