Your SlideShare is downloading. ×
James Fowler: The Power of Friends: Business Applications of Network Science
Upcoming SlideShare
Loading in...5

Thanks for flagging this SlideShare!

Oops! An error has occurred.


Introducing the official SlideShare app

Stunning, full-screen experience for iPhone and Android

Text the download link to your phone

Standard text messaging rates apply

James Fowler: The Power of Friends: Business Applications of Network Science


Published on

James H. Fowler earned a PhD from Harvard in 2003 and is currently Professor of Medical Genetics and Political Science at the University of California, San Diego. His work lies at the intersection of …

James H. Fowler earned a PhD from Harvard in 2003 and is currently Professor of Medical Genetics and Political Science at the University of California, San Diego. His work lies at the intersection of the natural and social sciences, with a focus on social networks, behavior, evolution, politics, genetics, and big data.

James was was named one of Foreign Policy’s Top 100 Global Thinkers, TechCrunch’s Top 20 Most Innovative People in Democracy, and Most Original Thinker of the year by The McLaughlin Group. Together with Nicholas Christakis, James wrote a book on social networks for a general audience called Connected. Connected has been translated into twenty languages, named an Editor’s Choice by the New York Times Book Review, and featured in Wired, Oprah’s Reading Guide, Business Week’s Best Books of the Year, and a cover story in New York Times Magazine.

We were delighted to host James in skatepark Waalhalla for #projectwaalhalla Social Science for Startups. See our meet up group for more events:

Published in: Technology, Health & Medicine

  • Be the first to comment

  • Be the first to like this

No Downloads
Total Views
On Slideshare
From Embeds
Number of Embeds
Embeds 0
No embeds

Report content
Flagged as inappropriate Flag as inappropriate
Flag as inappropriate

Select your reason for flagging this presentation as inappropriate.

No notes for slide
  • What might cause this clustering? REVIEWAll three are typically present, and it takes effort to tease them apart.We used a variety of strategies to sort this out.
  • With this kind of perspective, I came to see things differently. This network moves. Things flow in it. It changes and evolves. It is resilient to injury. It has a memory. It has a coherence and endurance across time. And so I came to see social networks as living things. And, we used a variety of techniques to study these living networks, to put them under the microscope. For instance one of our tricks was to exploit the directionality of friendship. ASK: ego/alter pairs.
  • The graphic shows the percentage increase in ego’s risk of obesity given that alter becomes obese. So, a friend becoming obese increases an ego’s risk of being obese. This holds if ego nominates alter, but not if alter nominated ego. Esteem is important.This directional data is also important because it suggests that confounding by unobserved factors is not the source of the relationship.And, the effect is gendered -- among friends, spouses, and siblings, further supporting the social nature of the effect at hand. many mechanisms:spread of behaviors – muffins and beer or a spread of norms – the idea of acceptable body sizeHeadline writers had a field day.I want to be clear, of course, that we do not think that our work should justify any sort of prejudice.we started exploring all sorts of other phenomenaAnd eventually, we became interested in emotions.
  • But look at these two different hospitals with a similar number of doctors and patient-days, both in NY.In the LEFT hospital, the specialists are central, but on the right, it is the primary care doctors.On the LEFT:Avg total costs in last 2 years of life: 104,000Avg total inpatient days in last 2 years of life: 48But in the RIGHT hospital.Avg total costs in last 2 years of life: 46,000Avg total inpatient days in last 2 years of life: 25Inpatient days: 270,000#MD’s: 851Mean normalized specialist centrality: 5.3Density: 0.25Inpatient days: 306,000#MD’s: 653Mean normalized specialist centrality: 3.0Density: 0.11
  • The narrow red loops indicate that the members of a dyad both named each other in the name generator. Some of the grey ties (for which only one individual named the other) result from the named individual not having completed the name generator. Note how the visual clusters of individuals all involve some red ties of reciprocation. Heavier individuals (darker green) also appear to be peripheral in the network and/or to cluster together. Our statistical analyses indicate that the peripheral or central position of the individuals is not significantly associated with BMI. The distance that separates individuals in the network, however, does significantly predict BMI. Individuals who are closer together have more similar BMI.
  • The largest connected network includes over 13,000 individuals. It is difficult to visualize such large networks without the graph being saturated by dots. Here we show a randomly selected set of 5461 individuals of the largest connected component.
  • But things diffuse in human populations in a particular way. In general, diffusion is not random.Because we live in networks. And networks have structure and ties – specific, biologically, socially, and mathematically grounded patterns of ties. The existence of these ties provides the actual routes by which things diffuse. Here is what it looks like, for an illustrative network of 105 people and their friendship ties.And there are all kinds of possible connections between people: relatives, co-workers, friends, casual acquaintances or business interactions, sexual relationships, and so on. And sometimes, we have multiple types of interactions with the same people – like have sex and do business with them.Different things spread across different sorts of ties. STDs spread across sexual ties, the flu across personal interactions. Some people are influenced in their smoking behavior by their friends, but in their charitable behavior by their co-workers or neighbors. And so on.
  • This method is not restricted to germs, but to anything that spreads in populations by person-to-person means, including: REVIEWBehaviors could include other health related phenomena: attitudes towards vaccination, vaccination itself, smoking, health information, attitudes and knowledge about safe sex, and so on.The key thing is that for this method to be useful, the thing in question must spread inter-personally, at least in part, and not just affect people independently, via a broadcast mechanism, for instance.
  • Backplane
  • Transcript

