ArticlePDF Available

The choice of complete anonymity versus pseudonymity for aggression online

Authors:

Abstract and Figures

Anonymity is a factor that could lead to disinhibited behavior which is something that could cause damage to many online communities. Anonymity is a generic term and should be analyzed further into different states such as pseudonymity and complete anonymity. In this paper a survey was conducted in order to determine the differences between the two anonymity states in relation to aggression. The findings show that in general, there is no difference in how aggressive a user responds between the two or when users answer with their real names. However, when the differences were tested in context of how strong of an opinion users felt they had about a topic, users that selected ”extremely strong” as their opinion, were found to respond more aggressively under the state of pseudonymity. Based on the evidence of this exploratory study, user-centered design could improve online community behavior by changing the design process, specifically to the approach of anonymity.
Content may be subject to copyright.
This is a preprint of an article published in eMinds: International Journal on Human-Computer Interaction, 2(8),
35-57, copyright c
2012 (University of Oviedo). This preprint has been updated to reflect changes in the final version.
The choice of complete anonymity versus
pseudonymity for aggression online
Michail Tsikerdekis
Masaryk University
Faculty of Informatics
Brno, Czech Republic
tsikerdekis AT gmail.com
Abstract Anonymity is a factor that could lead to disinhibited
behavior which is something that could cause damage to many
online communities. Anonymity is a generic term and should be
analyzed further into different states such as pseudonymity and
complete anonymity. In this paper a survey was conducted in or-
der to determine the differences between the two anonymity states
in relation to aggression. The findings show that in general, there
is no difference in how aggressive a user responds between the
two or when users answer with their real names. However, when
the differences were tested in context of how strong of an opinion
users felt they had about a topic, users that selected ”extremely
strong” as their opinion, were found to respond more aggressively
under the state of pseudonymity. Based on the evidence of this
exploratory study, user-centered design could improve online com-
munity behavior by changing the design process, specifically to
the approach of anonymity.
1 Introduction
Anonymity is a well known contributor to aggression online because of
the way that it helps individuals to act in a disinhibited way [14] [38] [31]
[32] [40] [16] [26] [29] [35]. Solutions about reducing this negative effect of
anonymity in social networking media have yet to be found. Furthermore,
there seems to be a void in literature and our understanding of the effects
of pseudonymity or complete anonymity in relation to aggression and how
these two anonymity states differ from each other.
Aggression can affect every aspect of online interaction and social net-
working media are no exception. The potential in economic damages but
also human life is immense as can be seen by examples of cyberbullying
and other ways of online aggression. One case was the suicide of a female
teenager that fell victim to a mother pretending to be a 16-year-old boy
[3]. So it becomes obvious that the danger is real for the victims but it
is not hard to imagine why a company’s reputation could be at stake if
aggressive incidents persist within their community.
Michail Tsikerdekis
Yet, online communities enable people experiencing a conflict to en-
gage in dialogue with people outside their borders, discuss their situation,
and reach peaceful resolutions [2]. Communities such as online support
groups have tremendous potential in handling specific conditions of dis-
tress [4]. Dialogue and communication is at the center of these online
communities for achieving their goals. Aggressive acts online may dis-
rupt these dialogues and cause unrepairable damage to these communities.
Hence, it may be beneficial to understand the causes of online aggression
and find ways of reducing it.
One way of regulating aggression in online communities is increas-
ing the policing force that consists of moderators, administrators and all
those alike within the community. This has the potential to resolve any
problems when and if they occur but it is not a preventing measure. Put
simply, it can heal the symptoms but it does not eradicate the cause.
Another downside about moderating in order to reduce violence online,
is that it requires labor which in turn produces additional cost for the
companies. In order to make an aggressive occurrence preventable, a shift
in the way that software engineers are trained to think and work today is
needed. Software engineers need to two ask questions about their software
and their design interfaces. How does the software affect the interaction
between the users and what can we do to change it and make it more
efficient towards the software engineer’s initial goal?
