RDF (the technology underneath FOAF) is all about representing knowledge. It's quite philosophical, actually: each RDF statement has a form similar to that used in predicate logic, wherein we can say things like "Plato wrote the Republic", "Aristotle is a student of Plato", or "Socrates knows Plato". This last relationship is a bit controversial in the FOAF world. Currently FOAF includes the "knows" relationship but nothing richer than that. Some have argued that FOAF needs to more fully define relationships, but I think the description of foaf:knows in the spec is fairly persuasive that describing detailed relationships is probably misguided. Part of the point of RDF is that no one RDF resource contains all the relevant information. Indeed, information emerges spontaneously from the web of RDF data. For instance, the fact that someone is in my roster may lead to their being listed in my FOAF file as someone that I know. The fact that the two of us are registered with the same chatrooms or on the same mailing lists may provide evidence of a stronger relationship. The fact that we are both JSF members, both work on the same open-source project, or have co-written a JEP may lead you to believe that we are colleagues. Further evidence regarding the closeness of our relationship might be gleaned from working out a network graph of our mutual "foaf:knows" connections (FOAF to SVG, anybody?).
Naturally, this might be putting too much faith in machines. Is there some knowledge that a friend or relative and I have that the machines do not? I think there is. The machines don't know that someone with the last name Ruhland just might be one of my 72 first cousins, and probably wouldn't be able to figure that out from "foaf:knows" alone no matter how many times it had iterated over various RDF data sources (though if lots of geneology data were in RDF, perhaps it would). The machines don't know that once or twice a week after lunch I take a walk with my friend Diz and we have wide-ranging discussions about technology, society, and just about anything else that strikes our fancy. So I think that something more than "foaf:knows" might be in order, but I'm not quite sure what it is.