MS&E 189 Lecture Notes - Lecture 6: Homophily, Social Search
MS&E 189 1/21/2016
Small World Phenomenon
Hypothesis: Any two people in this world can be linked through a relatively small number of
intermediate acquaintances
● A good introduction to thinking about network features density, distance, and logics
cohesion, brokerage
● Also helpful in thinking about how people perceive and leverage their social network.
Milgram’s Small World experiment
-Found 6 degrees of separation between citizens of rural areas (Kansas, Nebraska) with
stockbroker living in Boston
Implications:
-Not all paths equal due to:
1. Geography, foci location
2. Funneling
3. Homophily
Limitations:
● Does not capture structure of network, rather some form of social (decentralized) search
● Participants did not forward the folder to all contacts - not necessarily shortest path
● Small sample, non random start and end points, low response rate
Kevin Bacon - most central actor?
Not true: with 100 acquaintances per person, everyone should have 100^6 people in six
degrees
True: People will share multiple acquaintances (overlapping networks), some have larger,
others have smaller networks.
● Tendency towards closure and homophily, resulting in the redundancy of acquaintances
Watts and Strogatz Model
-Small worlds have high clustering, closure, homophily
-Implies regular connectivity pattern
-Short paths: actors can access others
-Implies random connectivity pattern
-Regular networks can be “enhanced” with just a little randomness
How small worlds emerge:
1. Start with regular highly clustered network
2. Add few random links
find more resources at oneclass.com
find more resources at oneclass.com
Document Summary
Hypothesis: any two people in this world can be linked through a relatively small number of intermediate acquaintances. A good introduction to thinking about network features density, distance, and logics cohesion, brokerage. Also helpful in thinking about how people perceive and leverage their social network. Found 6 degrees of separation between citizens of rural areas (kansas, nebraska) with stockbroker living in boston. Not all paths equal due to: geography, foci location, funneling, homophily. Does not capture structure of network, rather some form of social (decentralized) search. Participants did not forward the folder to all contacts - not necessarily shortest path. Small sample, non random start and end points, low response rate. Not true: with 100 acquaintances per person, everyone should have 100^6 people in six degrees. True: people will share multiple acquaintances (overlapping networks), some have larger, others have smaller networks. Tendency towards closure and homophily, resulting in the redundancy of acquaintances.