Applying the principles of social networks to estimate the spread of influenza

Influenza, also frequently referred to as the flu, is a common infectious and viral illness that spreads itself through, for example, coughing and sneezing. The flu is distinguished from the ‘common cold’, as it’s caused by dissimilar viruses, is more severe and often lasts longer. Approximately 375.000 people worldwide, dies from the flu each year. So how is the flu connected to social networks?

Figure 1:


We define a social network (group) to be a social structure, that consists of social individuals with interactions and ties with one another. Figure 1 displays the conceptual model of how a social network may look like.  As previously stated, the flu can be contracted either through direct or indirect contact with an infected individual, through coughs and sneezes or through the process of inhaling infected airborne particles. Scientists have performed a great number of studies on the subject of how social networks are correlated to the spreading of diseases. One in particular, by Nicholas Christakis and James Fowler (Christakis, Fowler, 2010), tracked how the influenza spreads in social networks. They used their knowledge of what social networks are and works in order to predict how the flu spreads. In order to determine the prediction, they adopted the paradox of friendship. The friendship paradox, first observed by the sociologist Scott Feld in 1991, states that ‘any given individual has less friends than what their friends has’. How is this true? The reason behind this paradox comes down to the fact that people who are identified as a friend, tends to be a more popular figure in a social group with stronger connections. They are what we define as a central individual in a social network.

Now, how do we make the connection between 1. The flu and 2. Social networks? From what we know, the flu is a disease spread through people and their interactions. One could make the assumption that an individual’s social network impacts the probability of becoming infected by the flu. Well-connected people in large social groups should, by theory, become infected before any individual with a smaller and less connected social group, even if they are in the same network. We theorise that this is due to the fact that central individuals in a network has more ties and interactions with other people. The core assumption is that the more people you interact with, the higher the probability of getting sick becomes.

Christakis and Fowler is the authors of numerous books about social networks. Some subjects they discuss are the dynamic spread of happiness, the spread of obesity, and the underlying power in social networks. They state that individuals that are considered central points in a social groups can act as an observation mark. This means that by observing central individuals in social networks, we can determine how the flu spreads. Mathematically speaking, these people are more likely to become infected by anything that’s spreading throughout the social group before anyone else.

A valuable question arises in this context; how do we identify which individuals are central in a social group? By simply looking at the network as a set of nodes and edges, the number of edges a node consists of, does not alone define the centrality of that given node. In fact, we have to look beyond just the number of edges and look at the quality of those edges. Are the set of connected nodes also central nodes or just nodes with a set of small connections? Is a node connected to multiple central nodes, the most central node?

Abhijit Banerjee from MIT theorised that the easiest approach in identifying the central individual in a social group is by asking people to identify the person in their social network that’s best able to spread gossip (Banerjee 2016). When people were asked, surprisingly most named a central figure in their social group, despite the fact that most people are not able to visualize the overall social structure of their network. One approach in achieving this is through constructing a simulated social network, then providing specific information to a given node and observe how that information spread throughout the network over time. The node that manages to spread the information furthest, can be defined as a central node. A commonly used and humorous analogy is the Six Degrees of Kevin Bacon, that states that ‘any two people on earth are six or fewer acquaintance links apart’. Researchers claim that these social networks are not static, but in fact grows over time. An interesting study done on old Australians claim that having a large social network late in your life can increase your life expectancy (Giles, 2005). This goes to show how short you have to look in order to find a connection between two people and how these social ties matter.

The experiment Christakis and Fowler performed on 744 Harvard undergraduates made significant findings. They asked the students to name their friends and from that constructed social networks. The authors then regularly checked the students for any flu symptoms. The report, ‘Social Network sensors for early detection of contagious outbreaks’, was composed after the experiments and shows that individuals that were identified as central nodes in a network contracted symptoms of the flu several weeks, from 14-69 days, before the flu peaked in the control group. This leads us to the conclusion that we can use social networks to determine when a disease will become widespread, by looking at central individuals in social networks.


  1. Nicholas A. Christakis & James H. Fowler, Social Network Sensors for Early detection of Contagious Outbreaks, September 15. 2010 [research article]. Available from: Accessed at: 06.11.17
  2. Lynne C Giles, Gary F V Glonek, Mary A Luszcz, Gary R Andrews, Effect of social networks on 10 year survival in very old Australians: The Australian longitudinal study of aging, August 2005 [research article]. Available from  Accessed at: 06.11.17
  3. Abhijit Banerjee, Gossip: Identifying Central Individuals in a Social Network, February 14. 2016 [research paper] Available from: Accessed at: 06.11.17