Social network representation: graphs or sets? for social network representation, what is better, sets or graphs ? What kind of feature the first gives that the second doesn't and viceversa?
 A: "Which is better, graphs or sets" depends crucially on what limitations you're imposing on how you use these.
For instance, as a potential foundation for mathematics, sets can be used to do essentially anything if you put your mind to it. For instance, sets can be used to represent a graph. You can represent a graph as a set of labelled nodes, and the edges $u\text{-}v$ are represented by unordered pairs $\{u,v\}$ (which will be singleton sets in the case of self-loops). Or you can represent a graph as a set of labelled edges, where the nodes are given as collections of edges $\{e_1, e_2, \ldots, e_k\}$ which meet one another (and where each edge can belong to at most two such sets).
More powerful than graphs are hypergraphs, which generalize graphs in that each 'edge' can link multiple nodes. If you picture social circles of friends as overlapping sets of people, what you're envisioning is a hypergraph; but conversely, a hypergraph is little more than a collection of overlapping sets.
You could represent the circles on Google+ as a sort of directed hypergraph, where each edge consists of a mapping from a single individual to the people that follow them, or from a single individual to all of the members of one of their circles. And of course, these functions (which are set-valued functions) can also be represented with sets if you're so inclined.
So it will depend on how creative you want to be with the particular tool you have at hand. Certainly, a graph on a set of people is more useful and informative (but requires more work to describe) than just having a set of people; but this is not the limitation of what you can "do with sets". 
