1 / 29

Studying Systems of Communities

Studying Systems of Communities. Brian Butler University of Pittsburgh Xiaoqing Wang University of Pittsburgh. Motivation Benefits & costs Commitment Conversation. Individuals. Microsoft.public.xml. Group. Structure Culture & ideology Type & identity.

plato
Télécharger la présentation

Studying Systems of Communities

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Studying Systems of Communities Brian Butler University of Pittsburgh Xiaoqing Wang University of Pittsburgh

  2. Motivation • Benefits & costs • Commitment • Conversation Individuals Microsoft.public.xml Group • Structure • Culture & ideology • Type & identity Existing Research on Online Community Sustainability • Existing online community research typically consists one individual or one community at a time.

  3. Microsoft.public.xml Online Communities are not Isolated Source: http://netscan.research.microsoft.com

  4. Strategies for Examining Systems of Communities • Visualization • Cases • Impact of communities on one another • Aggregate impact measures • Dyadic impacts and structure • Impact of communities on individual behavior • Aggregate impact measures • Dual mode person/community networks • System structure models • “Market share” models

  5. Visualizations Source: http://netscan.research.microsoft.com

  6. Cases • Focus on one system of communities • Descriptive and exploratory analysis • Useful for: • Clarifying how you will define the system or systems of groups • Selecting a strategy for larger scale analyses • Determining what the available data actually allows you to describe (and how hard it really is to construct the measures) • Example: USENET newsgroups related to breast cancer

  7. Defining the System(s) of Interest • Formal Names • Peers in the USENET hierarchy • Managed (sort of) • Membership Niches • Other places that the people go/are • Emergent • Content Niches • Other places that the messages go/are • Emergent

  8. Formal Names • Names act as boundary objects • Boundary Object: “[A]rtifacts, documents, terms, concepts…around which a [group] can organize their interactions” (Thompson, 2005, quoting Wenger, 1998, p. 105). • They also serve to define the neighborhood likely to be visible to individuals seeking content or people related to a topic • Operational question: What is the (a?) relevant neighborhood for alt.support.cancer.breast?

  9. %cancer% Neighborhood • 26 groups (alt.support.cancer.breast is not included) • Post Characteristics • Total Messages: 139812 Posts • Total Replies: 105768 Replies • Total Starts: 34044 (18271 Successful) • Thread Start Ratio: 53.6% • Average Posts/Month/Group: 220 • Maximum Posts/Month/Group: 2014 • Minimum Posts/Month/Group: 1 • People Characteristics • Total Unique Posters: 17960 (+ one record for “0 author”– ?) • Average Authors/Month/Group: 61 • Average 1 Time Posters/Month/Group: 35 • Average Returnees/Month/Group: 16

  10. %cancer% System – Posting Activity Correlation: -0.484 (p = 0.225, N.S.)

  11. %cancer% System – Author Dynamics %cancer% System alt.support.cancer.breast Pearson Correlations Avg Authors/Month : 0.738 (p = 0.037) Avg 1 Timers/Month: 0.806 (p = 0.016) Avg Returnee/Month: 0.562 (p = 0.181)

  12. %cancer% Neighborhood – Author Dynamics (II) Correlations: First Seen – 0.506 (p = 0.201); Last Seen – 0.216 (p=0.607); 1st Long – 0.507 (p = 0.200)

  13. Descriptive Analysis of the System Raises Questions… • Is it competition between one group and others? • Posting activity hints that it might be… • Or common external influence • 9/11/2001 might have suppressed activity in non-support groups and increased it in alt.support.cancer.breast --------------------------------------------------------------- • Is it a case of “there goes the neighborhood? • Positive significant correlations among average group participation measures and general downward trend suggest that it might be…. • Or increased fragmentation • More active groups would potentially lower the average number of participants in each one.

  14. Membership-Based Systems • Comembership ‘links’ • Two newsgroups have a comembership link when they share a common poster • This treated as a dyadic network even though it may not be dyadic (could also be mixed mode graph). • Non-directed ties; directed ties can represent movement • Co-memberships ego network vs. comembership network centered a target

  15. alt.support.cancer.breast (ASCB) Membership System • Full membership network • Any newsgroup that shares a member with alt.support.cancer.breast any time during the study period • 143,710 newsgroups (+ the group itself) • 39,275 newsgroups share a single member with ASCB • 70,307 (49%) have 5 or fewer members in common with ASCB • The most extensive membership tie is alt.support.cancer with with 1502 participants in common (this is 37% of the participants in ASCB)

