Facebook is probably using an algorithm that picks from among the friends that the user has interacted the most with. At least that's what I would do if I were implementing it. They already use such an algorithm internally to detect if your account is compromised (i.e., if the set of friends with whom you communicate changes sharply at some point.)
Given that Facebook might be doing this, it is likely that the top few friends would be of the opposite gender, in a heteronormative context. This intuitive expectation has been confirmed in research studies of social networks. e.g., http://www.cs.cmu.edu/~jure/pubs/msn-www08.pdf ("people tend to converse more frequently and with longer durations with the opposite gender").
Perhaps pc can confirm/refute whether this is true in his case?
For me (tried with a few refreshes to get multiple samples), it seems to show pictures of people who I have recently exchanged messages (rather than wall posts, or just about any other form of communication) with.
Given that Facebook might be doing this, it is likely that the top few friends would be of the opposite gender, in a heteronormative context. This intuitive expectation has been confirmed in research studies of social networks. e.g., http://www.cs.cmu.edu/~jure/pubs/msn-www08.pdf ("people tend to converse more frequently and with longer durations with the opposite gender").
Perhaps pc can confirm/refute whether this is true in his case?