There are several examples in which when these questions have been
addressed in similar studies, significant results were observed for the pro-
cedures in question. We now know that, emotional states can be de-
termined by body movements through unobtrusive mechanisms which if
implemented could improve collaborative environments by building more
meaningful interactions among their members [13]. Another example is
the importance of thoroughly investigating cognitive and affective factors,
when designing interactive media that may be critical for the future de-
velopment of virtual environment applications [37]. Arguably, and in a
similar way, user behavior can be predicted with the case of anonymity
and aggression and interaction can be engineered towards reducing ag-
gression online.
This paper examines the viability for software engineers to decide
whether or not to provide users with the option of complete anonymity
or pseudonymity and expect a reduction in the aggressive exchange of
messages within a network.
2 The concepts
2.1 Online aggression
Aggression and its subcategory of online aggression, is the delivery of
an aversive stimulus from one person to another, with intent to harm
and with an expectation of causing such harm when the other person is
2
The choice of complete anonymity versus pseudonymity for
aggression online
motivated to escape or avoid the stimulus [12]. Among human beings this
can take a variety of forms of violence such as mental, verbal or physical.
Availability, abundance and the free nature of online services provide
an individual with the perfect set of tools to hurt someone. Sometimes
the damage maybe even too hard for even moderators to contain. Online
studies on aggressive games have been conducted in order to determine
if in fact there is any correlation between games and aggressive behavior
but results have yet to produce a definitive answer. One study showed
that neither the disposition of the opponent nor the aggressive level of the
game type affected participant aggression [28]. Still, with so much variety
of online gaming and so many options about anonymity for each game,
results can be hard to generalize for all circumstances.
This variety of circumstances points to the fact that aggressive behav-
ior can be depended in a number of situational factors and experiences.
Examples of these factors can be the presence of violent objects such as a
gun [7], or experiences within a military setting which provide the social
context where servicemen learn aggression, violence, and murder [8]. But,
the most prevalent theory that clearly makes the connection between ag-
gression being affected by situational factors is the frustration-aggression
theory. The theory states that frustration can lead to anger, and that
anger triggers a hostile action [5].
In cases of online ’inappropriate’ behavior the presence of voice com-
munication has proven to significantly affect the way people act as it
was discovered in a research study regarding the well known prisoner’s
dilemma game [10]. If these situational factors that lead to frustration
and in turn aggression were to be understood, software engineers could
in theory develop the ’perfect’ environment in which it could provide the
user with the least amount of friction and best user experience for actually
preventing aggression up to a certain level.
2.2 Online Anonymity
Anonymity refers to the state of an individual’s personal identity, or per-
sonally identifiable information, being publicly unknown. There are a
couple of different issues that arise with the case of anonymity such as
the informative effect, group pressure effect and enforcement effect [21].
Some of the effects are positive for communities while others are not. As
an example, the presence of cues to identity positively affects interpersonal
perceptions, but at the same time decreases perceptions of solidarity or
entitativity [33]. Hence, the decision whether to provide users with the
option of anonymity has advantages and disadvantages. In another exam-
ple, research has shown that in votes and debates, people voting anony-
mously are more likely to change their vote and less likely to conform
with the group’s norm, behavior consistent with preventing groupthink
type behavior which could lead to ineffective and even risky decisions [1].
Providing the option of anonymity for a social networking brainstorming
3
Michail Tsikerdekis
group is the right choice according to the above study. Another study’s re-
sult that seems to agree with the above has shown that individuals under
anonymity states have reduced concerns about being positively evaluated
by others, and this creates an impersonal, task-oriented focus for group
interaction [20].
On the other hand when using anonymity the negative effects might
be too overwhelming for the stability of a community. The main effect
of anonymity as a general term is that it is often associated with aggres-
sion. Even though an individual’s identity might be traceable, simply the
heightened feeling of anonymity appears to be enough to promote disinhib-
ited behavior [38]. There is also a widespread agreement that anonymity
removes accountability out of the equation of online communication and
therefore reducing the core values of democratic tradition [9] [17]. Other
studies on group decision support systems have shown other issues that
could arise with the presence of anonymity such as the loss of credibility
and influence [27] and loss of accountability [11]. In addition anonymity
in GDSS could be addressed as something multidimensional and could be
subjective and context-dependent [24].
