Friends Only:Examining a Privacy-Enhancing Behavior in FacebookFred Stutzman and Jacob Kramer-Duffield, UNC
Changes in SNS adoption landscapeSource: Pew Internet and American Life Project, 2009, 2010
Changes in SNS landscapeSource: Pew Internet and American Life Project, 2009, 2010
Changes in FacebookTerms ofservicechangeOpen toregionalnetworksOpen toselectworkplacesOpen to most collegesEnd ofregionalnetworksPublicprofilesBeaconNews FeedintroducedFacebookat Harvard2004		2005		2006		2007		2008		2009
Changes in Facebook
Implications of ChangeShift from common identity to common bonde.g. Ren, Kraut & Kiesler, 2007; Sassenberg, 2002Uses, flows of social capital and social supporte.g.  Lampe et al. 2008Management of disclosureboundaries and contexts
Managing ContextsPresence of multiple social groupsBehavioral strategiesMental strategies“Least common denominator”Source:Lampinen et al., 2009
Context TensionConnections across status and power boundariesPropriety, work, familyHarms from crossed boundariesInadvertent disclosures across contextsSource:Skeels and Grudin, 2009
Putting Context in ContextFriendster“Burners, gay men, and bloggers”MyspaceTeens and mirror profilesTwitterPractical obscuritySource: boyd, 2006, 2007, Stutzman and Hartzog, 2009
Going Friends-OnlyMoving profile from network-viewable to viewable only by FriendsImplications:Searching, browsing and findingNetworked information contribution, appsPerson perception, relational formation
Going Friends-OnlyPublicly available information: Name, city, gender, photo, list of friends, fan pages, networks, friends list (temporarily)
Going Friends-OnlyUse Privacy Settings: 83.2%Friends-only: 58.3%2005: 5% (+/-1), 2006: 12% (+/- 1), 2007: 23% (+/- 4), 2008: 58% (+/- 5)
Going Friends-OnlyRQ:  What factors are associated with having a friends-only profile?Boundary regulation theories of privacyAltman, Derlega and ChaikinApplied in HCI, Social ComputingPalen and Dourish, Tufekci, DwyerApplied in Organizations (Allen et al.), Education (Mazer et al.), Communication (Petronio, Child et al.)OthersSelf
Petronio’s CPMProcess of Communication Privacy ManagementRule DevelopmentWho gets to know whatBoundary coordinationApplying rules-in-contextBoundary turbulenceReacting to events, managingand regulating rulesSource:Petronio, 2003
Study DesignWeb-based surveyPilot test, n=76Full survey, June 2008, n=494Response Analysis94% of respondents usedFacebookAnalytical sample, n=444Males, minorities, youngerstudents under-representedAnalysis is unweighted, FPCnot applied
Analytic PlanRQ:  What factors are associated with having a friends-only profile?DV: Having a friends only profileModelsNull and demographic baselinesRule developmentBoundary coordinationBoundary turbulenceEvaluationWithin comparison likelihood ratio testBetween comparison with AIC, BIC, ROC
Baseline ModelDemographic factorsSchool year, gender, ethnicityFacebook use factorsNumber of friends, length of membership, time spent on site
Rule DevelopmentWho do you tell what?H1: Strong ties = inward-focusedH2: Random ties = outside-focusede.g.Strahilevitz, 2005EvaluationNo items, blocks significantProblems: Lack of variation due to normative orientation in friending practices; InstrumentationScale: Lampe et al., 2006, 2008; Ellison et al. 2007
Boundary CoordinationCoordinating permeability rulesFriends can know my gossipPeople around campus shouldn’tSalient audiencesIntended audienceExpected audience
Boundary CoordinationExploring the effects of “expectancy violations”Violations effects codedAnalyzed simultaneouslyEvaluationWeak tie expectancy violations result in 3.31 increase in odds of beingfriends-onlyThe meaningful externalboundary?
Boundary TurbulenceMaintenance and (re)negotiation of disclosure boundariesPredictorsConversant privacy scaleAdvised someone to change FB profile/picture…Wall management scaleRemoved wall post (self/other)CovariatesGender, profile management effortAlphas: Conversant: .69, Wall: .73, Effort: .79
Boundary TurbulenceExploring effects of interpersonal privacy managementAnalyzed simultaneouslyEvaluationIndividuals who engage inhigher amounts of conversant management morelikely to have friends-only profileGender, effort not significant
Model ComparisonCompare predictive strength of the modelsNon-nested comparisons employ AIC, BIC, ROC, etc.All models represent incremental improvements
GoalsExplore the process of going friends-onlyApply theories of boundary regulation, and Communications Privacy Management, in SNS contextIdentify and prioritize models for further exploration
ImplicationsDynamic identification of functional network boundaryOpportunity to create non-reciprocal communication interfacesIncrease opportunities for collaboration around privacy settings
Thank you!Fred StutzmanFred.stutzman@unc.eduhttp://twitter.com/fstutzmanJacob Kramer-Duffieldjkramerd@gmail.comhttp://twitter.com/jaykaydee

