MEASURING TRUST IN ONLINE SOCIAL NETWORKS
The Effects of Network Parameters on the Level of Trust in Trust Games
with Incomplete Information
Parvaneh Afrasiabi Rad, Svante Edzen and Soren Samuelsson
Department of Computer Science, Electrical and Space Engineering, Luleå University of Technology, Luleå, Sweden
Keywords: Trust Level, Online Social Network, Trust Game, Incomplete Information, Network Parameters, Simulation.
Abstract: The most currently popular method for assessing trust in online social networks is Trust Game. The major
studies in this area have established results formed into hypotheses for the effects of a number of network
parameters on the extent to which individuals would place trust on each other. However, hypotheses for the
effects of a few number of network parameters, such as Indegree, are not deducible since the restrictive
game-theoretic assumptions that are imposed into the model do not let any such evidence available. To relax
the game-theoretic assumptions, we develop a model for games with incomplete information, based on a
game-theoretic model developed by Buskens (1998), and conduct a series of computer simulation of a
model of Iterated Heterogeneous Trust Games (IHTG). We compare the results with those of Buskens’
(1998) model and introduce Link-Strength as a new network parameter to investigate. Our results show a
positive effect of both Indegree and Link-strength on the level of trust in a noisy environment. In addition,
we come to a conclusion that current models can be fooled by the existing noise in the context of
information transmission, such as inactive users in our case.
1 INTRODUCTION
The effects of the communication channel and the
characteristics of interactions are the major
interesting areas in CMC studies on trust in social
networks (Riegelsberger, Sasse et al. 2003). In this
respect, scholars strive to derive hypotheses for the
effects of the structure of the communication
channel, which represents the patterns of
interactions, on the level of trust. The most currently
popular method for assessing trust in online social
networks is making use of “Trust Games” (Camerer
and Weigelt 1988; Kreps 1992; Kreps 1996; Snijders
1996; Dasgupta 2000; Buskens 2002).
The game-theoretic model for measuring trust
threshold developed by Buskens (1998) has been
acclaimed to be the first model that provides
hypotheses about both individual and global network
parameters, in addition to deriving hypotheses about
non-network parameters and their interaction effects
with network parameters (Buskens 2002, chap 3).
The model and analysis are applied for Iterated
Heterogeneous Trust Games (IHTG). The outcome
of the model suggests that network parameters
influence the extent to which trustors would place
trust on the trustee mainly through outdegree and
density, whereas other network parameters have not
been concluded to be influential on trust threshold
(Buskens 1995; 1998; Buskens 2002, chap 3;
Buskens and Raub 2008). However, such hypotheses
have been driven based on a Pareto optimal
equilibrium in trigger strategies for games with
complete information. Such context requires
postulating several assumptions and considerations
in various aspects in developing the model, whereas
most real situations are not governed by such
circumstances.
Focusing attention, Buskens’ (1998) game-
theoretic model takes a counterintuitive assumption
that the information is ‘always and accurately’ (p.
286) passed from one entity to another in the
network. The authenticity of the information,
however, is not promised in social networks. Such
facts that oppose the assumption of the reliability of
the information are referred to as ‘noise’ (ibid, p.
286). Refusing to incorporate the noise in the
context of information transmission due to the
restrictive game-theoretic assumptions, has led this
model to be unable to derive hypotheses about the
learning effects of embeddedness on trust, hence
531
Afrasiabi Rad P., Edzen S. and Samuelsson S..
MEASURING TRUST IN ONLINE SOCIAL NETWORKS - The Effects of Network Parameters on the Level of Trust in Trust Games with Incomplete
Information.
DOI: 10.5220/0003903005310539
In Proceedings of the 8th International Conference on Web Information Systems and Technologies (WEBIST-2012), pages 531-539
ISBN: 978-989-8565-08-2
Copyright
c
2012 SCITEPRESS (Science and Technology Publications, Lda.)
leaving network parameters such as indegree as un-
influential on the level of trust (ibid). In addition, the
game-theoretic assumptions impose some
circumstances to the context of information
transmission which is far from the realistic
environments. To extend the game-theoretic model,
so that predictions about the effects of network
parameters in the context of noise can be derived,
we should relax a number of selective assumptions.
