Skip to main content

Social Influence in Networks

Earlier in class we learnt about hubs and authorities, and how search engines score them according to their link structure – in general, the more connected they are to important nodes, the higher score they get. A similar scoring method goes on in our personal lives as well, amongst our social contacts. Each of us wields a certain influence over our friends and acquaintances, and vice versa. The amount of influence others have over us depends on many factors, including their likability, knowledge, authority, popularity, and even eloquence.

Yet, at the end of the day, it seems that how we determine others’ influence is based on a similar algorithm – we first deem all new acquaintances as equally influential. Then as they make recommendations to us we will assign them a score, kept in our mind, based on how much we like their tip. This internal score constantly gets updated as we have more suggestions from them, or even simply have conversations in which we exchange useful information. The more useful or enjoyable their knowledge is, the more influential they’ll be to us.

Similarly, if a person is influential to others, it is likely that we will find them holding more sway over us as well. In the hub-authority model, this translates to the influencer as a hub with a higher score due to his ability to influence others in addition to you. The basic hubs-authority approach is limited in capturing this type of influence, since hubs and authorities get scored based on their relative links, and not their popularity with users. We can take this additional influence into account by making a simple modification to the hub-authority system – allowing hubs to link to each other! Then, on each update loop, the score of the hub will be the sum of the score authorities it links too, and the sum of the hubs that link to it (not those that it links to). This is a useful mix of the scaled PageRank update rule and the hub-authority update rule. We determine a person’s influence based on how good their recommendations are (using hub-authority), and also how much influence they have over others (using PageRank).

The concept of influence is also gradually garnering more attention, especially in the online arena. Services like Klout, Kred and PeerIndex spring up to cater to a common (perhaps competitive) desire to know how influential we are (and who else is more/less influential than us), and also to serve marketers who like to target the high-influence crowd. Yet, with rankings come those who would try to game the system and reach the top more easily. Like the various SEO techniques to boost one’s search ranking, there could be people with multiple identities that all link to each other, agreements with others for mutual links, deliberate attention-seeking postings, and so on. Also, we should note that online influence may not even be indicative of real-world influence.  TIME magazine’s article shows Lady Gaga trailing closely behind President Barack Obama in a quick back-of-the-envelope influence score calculation, but like the article suggests, in the real world Obama is many times more likely to influence others and catalyze change than Lady Gaga.

This leads us to a deeper search for the meaning of influence. It is not merely enough to be engaging others. One needs to stimulate action in others, and also make others spread the word. An article in the NY Times gives us some insight into how companies like Klout compute influence: based on the quality rather than quantity of interactions. This holds true in real life as well; we are influenced more by people that we have deep conversations with, rather than others whom we only interact with on a superficial level.

Wikipedia also identifies some interesting types of influence:

  1. Compliance – agreeing outwardly to conform to peer pressure, related to the need to be liked
  2. Identification – agreeing so as to be more like the person you like, especially for celebrities
  3. Internalization – acceptance because of agreement both internally and externally, related to the need to be right

It seems that companies like Klout, via their measurements, seem to be measuring influence along the lines of identification. Identification results in a mimicking and widespread dissemination of the message, without any significant interaction or personal investment required.  With internalization someone would both be engaged and pass on the message, potentially persuading and influencing others in turn. However, because the message needs to be internalized before it can be passed, influence via internalization is deep but slow. Compliance creates a scenario of mere head-nodding, a low-quality interaction that generates little influence outside the social group.  Using the Lady Gaga and Obama examples from above, we can classify Lady Gaga’s influence as a case of identification, whilst Obama’s influence type will lean more towards internalization. A quick search of these two major figures on Klout reinforces this line of thought – Obama scores a respectable 88, but Lady Gaga has an even higher score of 92. Surprisingly, Justin Bieber is a top scorer with 100 Klout points. Is Justin Bieber really more influential than Barack Obama? Perhaps, in influencing our taste in clothes and music. But in the ways that really matter, in challenging our thinking and our behavior, I would say that Obama is many times more successful.

There are still more ways in which one can exert influence, and also be influenced in the offline and online world. As companies like Klout progress and refine their algorithms, it will be interesting to see if their concepts of influence and power will actually gain traction and change the way we interact as a society, like how social networking changed with websites like Facebook.


Leave a Reply

Blogging Calendar

October 2011