Friends Only: Examining a Privacy-Enhancing Behavior in Facebook

  • 1.
    Friends Only:Examining aPrivacy-Enhancing Behavior in FacebookFred Stutzman and Jacob Kramer-Duffield, UNC
  • 2.
    Changes in SNSadoption landscapeSource: Pew Internet and American Life Project, 2009, 2010
  • 3.
    Changes in SNSlandscapeSource: Pew Internet and American Life Project, 2009, 2010
  • 4.
    Changes in FacebookTermsofservicechangeOpen toregionalnetworksOpen toselectworkplacesOpen to most collegesEnd ofregionalnetworksPublicprofilesBeaconNews FeedintroducedFacebookat Harvard2004 2005 2006 2007 2008 2009
  • 5.
  • 7.
    Implications of ChangeShiftfrom common identity to common bonde.g. Ren, Kraut & Kiesler, 2007; Sassenberg, 2002Uses, flows of social capital and social supporte.g. Lampe et al. 2008Management of disclosureboundaries and contexts
  • 8.
    Managing ContextsPresence ofmultiple social groupsBehavioral strategiesMental strategies“Least common denominator”Source:Lampinen et al., 2009
  • 9.
    Context TensionConnections acrossstatus and power boundariesPropriety, work, familyHarms from crossed boundariesInadvertent disclosures across contextsSource:Skeels and Grudin, 2009
  • 10.
    Putting Context inContextFriendster“Burners, gay men, and bloggers”MyspaceTeens and mirror profilesTwitterPractical obscuritySource: boyd, 2006, 2007, Stutzman and Hartzog, 2009
  • 11.
    Going Friends-OnlyMoving profilefrom network-viewable to viewable only by FriendsImplications:Searching, browsing and findingNetworked information contribution, appsPerson perception, relational formation
  • 12.
    Going Friends-OnlyPublicly availableinformation: Name, city, gender, photo, list of friends, fan pages, networks, friends list (temporarily)
  • 13.
    Going Friends-OnlyUse PrivacySettings: 83.2%Friends-only: 58.3%2005: 5% (+/-1), 2006: 12% (+/- 1), 2007: 23% (+/- 4), 2008: 58% (+/- 5)
  • 14.
    Going Friends-OnlyRQ: What factors are associated with having a friends-only profile?Boundary regulation theories of privacyAltman, Derlega and ChaikinApplied in HCI, Social ComputingPalen and Dourish, Tufekci, DwyerApplied in Organizations (Allen et al.), Education (Mazer et al.), Communication (Petronio, Child et al.)OthersSelf
  • 15.
    Petronio’s CPMProcess ofCommunication Privacy ManagementRule DevelopmentWho gets to know whatBoundary coordinationApplying rules-in-contextBoundary turbulenceReacting to events, managingand regulating rulesSource:Petronio, 2003
  • 16.
    Study DesignWeb-based surveyPilottest, n=76Full survey, June 2008, n=494Response Analysis94% of respondents usedFacebookAnalytical sample, n=444Males, minorities, youngerstudents under-representedAnalysis is unweighted, FPCnot applied
  • 17.
    Analytic PlanRQ: What factors are associated with having a friends-only profile?DV: Having a friends only profileModelsNull and demographic baselinesRule developmentBoundary coordinationBoundary turbulenceEvaluationWithin comparison likelihood ratio testBetween comparison with AIC, BIC, ROC
  • 18.
    Baseline ModelDemographic factorsSchoolyear, gender, ethnicityFacebook use factorsNumber of friends, length of membership, time spent on site
  • 19.
    Rule DevelopmentWho doyou tell what?H1: Strong ties = inward-focusedH2: Random ties = outside-focusede.g.Strahilevitz, 2005EvaluationNo items, blocks significantProblems: Lack of variation due to normative orientation in friending practices; InstrumentationScale: Lampe et al., 2006, 2008; Ellison et al. 2007
  • 20.
    Boundary CoordinationCoordinating permeabilityrulesFriends can know my gossipPeople around campus shouldn’tSalient audiencesIntended audienceExpected audience
  • 21.
    Boundary CoordinationExploring theeffects of “expectancy violations”Violations effects codedAnalyzed simultaneouslyEvaluationWeak tie expectancy violations result in 3.31 increase in odds of beingfriends-onlyThe meaningful externalboundary?
  • 22.
    Boundary TurbulenceMaintenance and(re)negotiation of disclosure boundariesPredictorsConversant privacy scaleAdvised someone to change FB profile/picture…Wall management scaleRemoved wall post (self/other)CovariatesGender, profile management effortAlphas: Conversant: .69, Wall: .73, Effort: .79
  • 23.
    Boundary TurbulenceExploring effectsof interpersonal privacy managementAnalyzed simultaneouslyEvaluationIndividuals who engage inhigher amounts of conversant management morelikely to have friends-only profileGender, effort not significant
  • 24.
    Model ComparisonCompare predictivestrength of the modelsNon-nested comparisons employ AIC, BIC, ROC, etc.All models represent incremental improvements
  • 25.
    GoalsExplore the processof going friends-onlyApply theories of boundary regulation, and Communications Privacy Management, in SNS contextIdentify and prioritize models for further exploration
  • 26.
    ImplicationsDynamic identification offunctional network boundaryOpportunity to create non-reciprocal communication interfacesIncrease opportunities for collaboration around privacy settings
  • 27.
    Thank you!Fred StutzmanFred.stutzman@unc.eduhttp://twitter.com/fstutzmanJacobKramer-Duffieldjkramerd@gmail.comhttp://twitter.com/jaykaydee

Editor's Notes

  • #19 Gender emerges as the only significant variable; the odds of a male having a friends-only Facebook profile are 59% of the odds of a female having a friends-only Facebook profile.Interpreting the odds ratio, the addition of ten Facebook friends is associated with a 1% increase in the odds of having a friends-only Facebook profile.
  • #21 Paired T-Tests
  • #22 Expectancy violations were effects coded