In this study, we make assumptions about
incomplete information, and the existence of noise in
the context of information transmission. The new
context would let us investigate the effects of
additional network parameters, i.e. Indegree and
Link-strength. It is also closer to the context of real
social networks, in the sense of both assumptions
and structure. We boost the influence of the
circumstances of incomplete information by taking
sample networks that are very large in size, in order
to conceal the structure of the network from the
players. Simulations are run on 6 networks that are
sampled from Youtube, for their structure to be
closer to reality. The results are further analyzed to
derive hypotheses about the effect of a new set of
different network measures, indegree and link-
strength, on the level of trust in the context of noise.
We utilize the game-theoretic model, developed by
Buskens (1998), for its validity and make alterations
to its assumptions to form a new context.
The following section starts with the framework
of this study by introducing the sources of noise and
its effects. It also includes details of the model that is
developed in this study with the assumptions of
games with incomplete information. It follows with
a presentation of the simulation method in addition
to the results of the regression analysis of the
simulated data. The hypotheses driven from the
analysis of the results can be found at the end of that
section.
2 THE MODEL
As it has been previously discussed, the game-
theoretic model does not circumstantiate intuitive
hypotheses as far as indegree is concerned. Aside
from indegree effects, noteworthy is the reason for
such, which is due to the assumption that the
information about the trustee’s behavior is, always,
positive and accurately transmitted in the network.
After all, the role of indegree is more conclusive in a
“noisy” environment (1998, p. 286; Buskens 2002,
p. 90). A trustor could be reluctant to sanction a
trustee if she obtains information about the abuse of
trust form one trustee and she cannot verify the
information herself. She will decide to execute
sanction only in case she receives such negative
information repeatedly. The extent to which trustors
receive information in a network, indegree,
thereupon will have an effect on trust (Buskens
2002, chap 3). Regardless of the ways different
trustors could interpret incoming information about
the trustee’s behavior to further forward it to the
next trustor in the game, it is reasonable to conclude
that, in the context of noise, a trustor with larger
indegree is more certain about the accuracy of
information in hand by virtue of obtaining
information from multiple sources (Buskens 1998).
Also, the positive information about the behavior of
the trustee is more reliable when it is transmitted
through such trustor. Accordingly, an inactive user –
i.e. an actor with a large indegree and a small
outdegree, is a source of likely enough reliable
information, while not contributing to the flow of
information in the network.
We do not aim to manipulate the game-theoretic
context, deviate from trigger strategies, alter the
equilibrium or introduce a new one.
Notwithstanding, we will assume that subsisting
information about the behavior of the trustee is not
always considered to be accurate and neither is
perfectly transmitted between trustors. The origin of
such information, the amount and the order and
structure of its transmission is not a matter of
concern. Trustors, indeed, follow trigger strategies to
decide upon placing trust, however are triggered not
merely by receiving information about the abuse of
trust from the previous trustor, but as they are
“infected” by the information that they receive.
These arguments are valid only in the context of a an
information diffusion model that incorporates the
idea that ‘an actor does not receive information as a
package relinquished by the sender, but rather is
“infected” by the information given to him’
(Buskens and Yamaguchi 1999, p. 5). Therefore, the
information can be considered as unreliable not only
due to inaccuracy, but also resultant from
misinterpretation, information distortion during
transmission, etc.
2.1 Assumptions
Buskens’ (1998) game-theoretic model proposes that
Outdegree and Density are two network parameters
which predict almost all variance that could be
attributed for the trust threshold. These two factors
are weighted by the parameters of the game and
considerations of the equilibrium. Still, to satisfy the
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
532
assumption that users are triggered by the impact of
information, it is important to find out to what
degree they are infected. The answer to this question
cannot be explained only by Outdegree and Density,
since the amount of information obtained does not
mean that the receiver is certainly affected by it. In
order to estimate how much a piece of information
can infect a user, we include an additional network
parameter, namely Link-Strength, into the model to
measure the strength of a tie between two trustors. If
the relationship between two trustors is strong, one
can be influenced even with the lowest amount of
information obtained from the other. In other words,
users with some close friends, which are
characterized by strong friendship links, are more
likely to influence (or be influenced by) them than
those who hold many friendship bonds but almost no
close friends, provided that the two groups create
comparable amounts of content. So, the value of
Link-Strength can raise the effects of Outdegree and
Density on trust threshold. Here, we suggest for the
strength of a tie to be defined as the average number
of two-way interactions between two actors that are
connected by that tie. Link-Strength,
,
, is
positively related to the number of incoming and
outgoing interactions between two nodes of and .
