Yesterday I covered a recent study that explored how things appear popular within our social networks. It highlighted the so called majority illusion whereby something can be popular within your personal social network but not at all popular outside of that. This can distort your thinking into believing something is much more popular than it actually is.
The authors suggest therefore that the key to making something appear popular is finding the individuals within a network that are excessively popular, ie those that distort the mean.
A second study, by researchers at City College of New York looked at how such influential nodes can be identified. It rather counter-intuitively suggests that smaller can often be better than bigger.
“The problem of identifying the minimal set of influential nodes in complex networks for maximizing viral marketing in social media, optimizing immunization campaigns and protecting networks under attack is one of the most studied problems in network science,” the authors say. “So far, only intuitive strategies based mainly on ‘attacking’ the hubs to identify crucial nodes have been developed.”
The study, which is published in Nature, set out to tackle the issue via the creation of a scalable algorithm, which they tagged the Collective Influence algorithm. The authors believe it trumps anything currently in use in networks such as Twitter or Facebook.
“Through rigorous mathematical calculations, employing optimal percolation and state-of-the-art spin glass theory, we solved the optimal collective influence problem in random networks,” they say. “We show that the set of optimal superspreaders radically differ and is much smaller than that obtained by all previous heuristics rankings, including PageRank, the basis of Google.”
In layman’s terms
So what does this all mean? Well, it suggests that the most influential people in a network (in terms of the diffusion of ideas and content) aren’t actually the most connected people in that network.
Instead, the most influential people are weakly connected individuals who are strategically surrounded by hierarchical coronas of hubs. Thus, the authors suggest that the traditional vision of influence needs to be changed from bigger is better to smaller is smarter.
Nice to see a counter point to the usual rhetoric about everything being around a large social network. Thanks for sharing.
there's different sorts of importance within network structures – the ones above would be the nodes exhibiting high betweenness centrality – they are often known as information brokers or controllers as they have the ability to promote information diffusion into communities or restrict it.