Ever since I was introduced to the world of startups innovating the Health Care space (coined “Health 2.0”) I found myself surrounded by the most beautiful minds of the 21^st century: entrepreneurs and investors, engineers and thought leaders, infinitely passionate about using new ideas and technology to fix and solve problems in our modern Health Care. Their efforts result in saved lives, enhanced research experience, healthier lifestyle and new approaches to patient-oriented services. They influence, lead and inspire the cohort of fellow professionals and entrepreneurs. For the longest time I was trying to understand how to measure the influence of those who help others.

http://dictionary.reference.com/browse/influence

in·flu·ence

/ˈɪn flu əns/ Show Spelled [in-floo-uh ns] Show IPA noun, verb, -enced, -enc·ing.

–noun

1. the capacity or power of persons or things to be a compelling force on or produce effects on the actions, behavior, opinions, etc., of others: He used family influence to get the contract.

2. the action or process of producing effects on the actions, behavior, opinions, etc., of another or others: Her mother’s influence made her stay.

3. a person or thing that exerts influence: He is an influence for the good.

Recently I came across 2 sites that measure influence in the social media: Klout.com and Peer Index.net. While their approach is not identical, the influence scores are measured in terms of social outreach on the most popular social networks: Facebook, Twitter and LinkedIn. The methodologies vary as well: different variables, normalization criteria, noise filtration, etc. Those who are interested in the details can look it up here http://corp.klout.com/kscore and here http://www.peerindex.net/help/scores .

Understandably, in the age of social media and proliferating apps, social networks make their data accessible via API to engage a greater number of external services.  As a result, it is relatively cheap to get access to the vast amount of data, including social graphs, posts, comments, etc. Most entrepreneurs have a digital presence. Yet, few have the legacy. So let’s see how these people measure up to the “influencer” status:

What strikes me first is that Peerindex is either really bad at sampling or they are way off in their methodology. President Obama has an index of 60 and Mr. Godin just stinking 32. Moreover, not only don’t the absolute numbers make much sense, their ratios (which are a much better measurement of relevance and correlation) do not make sense either. What about Klout? While our President passed the mark of being higher than anybody on that short list, a score of 86 (out of 100) during my time at NYU was a B- at most. Who then would be worthy of an A on Klout ? Let’s talk about Esther Dyson, a Veteran of the Internet, the Brain, the Heart and the Soul of Health 2.0 movement, the Voice of integrity who was ranked in the top 11 world greatest entrepreneurs of all times by Msnbc. The Klout score of 46? Really? Compared to the real “Thought Leader” Paris Hilton with the score of 84? Either the Klout had uncovered the sociological problems in our society or…these analyses are not mature enough to produce good results.

It seems to me that computations based solely on the social network activity are unreliable and even quite comical when one’s influence on society is measured by the number of status updates. What if one influential multi-millionaire who set a precedent for future generations with his innovative approach to business and investment does not Tweet enough or, God forbid, does not tweet at all? Warren Buffet has only 1 tweet

and he is punished severely in his Influencer score.

I was somewhat relieved to see that Klout is adding “achievements” as the additional data point to measure the influence. It is unclear though how the scores will be affected, what is the achievements’ taxonomy and what would be the weight coefficients for the various types of such achievements.

I’m not opposed to using social networks’ activity as one of computational components in the overall Influence score.  Not using the other components is what creates a problem. Data, such as speaking engagements, success stories, and published books is not easy to find, but is crucial to the analysis. It could be found on Internet, semantically induced or even asked to be filled by the consumer (if having a higher score is important to people, they will take the time to provide this data). Editorial efforts and content curation would help too. Finally, I’m convinced that the better graph analysis would improve the metrics drastically.

My first steps as an entrepreneur were in the company called Information Laboratory where we developed a Software Analysis tool called SmallWorlds, pioneering some of the trends in the modern static analysis. It turned out that one does not need to run the software to see how stable it is. The robustness of code was determined by examining various properties of the structural graph formed between the components of the software. The Social Graph is similar to the Software Graph (this type of graphs called Small World network), and it obeys certain laws and principals. In layman’s terms it means that a graph contains very few nodes with a very high degree of connectivity, called hubs or connectors (Barabasi, “Linked” ) and many nodes with just a few connections. It is the presence of these connectors that create the six degree effect.

http://en.wikipedia.org/wiki/Power_law

Connectors are the true Influencers in our lives. The close proximity to them makes the others more recognizable and more attractive to be connected to. However, just having a social link to the hub is not enough. If I follow President Obama on Twitter it does mean that we are close. A good way to measure the true proximity to the connectors is to analyze the strength and semantic meaning of such connections.

Most of the social networks do not recognize the semantics of connections between people and that makes it very hard to measure the influence. For example, I have 2268 followers. Esther Dyson is one of them, and I follow her as well. We have a 2–way relationship semantically indistinguishable from many others in my 469 mutual followings.  If I could indicate that this mutual relationship is semantically special (mentor, friend, investor, role model) then it would be easier to understand the direction of the influence and the strength of our connection.

What about Authority? Scientists working with famous Hungarian mathematician Paul Erdős came up with the metric called Erdős Number, which in many cases served as the measurement of authority. The index was computed on the basis of co-authoring a research paper with Mr. Erdős, who had the number of 0, or someone who was assigned the number. Anyone who worked with Erdős would get the number of 1, and then the numbers increase by 1 on each additional level of interaction. Someone collaborating with someone who has Erdős number of 1 gets the number of 2 and so forth.

http://en.wikipedia.org/wiki/Erd%C5%91s_number

This not very sophisticated approach illustrates two points:

1. authority of 0 ( the maximum ) could be assigned manually (curated )
2. proximity to the higher authority is important.

Google’s Page Rank is a more sophisticated method for measuring the importance for a page, but is also rooted in the proximity and the direction of influence from more authoritative sources.

I think that Influence and Authority are close enough to be used interchangeably (if you are the influencer then you must be the authority and in reverse). Having said that, I’m not sure what that single number ( Klout score or Peerindex ) represents mathematically and semantically and, moreover, how to compare two scores. Is Paris Hilton almost twice as influential as Esther Dyson? I definitely hope not!  Perhaps in the world of fashion or bar hopping she outperforms Esther on all levels and one would certainly seek a fashion advice from Ms. Hilton.

It would be helpful to choose specific dimensions, the areas of influence (Professional Connections, Living/Friendship Connections, Health-related Connections) when comparing 2 people. It would be even better to let users pick the inputs into the formula and assign the importance of weights. Such measurement would be more accurate, more actionable, less deceiving and definitely less embarrassing.