7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining Privacy Concerns vs. User Behavior in Community Question Answering Imrul Kayes USF Nicolas Kourtellis Telefonica Research Francesco Bonchi Yahoo Labs Adriana Iamnitchi USF
19
Embed
Privacy Concerns vs. User Behavior in Community Question Answering
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy Concerns vs. User Behavior in Community Question Answering
Imrul KayesUSF
Nicolas KourtellisTelefonica Research
Francesco BonchiYahoo Labs
Adriana IamnitchiUSF
2
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
• Functionalities:– Q/A posts & comments, social networking,
leaderboards, and more.• Why do we need them?– Not all web-searches are successful!– Complicated / intricate questions!
3
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Motivation• Privacy of user data: important but unresolved issue• Privacy vs. platform usability:– Public content is helpful but users would prefer privacy
• Privacy-aware users are more engaged (e.g., on FB)– Is there a sweet-spot?
Priv
acy
Platform usability
4
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Main Idea• Utilize users’ modifications on privacy settings as
a proxy of privacy concerns• Group users into privacy categories:
• Keep active users with more than 10 Q/A (68%)• 4 privacy settings:– All public (84.4%), PU– Hide content (Questions/Answers) (2.5%), QA– Hide network (Followers/Followees) (0.9%), N– Hide content & network (12.2%), PR
PublicSemi-PrivatePrivate
*Kayes et al. “The social world of content abusers in Community Question Answering“, WWW’2015
9
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Retention
• Semi-private users have higher inter-event time• Private users have lower inter-event time
10
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Social Circles
Indegree = #followers, Outdegree = #followees
• Privacy-concerned users have larger social circles• Followers or followees
11
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Accomplishments
• Private & semi-private users have more points• Privacy-concerned users contribute more in YA
quantitatively
12
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Accomplishments
best answer percentage (BAP
• Privacy-concerned users contribute more in YA qualitatively
13
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Best Answers
• Privacy-concerned users select more Best Answers
14
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
• Community feedback via thumbs up and down:
• Privacy-concerned users have more average thumbs on Best Answers
Privacy & Best Answer Quality
15
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Abuse Reporting
• 90% of reports submitted by 8% of users
• Privacy-concerned users post more valid reports
16
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy & Deviance
• Deviance can indicate user engagement*• Private > semi-private > public users’ deviance scores
* Kayes et al. “The social world of content abusers in community question answering“, WWW’2015
17
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Summary1. 87.20% public profiles (default)2. Privacy-concerned users are more engaged:– higher retention– more social network contacts– contribute more & better content– higher perception on answer quality– better citizens in terms of reporting abuses– higher platform engagement (via deviance)
18
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Suggestions for better CQA platforms
• If privacy settings modified:– Likely commitment to the platform– Question recommendation and routing– Community moderation
• If privacy settings unmodified:– Incentivize for increased participation & retention
• Privacy settings in CQA sites– Prediction & recommendation via ML techniques
7th IEEE/ACM International Conference on Advances in Social Networks Analysis and Mining
Privacy Concerns vs. User Behavior in Community Question Answering