,
=
+
+
(1)
2.2 Solution of the Model
As the first step to incorporate the effect of noise in
information transmission into the model, we inspect
the model for the way inactive users would affect
measuring trust threshold. These actors barely send
any information out and are not as influential as
others in information transmission. However, the
network parameters that are assigned to them can
still fool the model and result in a higher level of
trust including inactive users in the calculations. An
extreme of such actors are those who have a high
Indegree together with a negligible Outdegree value.
Thereupon, to assign a zero value to the trust
threshold around all inactive users, the value of
is
multiplied by the hyperbolic tangent of their
Outdegree value. In such a way, in case the
outdegree is equal/close to zero, the trust threshold
will be equal/close to zero.
=
tanh
∗
+
∆
(2)
In addition, a major concern for measuring the
learning effects of network embeddedness is that
under the assumptions for the games with
incomplete information, the impact of the control
effects of embeddedness on the trust threshold is
lessened to a considerable degree (Buskens 2002,
chap 3). The reason is that in that situation, no
sufficient cues from the network are provided for the
trustee to control his behavior by a sanction
probability or a bad reputation aftermath. For that
reason, we propose to diminish the role of control
effects in the game-theoretic model and focus on the
impact of information diffusion. In this manner, the
parameters of the game will be defined as constant
in our model in order not to be influential on the
variation of the level of trust.
Moreover, for the sake of incomplete
information, the network structure is assumed to be
limited in the eyes of the trustee. This is
implemented by introducing networks with a large
number of nodes, the structure of which seems
extremely far from a trustee’s perception and is
unknown to him, except a few close relations.
3 SIMULATION
To achieve findings applicable to heterogeneous
networks, in which the assumptions for games with
incomplete information are applicable, we use a
simulation method. It is worthwhile to recall that, in
the model developed in this study, no specific trustee
is identified. In fact, the trustee is considered to be
the one with whom a trustor has interactions.
3.1 Sampled Networks
Earlier, in the development of the model, the size
and structure of the sample networks have been
contemplated in order to lead to a situation closer to
the assumptions for incomplete information.
Knowing that the results of Buskens’ (1998; 2002,
Chap. 3) game-theoretic model has shown no effects
of network size on the trust threshold, we feel free to
decide upon the number of nodes in the network.
Building the networks that would form the basis for
the simulation requires an exhaustive investigation
of many factors. Knowing that many methods for
creating sample networks carry considerable
drawbacks, we have decided to create networks for
simulation scenarios by sampling from an existing
online social network, Youtube. The only concern is
to fetch a number of network structures, representing
an actual social network, for simulation scenarios.
Networks are sampled starting from a randomly
selected user with an active profile and large number
of friends, using snowball sampling (Goodman
MEASURINGTRUSTINONLINESOCIALNETWORKS-TheEffectsofNetworkParametersontheLevelofTrustin
TrustGameswithIncompleteInformation
533
1961; Salganik and Heckathorn 2004). The
algorithm is provided with a random video ID
published by a user whose friends are added to the
network with the same structure and connections. To
do so, the Youtube network is crawled with a
snowball method to find them.
Sampling networks resulted in several networks
with thousands of nodes, among which 6 networks
with the number of users between 10,000 and 19,800
have been selected on which the simulation is to be
performed.
3.2 Experimental Design
Each of 6 abovementioned networks constitutes a
scenario for which the network parameters are
computed in the simulation. The values of
Outdegree, Indegree, Density, and Link-Strength are
calculated for every node in the network. These
values are further regressed on the values of
Indegree and Link-Strength to conclude the
influence of Indegree and Link-strength on trust
threshold in the context of noise. The (Spearman)
correlation coefficient between Indegree and Link-
Strength equals 0.093 in average for all 6 scenarios
which is low enough to make us confident to
perform their regression analysis separately.
However, the large number of cases in each scenario
is sufficient to distinguish the effects of the different
network parameters. For each node, two values of
trust threshold are calculated: model 1, and model 2.
Model 1 is the same as the solution introduced by
Buskens (1998; 2002) for the game-theoretic model.
