Fowler, James. Connected: The Surprising Power of Our Social Netowrks and How They Shape Our Lives. 23 November 2009.
From Driscollwiki
Fowler, James. Connected: The Surprising Power of Our Social Netowrks and How They Shape Our Lives. 23 November 2009
http://www.connectedthebook.com/
Why this research?
Why do people vote?
Imagining what would happen if people in Florida changed their votes?
- Nothing?
- If you vote, you get A. If you don't vote, you get A.
Game theorists, etc. show that it is irrelevant.
Robinson Crusoe model of social science: observing one person's decision to act.
Widower effect
Nicholas A. Christakis, MD, PhD
- Health and behavior, social science
When a person loses their spouse,
- They become more likely to die
Maybe it spreads beyond the one spouse to many other people?
How does behavior spread through social networks?
Gathering data w/ Christakis
Tracking 500 people over time
- Surveyed
- Examined by doctor
Initial estimated cost too high: $25m
- Looking for pilot opportunities
- Found Framingham, MA community health dataset
Framingham, MA community health
5000 people continue to come in year after year
- As few as 10 have dropped out
- Extremely uncommon
Green sheet, used to keep in touch with people over years
- Listed family, friends, workplace
- Kept for contact purposes
- But effectively traced the social networks
Unique to Framingham?
- How has the influx of immigrants affected the study?
Methodological notes
People asked to name "a close friend", some listed more than one
- Non genetic family members recorded as "friends"
Popularity of "Social networks"
Social networks have always been in play
- Online systems: Facebook, Myspace, etc. discursive substitutes for social networks
What is a social network?
Simplest social network: two people make a choice to connect.
- Form a friendship
- Pairs agglomerate to form large interconnected webs
Transitivity: probability that two of your friends are friends with each other
- Friends of friends
- Dense interconnections
How many dates to find Mr/Ms Right?
We shape our networks, our networks shape our experience
- Do you date at random?
- You date according to mutual friend/family recommendation
- Chicago sex survey suggests 2 in 3 relats based on this
Spread of behaviors, characteristics through a network
Obesity in Framingham
Best data in the Framingham set is Body Mass Index (BMI) (weight/height)
- Yielding Obesity variable
Do you need to dump fat friends?
- "Dumpers" actual became more obese
- Losing the social support has a more negative effect
Takeaway: Change cannot happen alone.
Clustering
- False positive: random chance
- Homophily: love of like, birds of a feather, selection effects
- Contextual effects: omitted variables, e.g. McDonald's opens nearby
Directionality of influence
Directionality of friendship
- Implies influence
- Namer / named
- Ego-perceived friend: influential friend has stronger influence
- Mutual friend: amplify the influence on each other
- Alter-perceived friend: almost no influence
Geographic proximity
- Do near friends have stronger influence than distant friends?
Emotional stampedes
- Clustering happy/sad people
Financial contagion
Run on Northern rock in England
- People withdrawing their money based on friends withdrawing their money
- Similar behavior found in stock market among traders
Microfinancing
Social ties guarantee the loan
- Via social support, pressure
- Higher re-pay rate than typical middle-class biz in Bangladesh
Voting Contagion
Typical story: new technology enabled Obama to reach more people, more intimately
- Fowler version: Obama campaign activated existing social ties
Three degrees of influence
Clusters extend to three degrees
- Friend's friend's friend
- Health: smoking, drinking, obesity
- Emotion: happiness, loneliness, depression
Evolutionary conjecture
There is an evolutionary/survival strategy in these group sizes.
Robin Dunbar's number (150+/-50)
Anthropologists doing studies of primate brain size
- Connected to size of social groups
- According to findings: homo sapien could manage social groups of ~100-150
- Evidence Roman centurian (~100) and mondern-day mil unit (~110)
5 x 5 x 5
Close connections: 5
- Three degrees of close connections: 125
Testing with genetics among social network
Heritability:
- Does your DNA effect relationships among your friends?
Social networks in online spaces
Influence requires deep, intimate social contact
- Looser online connects might not be strong enough
Average num of friends offline: 6
- Ave friends on FB: 100
How do you determine the "real" friends on Facebook?
- Use photo tagging
- Tends to reduce "friends" to ~6
- Tends to reproduce the 3-degrees principle
- Clustering among people who do / do not smile in their profile pictures
Movie, music taste spreading
- Movie tastes spread and cluster depending on network position
- Pulp Fiction in the center
- Love Actually in the periphery
Where is individual agency in this story?
Is there freewill?
- Mutual influence
- Individual agencial choices can have network effects
Influence on peers is actually greater than generally believed
- World-changing through behavior modelling
Implications for practice
Effective, productive team requires
- Core group of familiar people with history and
- New people who have diff perspectives, ideas
Echo chamber thinking
- Analysis of political bloggers suggests clustering by party
- But how must online connections be measured differently?
- Is a hyperlink as strong as a friend/acquaintance?