To be able to balance these positive and negative effects, an under-
standing of anonymity is essential. From a technical standpoint anonymity
can be divided into anonymity, unlinkability, linkability, undetectability,
unobservability, pseudonymity and so on [23]. Arguably these states of
anonymity are not perceived by the casual user and so for the purpose
of this study three states were examined closely. The anonymity states
were, pseudonymity where a nickname is used instead of a name, complete
anonymity where a user is denoted simply as anonymous and the state
where anonymity was absent and a user uses his or her real name. The
later was the control in this study.
2.2.1 Factors for disinhibited behaviour and complete anonymity
A number of factors contribute to the way people act while under the in-
fluence of anonymity. The reduced social cues model was one of the first
models that were developed to describe the nature of computer-mediated
communication. It argues that the reduced social contexual information
could have certain effects for groups such as effects of disinhibition and
liberation [19] [18]. A relevant theory coined the online disinhibition ef-
fect described the way people feel less restrained while online as well as a
number of factors such as dissociative anonymity, invisibility, asynchronic-
ity, solipsistic introjection, dissociative imagination, and minimization of
authority [31] [32]. The factors that are interesting for this paper and
are more associated with anonymity are dissociative anonymity but also
dissociative imagination.
Dissociative anonymity could be best explained as the sense of protec-
tion that one has under an ostensibly anonymous blanket while dissocia-
tive imagination describes the feeling of escapism, to throw away mundane
4
The choice of complete anonymity versus pseudonymity for
aggression online
concerns without having to worry about the consequences. The two may
sound similar but the later adds to the individual’s perception a sense that
the online space exists in a different realm beyond reality where different
rules may apply, even for one’s online artificially created persona which
is one’s pseudonym. Essentially the nickname takes on a life of its own.
Studies in real life conditions such as the Stanford prison experiment have
demonstrated the power of one believing that they are actually someone
else, essentially internalizing a role [42].
For the state of complete anonymity where an individual is completely
anonymous one can claim that dissociative anonymity as a factor plays
a big role. In fact dissociative anonymity is so strong that an individual
might even convince him or herself that he or she has no responsibility of
the online actions [31].
2.2.2 Pseudonyms & Disconnecting with the Online Self
On the other hand, when it comes to the use of pseudonyms there seem to
be more than two factors at play. Aside from the dissociative anonymity,
there is also the second factor of dissociative imagination which describes
the idea of one depicting that their online alter ego exists beyond the realm
of reality and therefore it could act in a different manner than one’s real
self. It is easy to understand how dissociative imagination is associated
more with the use of characters with nicknames/pseudonyms and not with
complete anonymity since there needs to be a character creation process
in order to contribute as a factor.
According to the above remarks an individual could grow an attach-
ment to his or her pseudonym self and become more aggressive just be-
cause the same rules that apply in the real world do not apply in the online
world. On the other hand a user with complete anonymity is lacking the
factor of dissociative imagination and therefore his or her actions may be
less aggressive.
2.2.3 Pseudonyms & Connecting with the Online Self
On the opposite side of the seesaw research on IRC nicknames and im-
pression formation seems to suggest that nicknames are an inherent part
of their Net-identity, and even of their real-life identity [6]. People find the
need to describe their traits, characteristics and appearances with their
nicknames and try to find the optimum way to do so. In a way they
try to encapsulate part of their personality into a nickname creating an
online extension of their real selves and not a completely different and
independent alter ego.
In fact the research that was mentioned above shows that people grow
a long term attachment with their nicknames [6]. Usually they prefer to
keep the same nickname and identity which, for the most part is connected
with the part of real self which they wish to share with others. In addition,
5
Michail Tsikerdekis
the same study concluded that most people when selecting a nickname do
not base their decisions in collective values but rather values related to
one’s self.
