1. Social network positions of trust,
credibility, prototypicality and social
comparison: An examination of
influence factors in an internet
community
Aleks Krotoski
SPERI
University of Surrey
BPS Social Section Conference
7 September 2006
3. Introduction: Persuasion
• Persuasion in online environments
– Lean medium?
– Effects on individual
• Deindividuation, social exclusion, loneliness
– Elaboration Likelihood Model of Persuasion (Petty
&Cacioppo, 1986)
• Peripheral route?
– Source factors
• Central route?
– Message
• Mediator: immersion
4. Introduction: Persuasion
• Persuasion in online environments
– Social Identity Deinviduation Effects (Spears
& Lea, 1992)
• Conformity with a perceived social identity
• Assumptions of anonymity
– Dynamic Social Impact Theory (Latané &
Bourgeois, 2001)
• Cultural homogeneity = proximity
7. Introduction: Persuasion
• Structural properties of persuasion
– Social Network Analysis
• Influence:
– Presence of tie
– Strength of tie
– Social Learning (Bandura et al, 1977)
– Structural Equivalence (Burt, 1987)
• Analytic techniques to pinpoint influential actors
• Measurement process to define network structures
8. Statement of Aims
• This paper examines the contribution of social
network variables as predictors of persuasion.
• Specifically, I look at the different
contributions which communication modes
have on persuasion in an online community
context.
9. Hypotheses
• Ratings of communication network tie strength for
different communication modes (e.g., public, private
and offline) will contribute more predictive power
for estimates of persuasion than a general
communication score.
• Communication tie strength for different
communication modes will be a greater mediator of
persuasion as communication privacy increases.
10. Method: Procedure and Respondents
• Second Life
– Immersive Virtual community
• “Virtual pub” (Kendall, 2002)
• “Third Place” (Deucheneaut & Moore, 2004)
11.
12. Method:
Procedure and Respondents
• Second Life
– Immersive Virtual community
• “Virtual pub” (Kendall, 2002)
• “Third Place” (Ducheneaut & Moore, 2004)
– Virtual identity
• “Avatar”-representation
– Synchronous, typed communication:
• Public communication
• Instant Message communication (Garton et al, 1997)
13.
14. Methods:
Procedure and Respondents
• Online survey
• 3 April 2006 – 8 June 2006
• Sociometric data collection
• 43 Residents
– Age: M= 32.9 years, SD = 8.13
– Offline gender: male 76.7%, female 23.3%
– Online gender: male 67.4%, female 32.6%)
• 657 avatars, 539 scores
15. Methods:
Independent Variables
• Social Network Communication
– 4 questions (α=0.782) (Garton et al, 1997; Correll,
1995)
• General communication
• Online public communication
• Online private communication
• Offline communication
16. Methods:
Dependent Variables
• Prototypicality (Self-categorization theory: Turner et al, 1987)
– One question: SIDE (Spears & Lea, 1991; Sassenberg & Postmes, 2002;
Postmes, 2001)
• Source Credibility (Renn & Levine, 1991)
– Four questions (α=0.862): Perceived expertise, likeability, believability
• Social Comparison (Perez & Mugny, 1996)
– Two questions (α=0.849): ATSCI (Lennox & Wolfe, 1984)
• General Trust
– Four questions (α=0.874):honesty (Renn & Levine, 1991), care
(Poortinga & Pidgeon, 2003), similarity (Cvetkovich, 1999),
trustworthiness (Renn & Levine, 1991)
• Domain-Specific Trust (Renn & Levine, 1991)
– Four questions (α=0.882): objectivity, honesty, perceived expertise,
reliability
18. Results: Single explanatory variable
(General Communication)
y β0 (Std. β (Std. σ 2e Loglikelihood
Error) Error) (fixed model LL)
Prototypicality 0.026 0.305 0.543 1292.354T
(0.101) (0.066) (0.035) (1335.299)
Credibility -0.093 0.519 0.531 1272.354T
(0.102) (0.071) (0.035) (1404.954)
Social Comparison -0.098 0.399 0.408 987.966T
(0.118) (0.064) (0.027) (1132.416)
General Trust -0.135 0.645 0.408 1114.31T
(0.098) (0.064) (0.027) (1345.777)
Domain-Specific Trust 0.035 0.271 0.347 1086.919T
(0.125) (0.055) (0.023) (1141.021)
*N=538; **N=539; σ2e: variance accounted for between avatars; Tp<0.000, df=2
• The predictive power of the estimate of the value of this measure
of General Trust is positively enhanced when we know how often
two people communicate in general.
19. Single explanatory variable:
General Trust & SNC categories
Explanatory Variable β0 (Std. β (Std. σ 2e Loglikelihood
Error) Error) (fixed model LL)
Online Public 0.085 (0.093) 0.370 0.476 1124.182T
Communication (0.052) (0.031) (1345.777)
Online Private 0.070 (0.094) 0.442 0.407 1115.396T
Communication (0.062) (0.027) (1345.777)
Offline 0.070 (0.090) 0.459 0.427 1159.681T
Communication (0.047) (0.028) (1345.777)
N=539; σ2e: variance accounted for between avatars; Tp<0.000, df=2
• Effect of interpersonal closeness on mode of communication (e.g., Garton et al,
1997)
• Offline communication contributes the most to the estimate of General Trust.
Online public communication contributes the least.
20. Results: Multiple explanatory
variables (General Trust)
Explanatory Variable β0 (Std. β1 (Std. β2 (Std. σ 2e Loglikelihood (fixed
Error) Error) Error) model LL)
Online public + online private 0.065 0.104 0.375 0.394 1144.879T
communication (0.121) (0.057) (0.074) (0.026) (1224.182)
Online public + offline 0.059 0.399 0.291 0.332 1057.941T
communication (0.085) (0.051) (0.051) (0.022) (1224.182)
Online private and offline 0.052 0.345 0.328 0.314 1038.486T
communication (0.087) (0.057) (0.046) (0.021) (1115.396)
N=539; σ2e: variance accounted for between avatars; Tp<0.000, df=3; *model rejected on basis of ill-fit
• Greatest improvement to the fit of a model occurs when offline communication
scores are added to the single-variable public communication model
• Adding online private communication to the online public communication model
renders the weight of online public communication insignificant, so this model is
rejected.
21. Summary
• Social network variables as mediators of persuasion
variables
– Empirical assessment of SNA assumptions
– Greatest effects on General Trust
• Communication tie strength’s effect on General Trust
increases as communication becomes more
private/intimate
• Supportive of Garton et al (1997) and others’ social network
analysis work
• Communication mode tie strength effect less
predictive than “general” strength measure
22. Conclusions
• Review of aims
• Implications
– Use of SN measurements in Social Psychology
– Assessing assumptions of cohesion made by Social
Network Analysis
• Further research
– Comparison with different types of network (e.g, trust-
based)
– Larger dataset (currently in collection)
– Network position effects on social influence
23. Thank you
A.Krotoski@surrey.ac.uk
SPERI
University of Surrey
BPS Social Section Conference
7 September 2006