  16. alt.support.cancer.breast (ASCB) Shared Membership Definitions • Filter 1: Participant level of activity • Logic: Groups that share active members are have a stronger link than those that share peripheral or minimal member • Measurement: • Active span (First Seen – Last Seen) • Number of posts • Active span > 1 day is used (In ASCB 1680 (41%) are meet this criteria) • Implementation: Individuals must be active in both newsgroups • Results: • 90,958 newsgroups (+ the group itself) • 39,275 newsgroups share a single member with ASCB • 59,701 (66%) have 5 or fewer members in common with ASCB • The most extensive comembership tie is alt.support.cancer with with 433 participants in common (this is 27% of the active participants in ASCB)

  17. alt.support.cancer.breast (ASCB) Shared Membership Definitions • Filter 2: Number of Shared Participants • Logic: The more active members that two group share, the stronger the connection between them • Measurement: • Number of unique authors shared by two newsgroups across the entire study period • Implementation: A cutoff 55 or more shared active participants (~3% of ASCB’s active participants) was used because it results in a comembership neighborhood that was the same size (# of groups) as the %cancer% neighborhood. • Results: • 26 newsgroups (+ the group itself) • Maximum shared participants (433 – alt.support.cancer) • Minimum shared participants (55) • Average shared participants (103)

  18. Membership System Description • 26 groups (alt.support.cancer.breast is not included) • Post Characteristics • Total Messages: 7,582,551 Posts • Total Replies: 6,456,678 Replies • Total Starts: 1,125,873 (674,344 Successful) • Thread Start Ratio: 60% • Average Posts/Month/Group: 5152 • Maximum Posts/Month/Group: 95,862 • Minimum Posts/Month/Group: 1 • Poster Characteristics • Total Unique Posters: 304,089 (+ one record for “0 author”– ?) • Average Authors/Month/Group: 535 • Average 1 Time Posters/Month/Group: 263 • Average Returnees/Month/Group: 164

  19. Membership System – Posting Activity Correlation: 0.014 (p = 0.975)

  20. Membership System – Author Dynamics ASBC Member System alt.support.cancer.breast Pearson Correlations Avg Authors/Month : -0.765 (p = 0.027) Avg 1 Timers/Month: -0.476 (p = 0.233) Avg Returnee/Month: -0.638 (p = 0.089)

  21. ASBC Member Links Over Time(Shared Authors for Each of the 27 Groups)

  22. Content System • Crossposting network -- which we couldn’t assess with this data set. Source: http://netscan.research.microsoft.com

  23. Formal and Membership System Overlap • 26 groups in Formal Name System • 26 groups in ASBC Member System • Only 5 groups in both alt.support.cancer.testicular sci.med.prostate.cancer alt.support.cancer.prostate alt.support.cancer sci.med.diseases.cancer

  24. Larger Sample Research Approaches • Impact of communities on one another • Aggregate impact measures • Dyadic impacts and structure • Impact of communities on individual behavior • Aggregate impact measures • Dual mode person/community networks • System structure models • “Market share” models

  25. System Characteristics • Size of the system • #Crossposting groups • #Comember groups • Activity in the system: • AvgPostCount in the CoMember system • AvgAuthorCount in the CoMember system • OverlapRate = AvgSharedAuthorPerNG/ActiveAuthors • Question: How does the size of the system and the activity in it affect a focal group?

  26. Communities Affects (Aggregate)- Findings • Content-based and relationship-based competition both reduce the ability to retain members. • High bond-based communities are more affected by characteristics of the member system • Low bond-based communities are more affected by the characteristics of the content–based system. • Communities with high levels of contribution are more affected by the characteristics of the content–based system

  27. Community System Characteristics Effects on Individual Behavior • Size of the content-based system is negatively associated with an individual’s likelihood of ongoing participation in a given community • The larger the content-based community system, the lower the impact of an individual’s recent contribution activity on their likelihood of ongoing participation in a given community • Possible explanation: Competition interacting with individual investment in their relationship with a community

  28. Other Approaches to Studying Systems of Communities • Impact of communities on one another • Dyadic impacts and structure • Impact of communities on individual behavior • Dual mode person/community networks • System structure models • “Market share” models

  29. Various Research and Practical Questions • Competition vs. Complementary • Life cycle issues • System life cycle • Community life cycle • Member life cycle • Role of communities within systems --------------------------------------------------------------------- • To link or not to link? • How many communities should we have (and does it matter)?

More Related