The latter is the value of trust threshold after
eliminating inactive users from the solution due to
the insignificance of their role in information
diffusion. The two models will be compared to make
deductions regarding additional networks parameters
in this study.
Both network and non-network parameters have
to be sampled for each simulation scenario. Network
parameters are calculated for each node in every
sampled network. Non-network parameters, on the
other hand, follow the same variation that is used by
Buskens (1998; 2002) and sampled independently
(in the probabilistic sense) for each network.
Noteworthy here is that, in each scenario, the value
of the game parameters that are involved in the
calculations of the trust threshold is set to be the
same for all trustors in a network. The reason is to
prevent its variation from being considered to be
effective in the calculations. This is perfectly in line
with the fact that introducing the assumptions of
incomplete information into the contexts of IHTG
would reduce the control effects of network
embeddedness (Buskens 1998; 2002) that are
implemented by the game parameters in this model.
The simulated system is a social network,
demonstrated by its graph with finite number of
nodes, for each the dependent variables are
computed to generate the simulation data. The
simulation environment is developed using Java and
Java Universal Network/Graph Framework (JUNG)
(2009). A “terminating simulation” (Banks, John S.
Carson et al. 1996, chap 12) is performed for each
scenario, as the termination circumstances for each
run is embedded in the simulation scenario
description. Each simulation scenario starts
traversing the network graph from a node to
compute the required values for network parameters
and trust threshold for both models, and terminates
when the computation is done for the last node in the
network. The system is studied for a single point of
time at which the network is sampled from Youtube,
hence assumed to be in a constant state during the
simulation. The outcome of the simulation is a set of
random values that constitute the “simulated data”
for further analysis. Here, the output consists of two
network parameters, Indegree and Link-Strength, in
addition to the two dependent variables of trust
threshold for both models 1 and 2,
and
respectively.
3.3 Analysis of the Simulated Data
The values for the dependent variables in each data
set do not fall below zero and even though they do
not always take a known value, they are known to be
elements in an interval. Thus, a regression analysis
of the dependent variables can be performed.
However, we cannot perform a linear regression
since the values of variables do not follow a normal
distribution. Therefore, to determine the correlation
between two variables, a Spearman regression
analysis is applicable (Sheskin 2004, p. 1360-1362).
In addition, to make sure if the output values from
the simulation are valid to be further analyzed, we
perform a confidence level t-test on the dependent
variables. The results show that after the first
simulation run the error in the average of the trust
threshold would not be more than 5% with the 95%
confidence, and with repeating simulation for 4
times we can be confident that with the probability
of 98%, the error would not exceed 2 percent.
Table 1 shows the results of the Spearman
regression analysis of the effects of Indegree and
Link-Strength on the trust thresholds for both
models. Spearman R-squared value for the
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
534
regression is given for each model in every scenario.
The second raw in the tables represents the
hypotheses on the effects as are derived from the
analytic results. The results obtained from the
regression, either strong or weak association
between the network parameters and trust
thresholds, are significant for all simulation
scenarios ( < 0.0001).
Table 1: rho of the Spearman regression of Indegree and
Link-Strength with trust threshold in both models for all
six scenarios (<0.0001).
Indegree Link-Strength
Model 1 Model 2 Model 1 Model 2
Hypothesis + ? + ?
Scenario 1 0.47 0.04 0.40 0.25
Scenario 2 0.58 0.08 0.40 0.23
Scenario 3 0.52 0.05 0.39 0.23
Scenario 4 0.56 0.04 0.40 0.22
Scenario 5 0.50 0.05 0.41 0.26
Scenario 6 0.52 0.04 0.37 0.24
The results of the effects of indegree on trust
threshold in model 1 show a moderate positive
correlation between two variables, meaning that the
value of the dependent variable, trust threshold,
increases in the Indegree of the trustors. This also
denotes that introducing noise in information
diffusion would result in a context in which
conclusions about the effects of Indegree on trust
threshold can be driven from the game-theoretic
model. Moreover, the results of Model 1 for
Indegree support the previous finding of Buskens
(1998) showing that ‘for given outdegrees, the
network structure centralized with respect to
indegree around the buyers with the highest
outdegree is the structure for which all buyers have
the highest trust threshold’ (p. 277). Considering that
Indegree is not an element of the computations for
trust threshold in the models, as it is calculated
based on the Outdegree and Density, however, it
cannot be claimed that the values for Indegree and
Outdegree are independent from each other, since
for a network as a whole, the aggregated number of
indegree equals to that of outdegree.
