Maximizing Acceptance of Targeted Friend Requests in Social Networks
Currently, the standard for social networking sites like Facebook and LinkedIn is to recommend new friends to users based on the existing connections they have. Users are given a list of potential friends that are typically friends-of-friends they have on the site already, since those are the relationships that are most likely to form given the principle of triadic closure. However, a group of researchers are looking to change this passive process of combing through a list of potential friends into an active one, allowing users to target others for friendship that they share no immediate mutual friends with.
Yang et al. propose that when actively targeting such an individual for friendship on a social network, the best method for maximizing the probability of request acceptance is to determine a path in the network between the user and their target and then to add each node as a friend sequentially down the path until they have achieve their desired friendship. Even though the user and his target are not directly tied, they almost certainly can be linked in this way because, as we know from class, the global friendship network is largely one giant component, with the vast majority of people in the world tied everyone else through it. This means between a user and any given target, there should be a path that can be taken between them. This is illustrated in the following diagram from the researchers:
In this instance, user s is trying reach their target t by finding a path between them to add friends along that is both short and carries a high probability of acceptance. s must first send a friend request to one of the nodes in the blue circle, and once accepted they must send another to one of their new friends-of-friends in the red circle. Upon acceptance of that second request, they now have direct mutual friends with t, which should hypothetically increase the probability that t will accept s’s friend request, whereas before they would likely have not accepted a random friend request out of the blue. This process operates under the principle of triadic closure as discussed in class; if s and t are both friends with the same third person, then there will be a tendency will be for them to close that triangle as s desires. Triadic closure is used in iterations here to ultimately achieve the end result of becoming friends with a specified individual.
These researchers found this process to be most successful using an algorithm they produced called SITINA that acts to maximize the probability that the request to the target friend will be accepted. They identified two primary factors that were key in determining this probability, the social influence factor, which involves environmental factors such as common friends, and homophily, which concerns similarities in personal characteristics between potential friends. Both of these were taken into account in the algorithm Yang et al. produced, which upon testing was discovered to operate with a higher probability of friend acceptance than when human intuition alone was used to attempt the process of adding friends in a path to the target node. They hope that social networking sites in the future will be able to implement their work as an alternative to the existing passive friend recommendation process.
Reference Article: http://arxiv.org/pdf/1302.7025v1.pdf