Dynamic Networks Xiang Zuo Dissertation Defense March ID: 565413
Download Presentation The PPT/PDF document "The Role of Social Ties in" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
The Role of Social Ties in Dynamic Networks
Xiang Zuo Dissertation Defense March 16, 2016
Prof. Adriana Iamnitchi (advisor) Prof. Kingsley A. Reeves, Jr.Prof. John Skvoretz
Committee:
Prof.
Mingyang
Li (chair)
Prof. Yao Liu
Prof.
Yicheng
Tu
Slide2
Social Ties
Social Ties are connections among people for sharing information, feelings, and experiences.2Slide3
Network Dynamics
3
The network topology changes over times
Information/behavior spreads in networks
http
://not-
ionic.tumblr.com
/Slide4
Outline
4
Background & Motivation
Indirect Ties
M
easurement
The Power of
I
ndirect
T
ies
Timing of link formation
Indirect ties and information diffusion
paths
Indirect ties improves performance of Friend-to-Friend systems
Quantifying Cheating
I
nfluence
Cheating is a complex
contagion
Factors that influence cheating
SummarySlide5
Prior Work
5
Social tiesProperties: Strong and weak ties [Granovetter ’73], Homophily [Mcpherson et al. ’01],
Triadic closure [
Rapoport ‘53]
.
Measurements
:
Katz
[Katz ‘53]
,
Jaccard
coefficient
[Salton et al.
’
83]
,
Neighborhood overlap
[Newman ’01], Adamic Adar
[Adamic et al.’03], Tie strength measurements
[Gilbert et al.’09]
.
Network dynamicsNetwork structure: communities growths and changes
[Backstrom et al. ‘06], group stability
[Patil et al. ‘13], shrinking diameter phenomenon [
Leskovec et al. ‘05, kumar et al. ‘10].
Diffusion/Contagion: most diffusions are small and shallow [Newman et al. ‘06], Hubs are firewalls to diffusion
[Watts et al. ’02, Centola et al. ‘07]
, Simple vs. Complex contagions
[
Centola
et al. ‘07]
, Complex contagions exhibit diverse diffusion
patterns
[
Centola
et al.
’
07, Romero et al. ‘11]
.
We know much less about social ties and network dynamics, especially indirect ties and the contagion of negative behavior.Slide6
My Research Contributions
6
Indirect tie measurementDevelop an indirect tie measurementThree advantages over state-of-art metricsIndirect tie in network dynamicsDefine two metrics to measure the relationships between link delays and tie strength.Predict information diffusion pathsContagion of cheating behavior
Empirically reveal that interaction network serves as a channel for cheating contagion.
Quantify factors that influence cheating with large real-world datasets.Slide7
Outline
7
Background & Motivation
Indirect Ties Measurement
The Power of Indirect
T
ies
Timing of link formation
Indirect ties and information diffusion
paths
Indirect ties improves performance of Friend-to-Friend systems
Quantifying Cheating
I
nfluence
Cheating is a complex
contagion
Factors that influence cheating
SummarySlide8
What
Is an Indirect Tie?An indirect tie is a relationship between two individuals who have no direct relation but are connected through other individual(s).
8Slide9
Why Study Indirect Ties?
9
Indirect ties are known to be a strong force shaping the
network.
L
ack
quantitative studies of the influence of indirect
ties in network dynamics,
especially for social distances longer than 2 hops.Slide10
Observations from Sociology
10
The strength
of a direct tie is related to the amount of interactions
.Multiple
types of interactions result in a
stronger
relationship than only with one type.
The strength of an indirect tie
decreases
with the length of the shortest path between the two
individuals.
SS
n
(A, B)
≠
SS
n(B, A).Slide11
Social Strength Metric
11
Example0.330.600.80
0.67
Zuo, The Power of Indirect Ties in Friend-to-Friend storage systems, P2P’14. Slide12
Evaluation: Using Indirect Ties for Link Prediction
12
Can we use a 2(3)-hop indirect tie between nodes that are not directly connected to predict whether a link will form between them?Zuo, The Power of Indirect Ties in dynamic networks, SocInfo’14. Slide13
Indirect Tie Measurements
13
Jaccard Coefficient (J)Adamic-Adar (AA)
Katz
Social Strength (SS)Slide14
Datasets
14
NetworksNodesEdgesEdge weightsObservation TimeTeam Fortress 2 (TF2)2,4069,7201-21,767300 daysInfectious Exhibition (IE)
4102,765
1-19190 days
Team Fortress 2
(TF2
)
is an online gaming
network.