In the second model, users with high values for
Indegree and Density do not necessarily contribute
to the trust threshold, unless having an Outdegree
value equal to at least 1. The considerable
correlation between the trust threshold and Indegree
drops dramatically to about zero after eliminating
the effects of inactive users from the computations
for trust threshold. The same fall is conspicuous in
the average trust threshold in each network, since its
value is set to be zero for inactive users. This
represents a roughly extreme case in which inactive
users can fool the model to return a fair value for
trust threshold as of their ample Indegree value,
whereas the actual trust threshold is less for those
users because of their latent contribution to
information diffusion.
According to the findings on the effects of Link-
Strength on trust threshold, it is reasonable to infer
that Link-Strength is probably an important factor
that should be considered in the analysis of social
networks in case of studying the learning effects of
network embeddedness. The results of the regression
in model 1 show a fair positive correlation between
the two variables so that about 39 percent of growth
in the trust threshold can be explained by the value
of Link-Strength. Even after removing the effects of
users with high Indegree value paired with a zero
Outdegree, this association falls to 25%, in average,
which is not weak enough to be completely
neglected. Such inference is intuitively justifiable. In
a noisy environment, if the relationship between two
trustors is strong, one can be influenced by even the
lowest amount of information received from the
other. Ergo, the information flowing between
trustors who have some close friends, i.e.
distinguished by strong connection links, are more
likely to be influential than between those who have
many friends but almost no close ones. This is
indeed the case under the circumstance that both
groups transfer comparable amounts of information.
The rate at which the Spearman’s rho falls after
inactive users are dismissed from the model 1 is
shown to be significantly higher for Indegree than it
is for Link-Strength. Intuitively speaking, an
explanation of such can be that inactive users do not
make strong friendship relations, thus the
elimination of those would not ensue omission of a
considerable number of strong links. Therefore, the
value of network parameters will not experience a
prodigious change in regards to the Link-Strength,
so its effects on the trust threshold will still remain
roughly the same. However, a reduction of those
effects is reasonably predictable. Furthermore, such
variation in the drop rates of rho values of Indegree
and Link-Strength can be interpreted so that the
conclusions for the effects of Link-Strength on the
trust thresholds are more reliable. Of course, we
make such statement under the circumstances that
the values of Indegree and Link-Strength are
considered to be calculated with independent
network elements.
4 SUBSTANTIVE IMPLICATIONS
The assumptions of the model in this study results in
MEASURINGTRUSTINONLINESOCIALNETWORKS-TheEffectsofNetworkParametersontheLevelofTrustin
TrustGameswithIncompleteInformation
535
diminishing the control effects of network
embeddedness, while altering the focus of the game-
theoretic model to the learning effects, influenced by
the role of information diffusion between trustors.
The model applies the previous findings, and the
results extend theoretical hypotheses for trust in trust
relations, and are in accordance with the existing
literature (Raub and Weesie 1990; Coleman 1994;
Weesie, Buskens et al. 1998).
The following hypotheses express the outcomes
of the model:
Hypothesis 1. In a context with noisy information,
trust increases with the value of Indegree of the
trustors.
Hypothesis 2. In a context with noisy information,
trust increases with the values of Link-Strength of
the trustors.
Hypothesis 3. In a context with noisy information,
the positive effects of Link-Strength on trust are
more promising and unyielding than those of
Indegree.
Hypothesis 4. In a context with noisy information,
the high Indegree value of users who do not
supplement information diffusion in a network do
not lead to an increase in the trust that can be placed.
Figure 1 illustrates the driven hypotheses in the
context of this study, while the previous hypotheses
for network parameters still remain valid. It also
shows the position of our model, and its
assumptions, related to Buskens’ (1998; 2002)
game-theoretic model.
Figure 1: The assumptions and outcomes of the model for
games with incomplete information.
5 CONCLUSIONS
The results of this study extend the findings of cases
with complete information, while those findings
remain valid in the newly developed context. The
hypotheses for two network parameters, Indegree
and Link-Strength, could not be driven in a network
of trustors where information is assumed to be
accurately transferred between Trust Game players.
