1. tie strength
predictors of tie strength often: total activity better:  ?
predictors of tie strength often: total activity better:  ? . dynamics of phone calls for 20M users (2.5B ties) . intensity no good to represent short-term ties  Dynamics of ties in massive communication networks Esteban Moro Egido,  Univ Madrid
predictors of tie strength often: total activity better:  ? . dynamics of phone calls for 20M users (2.5B ties) . intensity no good to represent short-term ties  Dynamics of ties in massive communication networks Esteban Moro Egido,  Univ Madrid answer: stability it's well-know: strong ties persist Also see “Predicting Tie Strength with Social Media” [CHI'09]
predictors of tie strength often: total activity better:  ? Me:  one can verify definitions of tie strength from the structure of the net. Why?
predictors of tie strength often: total activity better:  ? Me:  one can verify definitions of tie strength from the structure of the net. Why?  We know that: 1) Weak ties are bridges 2) Strong ties are embeeded
2. time
2) extract a reasonable communication net  1) make mobility prediction how many days of mobile data one needs to:
2) extract a reasonable communication net  Answer: 14 days! 1) make mobility prediction how many days of mobile data one needs to:
2) extract a reasonable communication net  Answer: 14 days! 1) make mobility prediction how many days of mobile data one needs to: Jean Bolot,  Sprint Mining Call and Mobility Data to Improve Paging Efficiency in Cellular Networks  [Mobicom'07]  We study the mobility patterns of cell phone users and develop mobility profiles... extract global trends from an evolving graph=  = segment dataset into time windows  (one static graph for each window) Jean Bolot,  Sprint Gautier Krings , UC Louvain/MIT
‘ breakfast’ and then ‘dinner’, ’ skipped breakfast’ and then ‘headache’ (it takes usually 6h) ‘ peace’ (it’s popular on sundays) ‘ happy hour’ (it’s popular not only on fridays) ‘ love you’ (popular during weekends) ‘ pregnant’ (on thursdays!!!) ‘ exhausted’ (usually at night) future: build a markov chain mdl to study  correlations   540M tweets in US Built a visualization tool for temporal occurencese.eg: Michael Macy,  Cornell Temporal trends of expression in twitter
3. performance & network diversity
Network Diversity and Economic Development Nathan Eagle, 1,2,*  Michael Macy, 3,4  Rob Claxton 1,5 diverse personal networks are linked to strong local economy
Future: put forward  whys  to avoid this... Network Diversity and Economic Development Nathan Eagle, 1,2,*  Michael Macy, 3,4  Rob Claxton 1,5 “ So keep building those social networks. It’s not a total waste of time. It just might be your own personal economic stimulus package.” diverse personal networks are linked to strong local economy
There's a group of connected people solving a problem. What's the best way of connecting those people? linear  fully connected Humans balance between  exploration  and  exploitation   David Lazer , Harvard
There's a group of connected people solving a problem. What's the best way of connecting those people? linear  fully connected Peak but no heterogeneity slow Humans balance between  exploration  and  exploitation   David Lazer , Harvard
There's a group of connected people solving a problem. What's the best way of connecting those people? linear  fully connected Peak but no heterogeneity slow Humans balance between  exploration  and  exploitation   David Lazer , Harvard Duncan Watts , Yahoo Study on  MechTurk . How performance is affected by net topologies and payoffs. Good news:  assignments are random and controlled
Causality [science 2006]  Duncan Watts , Yahoo
Causality [PNAS 2009]  Sinan Aral , NYU Stern/MIT
Good recommender systems promote homophily and kill diversity!
Good recommender systems promote homophily and kill diversity!  Kate Erlich , IBM Idea:   connect people based on 5 types of brokerage (coordinator, gatekeeper, etc.) Inspired by “Structures of Mediation: A Formal Approach to Brokerage in Transaction Networks” '89
Me:  distinguish brokers, embedded nodes, and hubs
4. Multiplex networks
face-to-face contacts Sandy Pentland , MIT “ friends and family ” 100 phones to mit members in a “residence”  surveys at different times– monthly, weekly, and asynchronously  study the subnetworks of those people (those whose religion is A, those living in floor B, those who have hobby C)
face-to-face contacts Sandy Pentland , MIT questions asked:  1) how influence (e.g., happiness) flows across those subnetworks  2) how to nudge people and and how to measure effectiveness (app store) 3) how friendship forms 4) how people react if they are able to control their personal data
JP Onnela,  Harvard
JP Onnela,  Harvard
2. time 1. tie strength  3. performance & net diversity 4. multiplex networks

Netsci10 report

  • 1.
  • 2.
  • 3.
  • 4.
  • 5.
    predictors of tiestrength often: total activity better: ?
