The Convergence of Social The Convergence of Social and Technological Networks and Technological Networks By Jon Kleinberg By Jon Kleinberg Presented by Jonathan Presented by Jonathan Willitts Willitts
Dec 21, 2015
The Convergence of Social and The Convergence of Social and Technological NetworksTechnological Networks
By Jon KleinbergBy Jon Kleinberg
Presented by Jonathan WillittsPresented by Jonathan Willitts
22
ContentsContents OverviewOverview
Why look into convergence of networksWhy look into convergence of networks
How it is possibleHow it is possible
The small world phenomenonThe small world phenomenon
Social contagion and the spread of ideasSocial contagion and the spread of ideas
Further directionsFurther directions
33
OverviewOverview Past decade has witnessed a coming together of the Past decade has witnessed a coming together of the
technological networks that connect computers on the technological networks that connect computers on the Internet, and the social networks that have linked humans Internet, and the social networks that have linked humans for millenniafor millennia
Sites such as Facebook, MySpace, Wikipedia, digg, Sites such as Facebook, MySpace, Wikipedia, digg, del.icio.us, YouTube and flickr have all developed from thisdel.icio.us, YouTube and flickr have all developed from this
Growing Pattern of movements to:Growing Pattern of movements to:• Form connections with others, Form connections with others, • Build virtual communitiesBuild virtual communities• Engage in self expressionEngage in self expression
Data being generated by these online worlds allows us to Data being generated by these online worlds allows us to observe human social interaction in a way never previously observe human social interaction in a way never previously possiblepossible
44
Why look into convergence of Why look into convergence of networks?networks?
Images of social networks offer glimpses of Images of social networks offer glimpses of everyday life from an unconventional vantage everyday life from an unconventional vantage point, such as:point, such as:• Flow of information through an organisationFlow of information through an organisation• Disintegration of social groups into rival factionsDisintegration of social groups into rival factions
Science advances when we can take something Science advances when we can take something that was once invisible, and make it visiblethat was once invisible, and make it visible
This is now happening with social networks, and This is now happening with social networks, and social processessocial processes
55
How it is possible?How it is possible? Collecting data on social networks has previously been Collecting data on social networks has previously been
hard work, requiring:hard work, requiring:• Extensive contact with group being studiedExtensive contact with group being studied• Limited to much smaller groups (tens or hundreds of Limited to much smaller groups (tens or hundreds of
individuals)individuals)
Online social interactions however leave digital tracesOnline social interactions however leave digital traces
This allows us to collect and use minute by minute data This allows us to collect and use minute by minute data on tens of millions of people to:on tens of millions of people to:• Monitor the way people seek connections and form Monitor the way people seek connections and form
friendships (Facebook)friendships (Facebook)• Coordinate with and engage in creative expressions Coordinate with and engage in creative expressions
(Wikipedia or flickr)(Wikipedia or flickr)• Observe news stories / witness controversy gathering Observe news stories / witness controversy gathering
around communities (bloggers)around communities (bloggers)
66
The Small-World Phenomenon / The Small-World Phenomenon / Decentralised Search Decentralised Search
Stanley Milgram experiments in the 1960sStanley Milgram experiments in the 1960s
Milgram asked a few hundred people in Boston and the Milgram asked a few hundred people in Boston and the Midwest to direct a letter towards a stockbroker in Midwest to direct a letter towards a stockbroker in MassachusettsMassachusetts• Gave each the targets name, address and occupationGave each the targets name, address and occupation• Only mail letter to someone they knew on first name basisOnly mail letter to someone they knew on first name basis• Mail with instructions to pass the letter onMail with instructions to pass the letter on
Successful letters reached the target in a median of 6 stepsSuccessful letters reached the target in a median of 6 steps
Play and then film: Six Degrees of Separation Play and then film: Six Degrees of Separation
Game: Six Degrees of Kevin BaconGame: Six Degrees of Kevin Bacon
88
Other work (MSN Messenger)Other work (MSN Messenger) Social network built from quarter-billion instant Social network built from quarter-billion instant
messaging accounts messaging accounts
Connected individuals who engaged in a conversation Connected individuals who engaged in a conversation over a 1 month periodover a 1 month period
Average length of the shortest path between any two Average length of the shortest path between any two people was 6.6people was 6.6
Very close to Milgram’s figure, obtained completely Very close to Milgram’s figure, obtained completely differentlydifferently• Milgram asked targets to forward letter to people they think Milgram asked targets to forward letter to people they think
might know the targetmight know the target• Studying “random” links made by conversations, participant Studying “random” links made by conversations, participant
unaware. unaware.