Similar to the above, another study seems to agree with the idea that
users adapt their avatars to reflect their own appearance and users who
perceive their avatars to be similar to their own appearance experience
as a result heightened private self-awareness [36]. Essentially the study
suggests that avatars which increase their owners’ self-focus may have an
influence on online behavior in the context of social computing.
2.2.4 Disconnecting versus Connecting with the Online Self
There are two conclusions that can be extracted from the above re-
marks about online anonymity and aggression. The first, is that complete
anonymity is an anonymity state where dissociative anonymity as a factor
could heighten disinhibited behaviour and in turn could lead to aggres-
sion. The second conclusion though, is contradicting. When it comes
to pseudonymity by using nicknames/pseudonyms, there are two powers
at play. The first is that people try to include part of their personality
into the online persona that they are creating while the second, dissocia-
tive imagination, tries to drive the creating process beyond the realm of
reality. The question is, which one of these two powers wins over and
therefore how pseudonymity differs from complete anonymity in relation
to aggression.
From a design perspective the benefits of knowing these differences for
software engineers can be extremely important because by simple changes
in the design, i.e. shifting from completely anonymous users to users with
pseudonyms or vice versa, can have a significant effect on the level of
aggression that a community produces through its users. In addition, if
by these alterations aggression can be reduced, moderation costs for social
networking media could also decrease [35].
3 Method
In order to be able to pinpoint how different anonymity states affect ag-
gression as factors a survey was created. Surveys consist of systematic and
standardized approaches for collecting information [22]. This standard-
ization is ideal for collecting similar data from groups that can be then
interpreted comparatively, which in turn reduces the researcher’s subjec-
tivity and produces highly reliable results. This survey aimed to explain
if complete anonymity and pseudonymity have an effect on how users re-
act and communicate through an online social network, specifically about
how much more aggressively they might behave. In addition the survey
would also explore any potential differences in the effects that complete
anonymity and pseudonymity may have on a user’s interactions.
6
The choice of complete anonymity versus pseudonymity for
aggression online
The population of interest was users with social networks and there-
fore the survey was conducted through Facebook as an application in the
form of a survey. Therefore everyone with a Facebook account that had
sufficient knowledge, about how to use a Facebook application could take
part in the survey. The sample was obtained with convenient/snowball
sampling methods. This was done in order to reduce the negative effects of
both sampling methods. The survey was advertised in several academic
and non academic groups on Facebook informing participants that the
survey was about interactions and interface design. The second sampling
method provided users with the ability to invite their friends into the
survey. This snowball sampling method is preferable for social networks
especially when individuals are sensitive about their privacy and show
unwillingness to participate in surveys without a friend of theirs opting-in
first. According to the theory of six degrees of separation, each individ-
ual had a statistical probability to be chosen using the snowball sampling
method. Although the topic still debatable, repeated studies have shown
that the distance between one person and another in a social network is
approximate to six [39] [41]. The application was originally written in En-
glish but later on translated to Greek as well. The translation was exact
and the sentences retained the original meaning that they had in the En-
glish language. Depending on the facebook language settings or operating
system settings, the language was selected automatically for the users.
Figure 1: The first page of the survey
The first page of the survey informed the participants that this was
a scientific research survey, that there are no right or wrong answers, to
answer honestly, answer as they would answer in a real life situation, and
to complete all the questions in the survey. In order to ensure the final,
all the survey was programmed with javascript code that would prompt
a user with an error message in case there was a question that was not
answered in the survey. Software tracked the progress of each individual
7
Michail Tsikerdekis
through the survey recording if it was completed successfully or not.
Figure 2: The second page of the survey
In the second page, people were asked to answer a series of questions
about highly controversial topics, as well as basic demographics such as
their gender and age group. The idea behind the use of highly controver-
sial topics was that people will have a bigger incentive to act aggressively
in the hypothetical scenarios in the later stages of the survey. In addition
the topics that were picked had more or less two poles in which people
could decide which side they were on. The topics were, death penalty,
abortion, and animal rights. As an example, participants were prompted
with the question “Do you believe in the death penalty?” with possible
answers “yes” or “no”. In addition underneath that question there would
be another which asked “How strong is your opinion about the death
penalty?” with possible answers “somewhat strong”, “very strong”, and
“extremely strong”. The reason behind this type of question was that it
was suspected that individuals that felt they had a strong opinion about
the topic of each scenario may be affected more by the different anonymity
states. Finally, after the individual finished answering the same questions
about the other two topics they were asked to type in a nickname for them-
selves. The nickname could be anything that they wanted aside from any
clear association with their name or surname.