  • 6.
    predictors of tiestrength often: total activity better: ? . dynamics of phone calls for 20M users (2.5B ties) . intensity no good to represent short-term ties Dynamics of ties in massive communication networks Esteban Moro Egido, Univ Madrid
  • 7.
    predictors of tiestrength often: total activity better: ? . dynamics of phone calls for 20M users (2.5B ties) . intensity no good to represent short-term ties Dynamics of ties in massive communication networks Esteban Moro Egido, Univ Madrid answer: stability it's well-know: strong ties persist Also see “Predicting Tie Strength with Social Media” [CHI'09]
  • 8.
    predictors of tiestrength often: total activity better: ? Me: one can verify definitions of tie strength from the structure of the net. Why?
  • 9.
    predictors of tiestrength often: total activity better: ? Me: one can verify definitions of tie strength from the structure of the net. Why? We know that: 1) Weak ties are bridges 2) Strong ties are embeeded
  • 10.
  • 11.
    2) extract areasonable communication net 1) make mobility prediction how many days of mobile data one needs to:
  • 12.
    2) extract areasonable communication net Answer: 14 days! 1) make mobility prediction how many days of mobile data one needs to:
  • 13.
    2) extract areasonable communication net Answer: 14 days! 1) make mobility prediction how many days of mobile data one needs to: Jean Bolot, Sprint Mining Call and Mobility Data to Improve Paging Efficiency in Cellular Networks [Mobicom'07] We study the mobility patterns of cell phone users and develop mobility profiles... extract global trends from an evolving graph= = segment dataset into time windows (one static graph for each window) Jean Bolot, Sprint Gautier Krings , UC Louvain/MIT
  • 14.
    ‘ breakfast’ andthen ‘dinner’, ’ skipped breakfast’ and then ‘headache’ (it takes usually 6h) ‘ peace’ (it’s popular on sundays) ‘ happy hour’ (it’s popular not only on fridays) ‘ love you’ (popular during weekends) ‘ pregnant’ (on thursdays!!!) ‘ exhausted’ (usually at night) future: build a markov chain mdl to study correlations 540M tweets in US Built a visualization tool for temporal occurencese.eg: Michael Macy, Cornell Temporal trends of expression in twitter
  • 15.
    3. performance &network diversity
  • 16.
    Network Diversity andEconomic Development Nathan Eagle, 1,2,* Michael Macy, 3,4 Rob Claxton 1,5 diverse personal networks are linked to strong local economy
  • 17.
    Future: put forward whys to avoid this... Network Diversity and Economic Development Nathan Eagle, 1,2,* Michael Macy, 3,4 Rob Claxton 1,5 “ So keep building those social networks. It’s not a total waste of time. It just might be your own personal economic stimulus package.” diverse personal networks are linked to strong local economy
  • 18.
    There's a groupof connected people solving a problem. What's the best way of connecting those people? linear fully connected Humans balance between exploration and exploitation David Lazer , Harvard
  • 19.
    There's a groupof connected people solving a problem. What's the best way of connecting those people? linear fully connected Peak but no heterogeneity slow Humans balance between exploration and exploitation David Lazer , Harvard
  • 20.
    There's a groupof connected people solving a problem. What's the best way of connecting those people? linear fully connected Peak but no heterogeneity slow Humans balance between exploration and exploitation David Lazer , Harvard Duncan Watts , Yahoo Study on MechTurk . How performance is affected by net topologies and payoffs. Good news: assignments are random and controlled
  • 21.
    Causality [science 2006] Duncan Watts , Yahoo
  • 22.
    Causality [PNAS 2009] Sinan Aral , NYU Stern/MIT
  • 23.
    Good recommender systemspromote homophily and kill diversity!
  • 24.
    Good recommender systemspromote homophily and kill diversity! Kate Erlich , IBM Idea: connect people based on 5 types of brokerage (coordinator, gatekeeper, etc.) Inspired by “Structures of Mediation: A Formal Approach to Brokerage in Transaction Networks” '89
  • 25.
    Me: distinguishbrokers, embedded nodes, and hubs
  • 26.
  • 27.
    face-to-face contacts SandyPentland , MIT “ friends and family ” 100 phones to mit members in a “residence” surveys at different times– monthly, weekly, and asynchronously study the subnetworks of those people (those whose religion is A, those living in floor B, those who have hobby C)
  • 28.
    face-to-face contacts SandyPentland , MIT questions asked: 1) how influence (e.g., happiness) flows across those subnetworks 2) how to nudge people and and how to measure effectiveness (app store) 3) how friendship forms 4) how people react if they are able to control their personal data
  • 29.
    JP Onnela, Harvard
  • 30.
    JP Onnela, Harvard
  • 31.
    2. time 1.tie strength 3. performance & net diversity 4. multiplex networks