Social recommender systems span beyond content recommendation. As mentioned in the introduction, social overload originates from both information and interaction overload. Since people are the key element that makes the web “social”, recommendation of people is a central pillar within the social recommender system domain. Terveen and Mcdonald \cite{Terveen_2005} coined the term “social matching” for recommender systems that recommend people to people. In their work, they explained why a people recommender is a unique RS, which is different than recommendation of other artifacts, and thus deserves its own special attention. Among other aspects, trust, reputation, privacy, and personal attraction have greater importance when it comes to people recommendation.
Social media sites and in particular SNSs define different types of explicit (or “articulated”) relationships among their users. The main dimensions of the relationship types are:
Symmetric vs. asymmetric. In some sites, such as Facebook and LinkedIn, a relationship between two users is reciprocated. In such a case, one user typically sends an invitation to connect to another user, who needs to accept the invitation. Once the other user accepts, the two are reciprocally connected on the site. On the other hand, asymmetric relationships, such as on Twitter or Pinterest, allow one user to “subscribe to” or “follow” another user. The other user does not necessarily need to follow the first user back and thus many asymmetric relationships are formed. Typically, user approval is not required in order to follow them.
Confirmed or non-confirmed. Some of the sites require the other side’s agreement for connecting or following, while others do not. Typically, symmetric networks require such confirmation and as long as it has not been received, no connection exists. Asymmetric networks do not typically require a confirmation and any user can choose to follow any other user, however there are exceptions to these norms.
Ad-hoc vs. permanent. Some of the sites encourage connection for an ad-hoc purpose, such as for people to meet at an event or partner for a joint task, while others encourage a long-term relatiships that is meant to last over months and years.
The site’s domain. The domain of the SNS has typically an important influence on the formed network. For example, Facebook is typically used for maintaining social relationships with friedns and aquaintances, while LinkedIn is a professional network meant for maintaining business relationships with colleagues and partners. The goals and charactersitics of a connection in each of these sites are therefore different, as they would be in SNSs for other domains, such as travel, art, cooking, question and answering, etc.
The different characteristics of people relationships in the different sites require different recommendation techniques. For example, a recommender for people to connect with on Facebook may seek to recommend familiar people, while a recommender for people to follow on Twitter may recommend people the user is interested in, even if they are not familiar. Recommending “celebrities” or popular people is probably a better strategy for a follower-followee network than for a friendship network.