The next stage of the survey was a series of three scenario based ques-
tions, based on the three different topics. Each scenario was prompted
separately on a single page. The system was evaluating the answers that
8
The choice of complete anonymity versus pseudonymity for
aggression online
Figure 3: Possible scenario variations, answers remain the same for all
cases
the users gave in the second page of the survey and created scenarios ac-
cording to the user’s alignment in a specific topic. For example if someone
was pro-choice he or she would have received a scenario about a childhood
friend that was raped, resulting in an unwanted pregnancy and that the
parents of the pregnant female wanted her to keep the baby due to reli-
gious beliefs. In the case of somebody being pro-life the scenario would
be the opposite (see figure 5 in Appendix). The user’s task was to write a
message to the parents in response to their decision, trying to explain his
or her feelings about the parents wishes. It was suspected that by mak-
ing a situation personal for the user he or she would be more inclined to
act aggressively. The participants had a list of already prepared answers
which they could choose from. A note should be made here that regardless
of the alignment of a user for a specific topic, pro-life or pro-choice, the
answers were exactly the same in order to avoid any bias. The answers
were four. They had a form of a personal message and moved progres-
sively from the first being extremely polite to the last being extremely
rude and aggressive. The answers in between the two extremes are lesser
versions of the extremes and stand for polite, rude or barely aggressive.
The reason for the four point Likert scale was to force users to decide
if they would respond politely or rudely. The choice was reinforced by
research that was conducted on flame wars. A flame war is when one or
more users engage in provocative responses overshadowing regular forum
9
Michail Tsikerdekis
discussion. Online flaming is most clearly associated with the expression
of antagonism [34]. Based on this if a user would decide to answer with
the third or the fourth choice it would potentially start a flame war or
even trigger an exchange of unpleasant messages.
The scenarios were created to cover the anonymity states in question,
including our control state which is the absence of anonymity. In ef-
fect, variations of the scenarios were created for someone answering with
their real name, for someone answering with their nickname and finally
for someone answering completely anonymously for all three scenarios,
abortion, death penalty, and animal rights. Regardless of the anonymity
state, the questions were exactly the same aside from the part which de-
fined how a user sends the message i.e (see figure 6 in Appendix). if in
the abortion scenario the case was about pseudonymity, it would prompt
“You are writing to her parents explaining your feelings about the matter
and you sign your message with your nickname instead” . In addition
the answers in the likert scale were no different from each other in all
the different anonymity states aside from the signature at the end of the
message which would alter depending on the anonymity state.
Figure 4: Possible variations of the order that an individual received the
questions under each state
The reason behind this was that each individual would receive scenar-
ios in different anonymity states. Essentially participants were randomly
assigned to different anonymity variations of the scenarios. In effect, par-
ticipants might have answered the scenario about abortion with their real
name while other participants might have answered the same scenario
with their nickname. The only thing that stayed as a constant for all
the subjects was the order of the scenarios according to the anonymity
state. The first scenario was always based on the control anonymity state
10
The choice of complete anonymity versus pseudonymity for
aggression online
where anonymity was absent, the second was based on pseudonymity and
the final was based on complete anonymity. Hence, differences between
responses could be analyzed later on between different anonymity states
and different scenarios but also within the same scenarios but between
different anonymity states. Therefore if people gave more aggressive an-
swers due to different scenarios and not because of the different anonymity
states it could be determined in the analysis later on. On the other hand
if in all three scenarios we see the same pattern that could be indicative
of the anonymity factor having an effect on the users’ responses.