99
Milgram FindingsMilgram Findings Milgram experiment not only showed that paths existed, Milgram experiment not only showed that paths existed,
but also that people knew how to find thembut also that people knew how to find them
People asked to find the target couldn’t know the course People asked to find the target couldn’t know the course the letter would take, or whether it would get therethe letter would take, or whether it would get there
The fact that so many did shows social networking ability to The fact that so many did shows social networking ability to funnel information towards far away targetsfunnel information towards far away targets
Caveat: the experiments were more successful the more Caveat: the experiments were more successful the more affluent and socially accessible the targetaffluent and socially accessible the target
Many chains failed to completeMany chains failed to complete
1111
Social contagion and the spread of Social contagion and the spread of ideas ideas
Milgram’s experiments aimed to focus Milgram’s experiments aimed to focus messages on particular targetsmessages on particular targets
Information on social networks tends to Information on social networks tends to radiate out in many directions at onceradiate out in many directions at once
Rumours, political messages and online Rumours, political messages and online videos can spread contagiously from videos can spread contagiously from person to person similar to an epidemicperson to person similar to an epidemic
1212
Diffusion of innovations Diffusion of innovations There is a pattern by which people influence each There is a pattern by which people influence each
other over periods of time (both online / offline) to other over periods of time (both online / offline) to • form new political and social beliefs, form new political and social beliefs, • adopt technologies and adopt technologies and • change personal behaviourchange personal behaviour
Process known as “Diffusion of Innovations”Process known as “Diffusion of Innovations”
Whilst outcomes may be clearly visible, the inner Whilst outcomes may be clearly visible, the inner workings have remained unknownworkings have remained unknown
1313
Recent studies of contagionRecent studies of contagion The probability of purchasing of books, DVDs and The probability of purchasing of books, DVDs and
music from a large online retailer increased with music from a large online retailer increased with the number of email recommendations a the number of email recommendations a potential customer receivedpotential customer received
Probability of joining a group in an online Probability of joining a group in an online community increases the more friends are community increases the more friends are already a member of the groupalready a member of the group
There is however a “diminishing returns” pattern There is however a “diminishing returns” pattern in which the marginal effect of each successive in which the marginal effect of each successive friend decreases. friend decreases.
1414
Contagion as a design principle Contagion as a design principle Decentralised search problem at the heart of the small Decentralised search problem at the heart of the small
world phenomenonworld phenomenon
Contagion in networks serves as a design principle for Contagion in networks serves as a design principle for spread of informationspread of information
Early distributed computing work proposed notion of Early distributed computing work proposed notion of “epidemic algorithms” where information updates would be “epidemic algorithms” where information updates would be spread between hosts using a probabilistic contagion rulespread between hosts using a probabilistic contagion rule
Lead to further research based on the fact that these Lead to further research based on the fact that these algorithms can be highly robust and simple to configure at algorithms can be highly robust and simple to configure at each nodeeach node
1515
Further directionsFurther directions Whilst most of this presentation has been focussed on dynamic Whilst most of this presentation has been focussed on dynamic
behaviour of individuals in social networks, an important behaviour of individuals in social networks, an important complimentary areas is how the structure of the network itself complimentary areas is how the structure of the network itself evolves over timeevolves over time
Recent studies of large datasets have shed light on important Recent studies of large datasets have shed light on important principles of network evolution such as:principles of network evolution such as:• Preferential attachment:Preferential attachment: in which nodes that already have many in which nodes that already have many
links will tend to acquire further ones at a greater ratelinks will tend to acquire further ones at a greater rate
• Triadic closure:Triadic closure: links are more likely to form between two people links are more likely to form between two people when they have a friend in commonwhen they have a friend in common
• Densification effects:Densification effects: in which the number of links per node in which the number of links per node increases as the network growsincreases as the network grows
• Shrinking diameters:Shrinking diameters: in which the number of steps in the shortest in which the number of steps in the shortest path between nodes can actually decrease as the total number of path between nodes can actually decrease as the total number of nodes increasesnodes increases