How does Facebook generate suggestions?

Question

Recently, I've received friend suggestions for my family members overseas - but here's the thing. I have no family members that use facebook, we share no common name, and we're not even on the same continent. In fact, I haven't communicated with them for over a decade.

Yet Facebook knows about them and knows that they are probably a good suggestion. My question is from a data perspective, where does Facebook get this information?

I'm not concerned about privacy, but rather the technical aspects involved. Is there a funky algorithm that makes it all work? Or is Facebook tapping into information about me that I have no idea about?

Answer 1

Facebook does do some weird stuff - I read that if someone does an import (gives FB permission to run through their gmail contacts for instance) and you show up in their contacts and they chose not to add you FB will still remember the connection and they could potentially show up as a suggestion for you.

http://blogoscoped.com/archive/2009-06-12-n15.html

http://consumerist.com/5301425/facebook-where-are-you-getting-these-crazy-friend-suggestions-from

Answer 2

I did an experiment recently, if I have a person's full name and ip address, I could know who that person is? Where he/she lives? Which school they went to and other public information.

Such a facility can be easily programmed (and many have already attempted to). Think this way - I know a person's last name and location. Now I could track down everyone who has the same last name and reside in the same locality. This way I can have a simple social graph of a family tree. Match more details and I get a better picture of 'relationships' these people share. (Try it out).

I can't even imagine the amount of data a 200 million strong social network can produce. I guess they do it this way - a) There are 4 people A, B, C, and D. 'A' has emails of 'B' and 'C' but not of 'D'. Both 'B' and 'C' have email of 'D'. Now I could say there is a 60% chance of A knowing D.

b) 'A' visited the profile of 'D'. This is the connect between them - A->B->C->D. Now, there is also a person 'E' who is directly connected to both A and D. So shorter chain -> A->E->D. With these 3 pieces of information, I could there is a 70% chance that A knows D. In this case 'D' did nothing and got a friend suggestion.

So I think its quite easy to generate a social graph, all you need is some useful data and some motivation.