Infectious Exhibition
(IE) is a face-to-face interaction network collected from a 3-month science gallery in Ireland
Slide15
Link Prediction Results
15
ClassifiernNetworkMetricPrecisionAUCDecision Tree2 TF2Social Strength0.75±0.010.77±0.09Adamic Adar 0.71±0.040.71±0.06
Jaccard
0.51±0.07
0.51±0.08
Katz
0.69±0.01
0.68±0.05
IE
Social
Strength
0.84±0.01
0.87±0.01
Adamic
Adar
0.69±0.02
0.70±0.03
J
accard
0.69±0.07
0.68±0.04
Katz
0.66±0.030.65±0.01
3 TF2
Social Strength0.63±0.020.64±0.03
Katz
0.52±0.07
0.54±0.03
IE
Social
Strength
0.65±0.01
0.66±0.01
Katz
0.62±0.05
0.62±0.07
Social strength can measure strength of indirect ties. Slide16
Outline
16
Background & Motivation
Indirect Ties
M
easurement
The Power of Indirect
T
ies
Timing of link formation
Indirect ties and information diffusion
paths
Indirect ties improves performance of Friend-to-Friend systems
Quantifying Cheating
I
nfluence
Cheating is a complex
contagion
Factors that influence cheating
SummarySlide17
Timing of Link Formation
Link formation delay: the interval between the time when the link formation conditions are met and the time when the link forms.
17Is there any connection between the strength of an indirect tie and the delay of link formation? Zuo, The Power of Indirect Ties in dynamic networks, SocInfo’14. Slide18
Link Formation Delay Definition
18Slide19
Tie Classification
19
Classify indirect ties into strong and weak with three criteria:Slide20
Tie Strength vs. Link Delay
20
33% (strong) vs. 7% (weak)Strong indirect ties form direct links quicker both in 2 and 3 hops. Slide21
Outline
21
Background & Motivation
Indirect Ties
M
easurement
The Power of Indirect
T
ies
Timing of link formation
Indirect ties and information diffusion
paths
Indirect ties improves performance of Friend-to-Friend systems
Quantifying Cheating
I
nfluence
Cheating is a complex
contagion
Factors that influence cheating
SummarySlide22
Concerns of Cheating in Real Life
22
The factors that influence cheating remained untested outside of controlled laboratory experiments and small, survey-based studies.Slide23
Cheating in Online Games
23Slide24
Why Study Cheating in Online Games?
24
Cheating is widespread in online games [Pritchard ’00]. In-game behavior closely mirrors real-world social behavior [Szell et al. ‘10].
Ties are supported by real gaming interactions [Blackburn et al.‘13].Slide25
Our GoalsUnderstand what
factors inherently affect contagion of cheating behavior based on sociological and psychological theories.Observing unpunished cheaters In-group vs. out-group influence Social status influences cheatingAwareness of repercussions Neighborhood Structure
25Zuo, Bad Apples Spoil the Fun: Quantifying Cheating in Online Gaming, ICWSM’16. Slide26
DatasetsSteam and Steam Community
Provides the Valve Anti-Cheat (VAC) service> 1.5 million accounts were banned, by 2014 GameMeIntegrates Steam player profilesProvides real-time statistics of players and game servers
26Slide27
Friendship and Co-match Graphs
Friendship graph Players are nodes; edges represent their declared friendship on Steam profilesCo-Match graph Players are nodes; edges represent two players that played in at least one match27
GraphNodesEdges# CheatersObservation Time (days)Friendship3,148,28944,725,277223,527 (7.1%)2,511Co-match167,4321,130,5952,359 (1.4%)32Slide28
MethodologyThe influence of timing condition
Compensates for missing cheating timeTA ± ω < TB (ω = 0, 1, 3, 7, 14, 21 days)28
ABTATB<±ωSlide29
Factor I: Observing Unpunished Cheaters Aggravates Cheating HypothesisObserving unpunished cheaters in action increases the likelihood of cheating.