A pilot study of 10 people was conducted in order to determine any
errors with the survey. Errors in the text and the system were fixed
after reviewing the pilot results and feedback from the participants. The
anonymity states for each scenario were highlighted with bold colors in
order to avoid having users that might have missed the different anonymity
states.
Hypothesis Hypothesis statement
H1The way people respond is significantly different be-
tween the anonymity states of the study
H2There are significant differences between males and
females in the way they respond under anonymity
states
H3Users with stronger opinions in topics will respond
more aggressively under different anonymity states
H4Scenarios affect the differences between different
anonymity states
H5The state of pseudonymity may produce more ag-
gressive results compared to the case of complete
anonymity
Table 1: Main hypotheses of the study
4 Results
The total participants that took part on the survey were 290 of which 163
successfully went through the entire content of the survey, filling out all
the questions until the end. Almost all of the nonresponse cases occurred
in the first page of the survey. Therefore, all the data used for the analyzes
below consisted from the 163 participant sample and the rest was excluded
from the study. Of the 163 participants, 87 were males and 76 were females
or percentage-wise 53.4% men and 46.6% women. This is similar to the
percentages of men and women in the world population according to the
11
Michail Tsikerdekis
National Institute for Demografic Research (Institut national d’´etudes
emographiques). Most of the participants fell under the 20-30 age group
which accounted for 70.6% of the total sample. The 13-20 age group
accounted for 11.7%, the 30-40 for 12.9%, the 40-50 for 2.5% and the 50+
for 2.5% of the total sample. Especially for the categories of 40-50 and 50+
because the sample was so small, any correlation analysis with age groups
had to be treated with caution and scepticism. Finally, although there
were no questions asking the participants about their country of origin
since the survey was not designed to test for geographical differences, an
attempt was made to trace back individuals to their country of origin
through their account identifiers. A note should be made here that some
of the participants did not publicly reveal their location and therefore their
information was coded as Unknown. In terms of geographical distribution
per continent, based on a seven continent model, Europe accounted for
68.71%, North America for 12.27%, Asia for 6.75%, South America for
2.45%, Africa 1.64 %, Australia for 0.61% and Unknown for 7.36%.
The missing values that were found for the answers from all the data
available were 3, one answer for each anonymity state has not been prop-
erly received. These errors can be attributed to faulty communication
that could have occurred between the client sending the information and
the server receiving it, i.e. session timeout. These 3 missing answers ac-
count for 0.6% of the total 489 answers that were received from all the
scenarios and anonymity states combined and were not expected to have a
significant effect to the analysis that followed. Hence, they were registered
as missing values in the statistical analysis tool that was used.
4.1 Analysis as one group and within group differ-
ences for all scenarios
Because of the way the survey had been designed, the data could have
been analyzed in various ways. At first, the data was treated as com-
ing from one group of participants and analysis was performed for within
group differences. This type of analysis was chosen because the partici-
pants progressed through the different anonymity states which could be
perceived as treatments. The group consisted of all the scenarios be-
ing answered in all of the anonymity states. The median and standard
deviation values were, for the control state M= 1.85, SD = 1.031,
for pseudonymity M= 2.06, SD = 1.059, and for complete anonymity
M= 1.98, SD = 1.009. The first hypothesis was that there would be dif-
ferences between the different anonymity states. A Friedman’s test showed
no statistically significant differences between the different anonymity
states, χ2(2) = 4.671, N = 160, p = 0.097. Although the H1hypothe-
sis according to the above had been rejected, it could be due to the fact
that the answers to the questions were based on various scenarios. If that
is the case, analysis between groups across each individual scenario should
yield different results. This analysis is presented later on in this paper.
12
The choice of complete anonymity versus pseudonymity for
aggression online
Correlation tests were also conducted to see any possible relationships
between genders and how they answered the questions. No statistically
significant correlations were found for the states of Absence of anonymity
ρ= 0.082, N = 162, p = 0.303, Pseudonymity ρ= 0.128, N = 162, p =
0.105, Complete anonymity ρ=0.010, N = 162, p = 0.903.
