Week4

From INLS 490

Revision as of 14:24, 18 September 2007 by Fred (Talk | contribs)
(diff) ←Older revision | Current revision (diff) | Newer revision→ (diff)
Jump to: navigation, search

Week 4

Contents

[edit] Video

Social Networking in Plain English by Lee LeFever, CommonCraft

[edit] Week 4


[edit] Questions

[edit] Network Dynamics, Part II

[edit] Scale-Free Networks

  • Non-random network, as compared to the Erdos-Renyi and Small World Models
  • Growth is organic, with nodes coming to the network over time intervals
  • Nodes demonstrate preferential attachment, meaning they naturally choose to connect with highly-connected nodes (the authorities)
  • The rich-get-richer phenomenon?

[edit] Questions

  • All clear on scale-free networks, preferential attachment, hubs, authorities, etc?
  • Can we articulate how hubs and authorities emerge in the blogosphere, or on the web?
  • Do we understand how search relevance algorithms leverage network models?
  • Questions about network models?

[edit] Golder et. al.

  • Messaging within FB exhibits temporal rhythms that are robust and consistend across campuses and seasons
    • College student weekend - Midday Friday-Sunday
    • College student week - Sun Eve - Friday
  • Grouping effects/clusters at schools - there were similarities in messaging patters at schools
  • Nearly all communication was found to occur between friends, and only a small proportion of friends exchanged messsages (90.6% msgs/87.5% pokes between friends)
  • Compared to email, FB messages are spares .97msgs/week/user
  • Are messages and pokes a satisfactory proxy for gauging and understanding social activities on a large scale? How would you improve this study?
  • 15.1% of friend pairs exchanged messages - what does this say about social network friendship?
  • Pokes bias heavily towards same-school - does this make sense? If so, why?
  • Highest levels of reciprocity in established-friends, same-school messages, lowest reciprocity in non-friend, different-school messages.
  • Facebooking and schoolwork correlation - does this make sense?
  • FB use decreases during traditional periods of "social time" - what does this tell us?
  • No significant shift in patterns over summer time - Facebook as friend-maintenance tool.

[edit] Kumar et. al.

  • Members of a social network
    • Singletons - People who have joined but not added friendships
    • Giant component - people who comprise the mass of the linkages
    • Middle region - Isolated communities that don't interact with the giant component
  • Kumar - the likelihood that two isolated communities will merge is unexpectedly low - how do our graph models explain this?
  • Isolated communities and stars
    • Star - a single "charismatic" individual connecting a group
    • Stars are less important thatn the giant component - they are an edge phenomenon
  • How an individual joins a network
    • They actively seek it out (loners and giant component)
    • They are invited by a friend or colleague (stars, middle region - stars wish to migrate their offline contacts online)
  • Model of biased preferential attachment says that it is easy to find connections in the giant component, but more difficult in the isolated communities.
  • RQ1 - Are friendships reciprocal in online social networks?
    • Flickr: 70.2%
    • Yahoo 360: 84%
    • Friends tend to reciprocate quickly, usually in a day [Image]
    • What do high levels of reciprocity say about a network? What are levels of reciprocity lower in Flickr?
  • RQ2 - How does density in social networks change over time
    • 1) Early adopters join and add lots of friendships
    • 2) More users join and dilute the densite of the pool
    • 3) Organic growth - people narutally add friends [Image]
  • RQ3 What is the dynamics of component formation and evolution in OSN's?
    • Over time, the density fractions change - more singletons = weaker network?
  • RQ4 How do components merge with each other as nodes and edges arrive in OSNs?
    • Singletons merging with the non-giant component (stars)
    • Non-giant component merging with the giant component (organic growth)
  • RQ5 Do the compnents in the middle region have any special structure?
    • Middle region large composed of stars - Flickr 92.8% stars, Y360 88.7% stars
  • RQ6 How does the diameter of the social network behave as a function of time?
    • Avg vs. Effective diameter
    • Flickr - Avg 6.01, Eff 7.61
    • Y360 Avg 8.26, Eff 10.47
    • Diameter has high correlation with density over time
    • In the third stage diameter starts shrinking [Image]
  • RQ7 - Does the giant component have a reasonably small core of nodes with high connectivity?
    • Giant component characterized by well-connected core at the center
  • Model of networks - types of users
    • Passive - Join out of curiosity, don't engage in significant activity
    • Inviters - interested in migrating an offline community into the online space
    • Linkers - full participants


[edit] Presentations

[edit] Quick Links