    • 1.
    • 2. Who Are Your Friends? Who do you discuss Important Matters with? Who do you spend your Free Time with? Connected
    • 3. One Pair Connected
    • 4. Connected Many Pairs
    • 5. Connected Interconnected
    • 6. Connected Social Network
    • 7. The Power of Friends Connected Friends as Data Friends as Motivators Friends as Multipliers Friends as Sensors
    • 8. The Framingham Heart Study Connected Original Cohort 1948 N = 5,209 Offspring Cohort 1971 N = 5,124 Gen 3 Cohort 2002 N ~ 4,000
    • 9. Obesity Clusters FHS NETWORK Connected
    • 10. Three Degrees Of Association FHS NETWORK Connected
    • 11. HOMOPHILY Causes of Similarity and Clustering INFLUENCE CONTEXT Connected
    • 12. The Spread Of Obesity Connected FROM 1971 TO 2003
    • 13. Spread of Obesity Connected NA Christakis and JH Fowler, “The Spread of Obesity in a Large Social Network Over 32 Years,” New England Journal of Medicine 2007; 357: 370-379 Ego-Perceived Friend Mutual Friend Alter-Perceived Friend Same Sex Friend Opposite Sex Friend Spouse Sibling Same Sex Sibling Opposite Sex Sibling Immediate Neighbor Small Workplace Co-worker 0 100 200 300 PERCENTAGE INCREASE IN RISK OF OBESITY SOCIAL CONTACT
    • 14. 20011971 Smoking Clusters FHS NETWORK Connected
    • 15. Drinking Clusters FHS NETWORK Connected
    • 16. ANGER HAPPINESS Reading Emotions Connected
    • 17. Happiness Clusters FHS NETWORK Connected
    • 18. Generosity Cascades EXPERIMENTAL NETWORK Connected
    • 19. How Do We Take Our Natural Social Networks Online? Connected
    • 20. Connected
    • 21. Online Networks Connected FULL NETWORK
    • 22. Online Networks Connected NO EFFECT!
    • 23. Online Networks REAL FRIENDS Connected K Lewis, J Kaufman, M Gonzalez, A Wimmer, and NA Christakis, “Tastes, Ties, and Time,” Social Networks 2008; 30:
    • 24. Online Networks REAL FRIENDS Connected PLUS FACEBOOK K Lewis, J Kaufman, M Gonzalez, A Wimmer, and NA Christakis, “Tastes, Ties, and Time,” Social Networks 2008; 30:
    • 25. Photo Tagging FACEBOOK Connected
    • 26. Smiling Clusters FACEBOOK NETWORK Smilers Non Smilers Connected
    • 27. Obesity Clusters FACEBOOK NETWORK Connected
    • 28. Viral Voting FACEBOOK NETWORK Connected
    • 29. Viral Voting FACEBOOK NETWORK Connected
    • 30. Viral Voting FACEBOOK NETWORK Connected Direct Effect Indirect Effects 0 50k 100k 150k 200k 250k 300k 350k IncreaseinValidatedVote Friends CloseFriends
    • 31. Connected
    • 32. Connected Initially targeted High influence & high receptiveness Second wave More receptive Third wave Increasing acceptance Measuring susceptibility-- Intervention is more effective Express Scripts
    • 33. Email Data at Healthways -Each node represents an employee -Each line represents >100 emails transferred between nodes
    • 34. BMI Ranks and Obesity at Healthways Red lines show bi-directional ties Grey lines are directed ties Body Mass Index (BMI) > 30 is considered obese
    • 35. Bikewalk Program in Blue Zones by Healthways -Each node represents an individual -Green nodes are predicted adopters for the Bikewalk program
    • 36. Connected
    • 37. Connected Network
    • 39. Connected Population RANDOM PEOPLE
    • 40. Connected Population PEOPLE & FRIENDS
    • 41. Connected Observed Differences in Epidemic Curves CUMULATIVEINCIDENCE OFINFLUENZA DAILYINCIDENCE OFINFLUENZA DAYS SINCE SEPTEMBER 1 DAYS SINCE SEPTEMBER 1 0 20 40 60 80 100 120 0.00000.00040.00080.0012 0 20 40 60 80 100 120
    • 42. Connected PATHOGENS INFORMATION NORMS BEHAVIORS Contagious Outbreaks
    • 43. Connected Twitter Data! 2/3 of the Twittershpere • 476,553,560 tweets • 40,171,624 users • 1,468,365,182 follows June-December 2009 Kwak, H., Lee, C., Park, H., & Moon, S. (2010). What is Twitter, a social network or a news media. Proceedings of the 19th international conference on World wide web, 591–600.  66,935,466 tweets using a hashtag  4,093,624 different hashtags  1,620,896 users using hashtags
    • 44. Connected Sensor vs. control – specific examples
    • 45. Connected Global view of lead times
    • 46. Connected More Active AND More Diverse
    • 47. ConnectedConnected Early Warning, Even Out of Sample
    • 48. Powerful Friends Connected Friends as Data Friends as Motivators Friends as Multipliers Friends as Sensors
    • 49. Connected
    • 50. Connected
    • 51. Connected Realize Your Own Network Power
    • 52. Connected Thank You!