One of the original hypotheses was that there might be some sort of
relationship between how strong the users feel that their opinion is about
a topic and how they respond to certain situations. Since in all of the
scenarios the problems are personal for the individuals, what might affect
how aggressively they respond would lie in how strong of an opinion users
felt they had about each topic. Spearman tests were performed. The
tests failed to show any statistically significant correlation between how
strongly opinionated a user was about the topic related to the scenario
and the aggressive response for the two anonymity states of complete
anonymity, ρ= 0.083, N = 162, p = 0.296, and the state where anonymity
was absent, ρ= 0.068, N = 162, p = 0.393. On the other hand, there was
a statistically significant correlation for the state were pseudonymity was
present, ρ= 0.288, N = 162, p = 0.000. This could also be seen at the
Table 2 where the more participants felt that they had a stronger opinion
about a topic, the more abrasively they would answer in the scenarios.
Strength of opinion
/ Response
Very polite Polite Rude Very rude Total
Somewhat strong 29 17 3 5 54
53.7% 31.5% 5.6% 9.3% 100%
Very strong 20 22 8 7 57
35.1% 38.6% 14.0% 12.3% 100%
Extremely strong 13 14 12 12 51
25.5% 27.5% 23.5% 23.5% 100%
Table 2: Percentages of responses per strength of opinion for the
pseudonymous state
4.2 Analysis as three different groups according to
the different topics of scenarios
After completing the first way of analyzing the data, there was a need to
see if the different topics that scenarios were based could have affected the
results. In other words individuals for example, might have felt the need
to be more polite in the case of the death penalty scenario, while in the
case of the abortion scenario they might wanted to act more aggressively.
Since each individual answered each topic in one of the different anonymity
states, the data were split according to each topic and the same analysis
13
Michail Tsikerdekis
was conducted but this time for each group separately.
Starting off with the scenario about abortion, a non-parametric one-
way analysis of variance test was performed. The Kruskal-Wallis test
failed to show a statistically significant difference between the different
anonymity groups, K= 1.383, df = 2, N = 162, p = 0.501. Similarly,
Kruskal-Wallis tests failed to show statistically significant differences for
the death penalty scenario, K= 2.081, df = 2, N = 161, p = 0.353, and
for the animal rights scenario, K= 5.529, df = 2, N = 163, p = 0.063.
4.3 Analysis as three different groups according to
different states of opinion for the answers of the
participants
In the first analysis, within groups analysis, a correlation was discovered
between how strong of an opinion individuals felt they had about a specific
topic and how aggressive their answer was. In the case of the pseudony-
mous state where a statistically significant correlation was found, the deci-
sion was made to divide the groups according to how strong their opinions
were. The outcome of this was three different groups that contained an-
swers for all three scenarios and anonymity states but were divided based
on how strong of an opinion users felt they had about each particular
topic. In other words, a user might have answered that he or she had
an extremely strong opinion about the death penalty but very strong for
the topics of animal rights and abortion. In effect that means that his or
her answer for the death penalty scenario was moved to the group with
the rest of the extremely strong opinion cases while the other two were
moved to the group with the very strong opinion cases. Hence, each group
contained answers from all three topics, and anonymity states but all of
the answers came from individuals that had the similar state of opinion
about each of the topics. Since we found that the different topics do not
affect the outcome of the answers this analysis will show how the strength
of opinion about the topics can affect the participants’ answers.
The first group consisted of answers to scenarios that came from users
that their beliefs on the topic were “somewhat strong” according to the
questionnaire. A Kruskal-Wallis based on the ordinal four point Likert
scale, showed no statistically significant difference between the different
anonymity states, K= 0.338, df = 2, N = 149, p = 0.844.
The next group to be tested was the one for those topics that the
participants felt that their opinion was “very strong”. Again Kruskal-
Wallis analysis was conducted for the ordinal data where no statistically
significant differences were found although the value of p was significantly
lower than in the previous test, K= 2.141, df = 2, N = 162, p = 0.343.