Adding the possibility of existing corrupt pieces to
the flow of information in a network creates an
environment that is closer to reality in which
learning about the behavior of the trustee is more
complex and affected by more parameters.
Particularly, games with incomplete information are
to be utilized for analyzing the learning and control
effects of embeddedness in an integrative manner.
However, in respect of theoretical modeling,
relaxing game-theoretic strong rationality
assumptions and introducing more realistic ones
about how actors use relevant information that they
obtain, seldom can come to a balance with analytic
tractability. In addition to the intricacy of models
with more realistic assumptions, knowledge about
what realistic assumptions could be is limited
because the effects of learning and control
mechanisms has not yet been successfully cleared up
by empirical researches (Buskens and Raub 2008).
Corresponding to the situation, we have tried to
relax less disturbing assumptions and introduce a
few ones regarding to the games with incomplete
information to be able to extract results for the
learning effects of network embeddedness on the
level of trust.
We have shown that trustors with higher
Indegree have the capability to certify the positive
information about a trustee while they receive wide
variety pieces of information about others’
experiences with the trustee. The assumptions of this
model approximate those of the contagion models
for information diffusion in heterogeneous networks
(Buskens and Yamaguchi 1999) rather than the
assumptions in transit models for such (Friedkin
1992; Yamaguchi 1996). Contagion models measure
the extent to which an individual is influenced by
information that is flowing in a network, whereas in
transit models it is sufficient for an actor to obtain
the information to be affected by it. The unrealistic
assumptions of the transit models let them
overestimate the effects of a number of network
parameters. Here, we argue that not all the values for
network parameters are conclusive and cannot be
considered as effective on the trust level. We suggest
that the Outdegree value should be weighted by the
strength of the links through which information is
transmitted between two actors, so that it would be
possible to conclude the extent to which the
information in flow is actually influential.
To link these results to the discussions on the
control and learning effects of embeddedness, it can
be concluded that a piece of negative information
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
536
about the trustee per se, and the opportunity for the
trustor to exit the iterated trust games, does not
promise additional control opportunities for the
trustor in this model. Besides, adding the
opportunity of spreading “voice” in a network
creates more learning as well as control possibilities
for the trustors. This is in agreement with the
assertion by Buskens (2003) that ‘… the two aspects
of voice [control and learning] need to be combined
by the trustors to enable more trust in the trustee’ (p.
246). Such a result can be in favor of the situations
in which trustors can sometimes experience negative
outcomes while the trustee has not intentionally
abused trust (for examples, see Radner 1981; Porter
1983). Learning is expected to be more important in
such situations where a piece of negative
information about trustee’s behavior, solely, should
not fundamentally ensue in an exit option. Another
such situation is one where the trustee does not have
a fixed type, thus the trustors would hesitate to exit
and more like to observe the changes in the trustee’s
type. Therefore, every experience with the trustee
would be worthwhile to the trustors (Mailath and
Samuelson 2001).
This model also carries some restrictions. Even
though we have minimized the effects of the game
parameters in the model to be able to aim the
attention at the effects of information flow between
trustors, the trustors are still considered as playing
successively with the trustee. Therefore, we cannot
claim that the model can be applied to the situations
in which trustors can play simultaneously with the
trustee. The simulation study context that we have
implemented in this study can be considered as one
in which the order of the Trust Games is not a matter
of importance (see Buskens 2003), however, no
assumption is made in this regard. Another
important disadvantage of this model is that trust is
not investigated joined together with distrust. In fact,
it is reasonable to deduce that the existence of an
amount of negative information on the trustee’s
behavior could take a few more steps than just
reducing the trust level and ensue distrust.
Measuring distrust, involves different factors while
the relative assumptions are still ambiguous since
the topic has not yet attracted enough attention of
scholars. Still, extended assumptions, as mentioned
above, would modify the model to one that is closer
to real situations.
Certainly, the discussions on how such
assumptions can be changed or extended encompass
a wide range of considerations. However, in this
respect, Buskens (2003) states that ‘… I think that
we lack considerable knowledge about what actually
reasonable assumptions are especially related to
information availability of actors, information
exchange, among actors, and how actors actually use
this information [to update their beliefs, or decide
upon sanctioning the trustee] …’ (p. 247).
