Triadic Closures in the Recommendation Engine of LinkedIn
https://www.zdnet.com/article/how-linkedins-people-we-may-know-feature-is-so-accurate/
The main purpose of social networks is to facilitate ties being established between different social circles and individuals, and encourage the formation of new relationships. In this article, titled “How Linkedin’s ‘people we may know’ feature is so accurate,” the author Eileen Brown discusses LinkedIn’s (a social media platform for professionals) “people you may know” feature, which is essentially a recommendation engine that aims to recommend to users other users they may wish to connect with. Brown focuses on how the recommendation system typically tends to behave, and describes a particular hypothetical scenario that is often seen in LinkedIn’s recommendation system: if a person named Jim has a connection with you, and Jim also has a connection with another person named Sue, it is highly likely you will receive a recommendation to connect with this person Sue. Essentially, LinkedIn has decided that you might like to connect with Sue because you and Sue have a common factor: Jim. LinkedIn uses email hashes to read who is sending emails to who, and tracks who Jim may be connecting with at a given time. If Jim emails both you and Sue at around the same time and are connected with both of you, LinkedIn will detect this and assume that you and Sue may also like to connect. After receiving a suggestion from LinkedIn to connect with Sue, there is now a higher likelihood of you choosing to connect with Sue and establish a tie with her, as you now are aware of who she is and aware that she is also connected with gym (and the same if Sue received a suggestion to connect with you).
This resource connects to the topic of the Triadic Closure that we have been covering in this course. Essentially, the Strong Triadic Closure Theory states that if there is a strong tie between nodes A and D and a strong tie between nodes A and C, there must be a tie between nodes D and C. Beyond this, even more generally, the principle of triadic closure itself states that if two nodes in a network both have a tie to the same node, then there is an increased possibility that these two nodes will also establish a tie at some point in the future. Specifically in relation to social networks, this principle essentially states that if two people in a network both share a friend in common, then these two people have an increased possibility of also being friends later on. This phenomenon is reflected in LinkedIn’s recommendation algorithm, as it encourages the formation of connection between individuals who have a connection in common with another person; essentially, as discussed above, if person A has a connection with person B and person C, then person B and person C are more likely to also form a connection later on in the future. This phenomenon is directly reflected in the situation discussed in the article, with Jim being person A, and Sue and yourself being person B and person C. Jim is essentially connected to both you and Sue, which (through LinkedIn’s recommendations system) creates a higher possibility of you and Sue also becoming connected.
As discussed in class, triadic closures are favorable because they help to facilitate connections across large networks. There is increased opportunity for two people B and C to meet if they both have a common friend in person A. Moreover, there may be incentives for person A to introduce person B and person C, whether it be for work-related reasons, social reasons, or etc. Moreover, it may also be more likely that person B and person C are compatible and share similarities that may also encourage the formation of a connection/tie between them. Thus, the principles of the triadic closure (a phenomenon we have discussed in class) is reflected in the recommendation system used by LinkedIn (the feature that recommends to users specific people that they may be interested in connecting with).