Stronger Networks for Job Search
https://www.fastcompany.com/3041091/this-startup-intends-to-be-an-intimate-version-of-linkedin-with-more-meaningful-work-connect
Networks are extremely prevalent in finding new jobs, as connections are often the easiest way to find an entry to a company. However, sites such as LinkedIn often become bloated, as people connect with so many people. Their networks can often reach into the thousands of people, devaluing their connections to their peers. Thus, when people actually go to find a job, there is much less motivation to help a connection, as you may not actually have a strong tie to that person. A new startup, Shapr, aims to solve this problem by limiting the number of connections in a user’s network. The idea behind this limitation is that your connections actually mean something on this new platform. With a small, tightly nit network, users can actually look to their connections for real help in a job search.
This idea of limiting the size of a network is very interesting. Since the motivation is ultimately about strengthening the ties in your network, it can take advantage of some network theory that LinkedIn may not be able to use effectively. For example, the idea of Strong Triadic Closures will apply significantly in Shapr networks. If a user A is connected to B and C on Shapr, there is a very high chance that B and C want to be connected, as Shapr’s ties are very strong. Thus, Shapr can utilize this Strong Triadic Closure property to build very strong recommendation systems to create the best 50 person network possible for their users. On the contrary, a site like LinkedIn may not be able to leverage this idea effectively. Since users of LinkedIn often have weaker ties to their connections, the Strong Triadic Closure property may not apply at all (it only applies given two strong connections). Thus, LinkedIn would be implementing a sometimes ineffective heuristic, leading to poor recommendations. While both these sites probably use Machine Learning for their recommendation systems, I think this principle still would apply to a learned system. Ultimately, Shapr will have an easier time learning a recommendation system, as all their network ties are much stronger.