Therefore, it would be fruitful to develop such
experimental designs that allow for testing both the
implications of theoretical models and the way
actors use the information obtained while playing a
game. For studying the learning effects, it is more
favorable to analyze the decision making process of
actors rather than the decision itself (ibid). Such
contemplative propositions would extremely add to
the complexity of the current models of Trust Games
and can form cases for further research.
ACKNOWLEDGEMENTS
Clarifying conversations with PhD. Saeed Dastgiri,
professor at Tabriz University of Medical Sciences,
Iran, and PhD. Ali Ardalan are gratefully
acknowledged.
REFERENCES
Aberer, K. and Z. Despotovic (2001). Managing trust in a
peer-2-peer information system, ACM.
Artz, D. and Y. Gil (2007). "A survey of trust in computer
science and the semantic web." Web Semantics:
Science, Services and Agents on the World Wide Web
5(2): 58-71.
Banks, J., I. John S. Carson and B. L. Nelson (1996).
Discrete-Event System Simulation, Prentice-Hall
International, Inc. .
Beth, T., M. Borcherding and B. Klein (1994). "Valuation
of trust in open networks." Computer Security—
ESORICS 94: 1-18.
Bonatti, P., C. Duma, D. Olmedilla and N. Shahmehri
(2005). An integration of reputation-based and policy-
based trust management. The Semantic Web Policy
Workshop.
Bonatti, P. and D. Olmedilla (2005). Driving and
monitoring provisional trust negotiation with
metapolicies, IEEE.
Bos, N., J. Olson, D. Gergle, G. Olson and Z. Wright
(2002). Effects of four computer-mediated
communications channels on trust development,
ACM.
Brainov, S. and T. Sandholm (1999). Contracting with
uncertain level of trust, ACM.
Bratley, P., B. L. Fox and L. E. Scharge (1987). A Guide
to Simulation. New York, Springer-Verlag.
Burt, R. S. (1987). "Social contagion and innovation:
Cohesion versus structural equivalence." American
Journal of Sociology: 1287-1335.
Burt, R. S. and M. Knez (1996). "Trust and third-party
gossip." Trust in organizations: Frontiers of theory and
MEASURINGTRUSTINONLINESOCIALNETWORKS-TheEffectsofNetworkParametersontheLevelofTrustin
TrustGameswithIncompleteInformation
537
research 68: 89.
Buskens, V. (1995). "Social networks and the effect of
reputation on cooperation." ISCORE paper 42.
Buskens, V. (1998). "Network Construction Methods for
the Simulation of Stochastic Blockmodels DRAFT."
Buskens, V. (1998). "The social structure of trust." Social
Networks 20(3): 265-289.
Buskens, V. (2003). "Trust in triads: effects of exit,
control, and learning." Games and Economic Behavior
42(2): 235-252.
Buskens, V. and W. Raub (2008). "Rational choice
research on social dilemmas: embeddedness effects on
trust." Handbook of Rational Choice Social Research.
New York: Russell Sage.
Buskens, V. and A. Van de Rijt (2008). "Dynamics of
networks if everyone strives for structural holes." ajs
114(2): 371-407.
Buskens, V. and K. Yamaguchi (1999). "A new model for
information diffusion in heterogeneous social
networks." Sociological Methodology 29(1): 281-325.
Buskens, V. W. (2002). Social networks and trust, Kluwer
Academic Pub.
Camerer, C. and K. Weigelt (1988). "Experimental tests of
a sequential equilibrium reputation model." Econo-
metrica: Journal of the Econometric Society: 1-36.
Coleman, J. S. (1964). "Collective Decisions*."
Sociological Inquiry 34(2): 166-181.
Coleman, J. S. (1994). Foundations of social theory,
Belknap Press.
Coleman, J. S., E. Katz, H. Menzel and C. U. B. o. A. S.
Research (1966). Medical innovation: A diffusion
study, Bobbs-Merrill Co.
Dasgupta, P. (2000). "Trust as a Commodity." Trust:
Making and Breaking Cooperative Relations,
electronic edition, Department of Sociology,
University of Oxford: 49-72.
Falcone, R. and C. Castelfranchi (2004). Trust dynamics:
How trust is influenced by direct experiences and by
trust itself, IEEE Computer Society.
Friedkin, N. E. (1992). "An expected value model of
social power: Predictions for selected exchange
networks." Social Networks 14(3-4): 213-229.