In-lab experimental supportGino et al. [Gino, Ayal, and Ariely’09] in-lab, controlled experiments.AssumptionsA player cheats in all matches played before he is VAC-banned.All players with whom the cheater played before being VAC-banned noticed he was cheating and not punished.We consider all players who get VAC-banned in an interval that satisfies the influence timing condition were influenced by him.
29Slide30
Factor I: Results
30Slide31
Factor II: In-group vs. Out-group Influence
HypothesisPeople are influenced by members of their groups more than by out-group members. In-lab experimental supportGino et al. [Gino, Ayal, and Ariely’09] in-lab, controlled experiments.AssumptionIn-group members are players in the same team while out-group members are players in different teams.
31Slide32
Factor II: Results
32Slide33
Factor III: Social Status Influences Cheating Hypothesis
Social status influences the decision to cheat.Theory Upper-status individuals are more likely to engage in unethical behavior [Piff et al. ’09].AssumptionSteam level is an indicator of a player’s social status in the gaming world.33Slide34
Factor III: Results
34Slide35
Outline35
Background & Motivation
Indirect Ties
M
easurement
The Power of Indirect
T
ies
Timing of link formation
Indirect ties and information diffusion
paths
Indirect ties improves performance of Friend-to-Friend systems
Quantifying Cheating
I
nfluence
Cheating is a complex
contagion
Factors that influence cheating
SummarySlide36
Summary
Developed new methods of tie measurement based on sociology observationsTreat ties as asymmetric in strength;More accurate tie strength measurement;Quantify indirect tie strength at any social distance. Enabled novel applications in social computingIndirect ties infer timing of link formation; Indirect ties predict information diffusion;Improve data availability in Friend-to-Friend systems.Combined the power of computing, technology and big data to answer fundamental questions in social science
Cheating spreads in interaction-based networks;Tested several factors that motivate cheating at scale.36Slide37
Publications Xiang Zuo, Clayton Gandy, John Skvoretz, Adriana
Iamnitchi. “Bad Apples Spoil the Fun: Quantifying Cheating Influence in Online Gaming ”, Proceedings of 10th International AAAI Conference on Web and Social Media (ICWSM), Cologne, Germany, May 2016 (acceptance rate: 17%).Xiang Zuo, Jeremy Blackburn, Nicolas Kourtellis, John Skvoretz, Adriana Iamnitchi. “The Power of Indirect Ties”, Journal of Computer Communications, 2016.Xiang Zuo, Jeremy Blackburn, Nicolas Kourtellis, John Skvoretz, Adriana Iamnitchi. “The Influence of Indirect Ties on Social Network Dynamics”, International Conference on Social Informatics, Spain, 2014.Xiang Zuo, Jeremy Blackburn, Nicolas Kourtellis, John Skvoretz, Adriana Iamnitchi. “The Power of Indirect Ties in Friend-to-Friends Systems”, IEEE International Conference on Peer-to-Peer Computing, London, 2014.
37Slide38
Other Publications Xiang Zuo, Adriana Iamnitchi
. “A Survey of Socially Aware Peer-to-Peer Systems”, ACM Computing Surveys, 2016.Imrul Kayes, Xiang Zuo, Da Wang, Jacob Chakareski. “To Blog or Not to Blog: Characterizing and Predicting Retention in Community Blogs”, Proceedings of 7th ACM/ASE International Conference on Social Computing (SocialCom), China, 2014.Xiang Zuo, Alvin Chin, Xiaoguang Fan, Bin Xu, Dezhi Hong, Ying Wang, Xia Wang. “Connecting People at a Conference: A Study of Influence between Offline and Online Using a Mobile Social Application”, IEEE International Conference on Cyber, Physical and Social Computing (CPSCOM), 2012.Jeremy Blackburn, Ramanuja Simha, Nicolas Kourtellis, Xiang Zuo, Matei Ripeanu, John Skvoretz, Adriana Iamnitchi. “Branded with a Scarlet “C”: Cheaters in a Gaming Social Network ”, Proceedings of 12th International Conference on World Wide Web, (WWW), 2012.Jeremy Blackburn, Ramanuja Simha, Clayton Long, Xiang Zuo, Nicolas Kourtellis, John Skvoretz, Adriana Iamnitchi. “Branded with a Scarlet “C”: Cheaters in a Gaming Social Network ”, HPDC/SIGMETRICS 2011 Student Posters (Best Student Poster Award).
38Slide39
Q & A
39