Friedman, B., P. H. Khan Jr and D. C. Howe (2000).
"Trust online." Communications of the ACM 43(12):
34-40.
Fudenberg, D. and J. Tirole (1991). Game theory. 1991,
MIT Press.
Gibbons, R. (2001). "Trust in social structures: Hobbes
and Coase meet repeated games." Trust in society:
332-353.
Golbeck, J. and J. Hendler (2004a). "Accuracy of metrics
for inferring trust and reputation in semantic web-
based social networks." Engineering Knowledge in the
Age of the SemanticWeb: 116-131.
Goodman, L. A. (1961). "Snowball sampling." The Annals
of Mathematical Statistics: 148-170.
Granovetter, M. S. (1973). "The Strength of Weak Ties."
The American Journal of Sociology 78(6): 1360-1380.
Harsanyi, J. C. (1995). "A new theory of equilibrium
selection for games with complete information."
Games and Economic Behavior 8(1): 91-122.
Harsanyi, J. C. and R. Selten (1988). "A general theory of
equilibrium selection in games." MIT Press Books 1.
Haythornthwaite, C. (1996). "Social network analysis: An
approach and technique for the study of information
exchange." Library & Information Science Research
18(4): 323-342.
Janssen, M. A. (2011). Small World Networks. Games and
Gossip, OpenABM Consortium.
Jarvenpaa, S. L. and D. E. Leidner (1998).
"Communication and trust in global virtual teams."
Journal of Computer Mediated Communication 3(4):
0-0.
Kreps, D. M. (1992). "Game theory and economic
modelling." OUP Catalogue.
Kreps, D. M. (1996). "Corporate culture and economic
theory." Firms, organizations and contracts: a reader in
industrial organization: 221–275.
Lia, N., W. Winsborough and J. Mitchell (2003).
"Distributed credential chain discovery in trust
management." Journal of Computer Security 11(1):
35–86.
Lipnack, J. and J. Stamps (1997). Virtual teams: Reaching
across space, time, and organizations with technology,
John Wiley & Sons Inc.
Mailath, G. J. and L. Samuelson (2001). "Who wants a
good reputation?" Review of Economic Studies 68(2):
415-441.
McGuire, W. J. (1966). "Attitudes and opinions." Annual
review of psychology 17(1): 475-514.
Nash, J. (1951). "Non-cooperative games." The Annals of
Mathematics 54(2): 286-295.
Naylor, T. H., J. Finger, J. L. McKenney, W. E. Schrank
and C. C. Holt (1967). "Verification of computer
simulation models." Management Science: 92-106.
Porter, R. H. (1983). "Optimal cartel trigger price
strategies* 1." Journal of Economic Theory 29(2):
313-338.
Radner, R. (1981). "Monitoring cooperative agreements in
a repeated principal-agent relationship."
Econometrica: Journal of the Econometric Society:
1127-1148.
Raub, W. and J. Weesie (1990). "Reputation and
Efficiency in Social Interactions: An Example of
Network Effects." American Journal of Sociology
96(3): 626-654.
Riegelsberger, J., M. A. Sasse and J. D. McCarthy (2003).
"The researcher's dilemma: evaluating trust in
computer-mediated communication." International
Journal of Human-Computer Studies 58(6): 759-781.
Salganik, M. J. and D. D. Heckathorn (2004). "Sampling
and Estimation in Hidden Populations Using
Respondent Driven Sampling." Sociological
Methodology 34(1): 193-240.
Sheskin, D. (2004). Handbook of parametric and
nonparametric statistical procedures, CRC Pr I Llc.
Snijders, C. (1996). Trust and commitments, Purdue
University Press.
Weesie, J., V. Buskens and W. Raub (1998). The
management of trust relations via institutional and
WEBIST2012-8thInternationalConferenceonWebInformationSystemsandTechnologies
538
structural embeddedness. The Problem of Solidarity:
Theories and Models. P. Doreian and T. J. Fararo.
Amsterdam: 113-138.
Yamaguchi, K. (1996). "Power in networks of
substitutable and complementary exchange relations:
A rational-choice model and an analysis of power
centralization." American Sociological Review: 308-
332.
MEASURINGTRUSTINONLINESOCIALNETWORKS-TheEffectsofNetworkParametersontheLevelofTrustin
TrustGameswithIncompleteInformation
539