The final group that was tested was the one that consisted of answers
from participants that felt that their opinion about the topics was “ex-
tremely strong”. In this case the Kruskal-Wallis test showed a statistically
significant difference, K= 7.254, df = 2, N = 175, p = 0.027. Post hoc
14
The choice of complete anonymity versus pseudonymity for
aggression online
Anonymity State N Mean Rank
Absence of Anonymity 64 79.64
Pseudonymity 51 103.05
Complete Anonymity 60 84.12
Total 175
Table 3: Mean Ranks of Kruskal-Wallis test for the group with users that
felt they had an extremely strong opinion about the topics
Anonymity States Difference in
Mean Ranks
Least Significant
Difference between
Mean Ranks
Absence of Anonymity –
Pseudonymity
23.41 22.766
Absence of Anonymity –
Complete Anonymity
4.48 21.795
Pseudonymity – Complete
Anonymity
18.93 23.099
Table 4: Difference in Mean Ranks & Least Significant Difference for the
group with users that felt they had an extremely strong opinion about
the topics
tests for Kruskal-Wallis were conducted in order to determine where the
significant difference lied between the groups. The least significant differ-
ence between mean ranks was determined between all three groups and it
was compared with the mean ranks from the Kruskal-Wallis test with an
alpha = 0.05 [25] [30]. The results are shown in table 3 and 4. As can be
seen in table 4, the difference in the mean ranks between the case where
anonymity was absent and the case of pseudonymity was higher than the
least significant difference between the ranks and therefore, this is where
the statistically significant difference between the groups was. Looking
at the percentages for the answers of these two groups, it was obvious
that people that had answered with their pseudonyms chose more aggres-
sive responses than when compared to the cases where people sent the
messages with their real names.
15
Michail Tsikerdekis
5 Discussion and Conclusions
Looking at the results we can see that up to a point they coincide with
the previous empirical findings on anonymity and aggression. Even in
cases where no statistically significant differences were found the means
for each anonymity state were slightly higher than the mean of the con-
trol state. Since previous studies never compared the difference between
pseudonymity and complete anonymity this paper had to rely on theoret-
ical foundations of literature in order to form its hypotheses and assess
the expectations for the results.
As can be seen in the literature, aggression is affected by a number
of situational factors [5] [7] [19] [18] [8] [10] [31] [32]. According to this
study’s findings, just having the option of speaking out anonymously or
pseudonymously does not lead to an aggressive response alone. Under the
right circumstances though, the presence of pseudonymity can contribute
to aggressive exchange of messages which would have been absent or of
lesser effect if complete anonymity was used instead.
There is an undoubtedly strong correlation between how strong of an
opinion users feel they have about a topic and their aggressive response
in the case of pseudonymity. That correlation had to be further explored
in order to determine the level of association and as it can be seen by
the results, individuals that feel they have an extremely strong opinion
about a specific topic can and will become more aggressive in the way
that they respond in their messages. Therefore the initial suspicion that
dissociative imagination plays a big role in the case of nicknames has been
partially confirmed. In addition, it should be noted that while there was
no statistically significant difference between the case where people used
their real names and the case of complete anonymity, the mean ranks of
anonymity were slightly higher in every test. That fact gave the case of
complete anonymity a unique feature of being the intermediary between
the state where anonymity was absent and pseudonymity. In other words,
since it was no different in comparison to the other two states, although
the other two states differed from each other, we could conclude that it
lies somewhere in-between.
Of course all of the above apply only to the case where individuals felt
they had an extremely strong opinion about a topic. For the rest of the
cases the differences were not significant enough although, this does not
mean that it should not be taken seriously. The study has not explored
how one shifts from having a very strong opinion to an extremely strong
and therefore assuming that a social network contains only individuals
with moderate beliefs could be hasty and inaccurate.
5.1 Recommendations
According to the results and conclusions of this research, several sugges-
tions can be made for the software engineers designing social networking
16
The choice of complete anonymity versus pseudonymity for
aggression online
media. The first and most important, is knowing the population of the
social network. Understanding their beliefs and how strong the users
opinions are or even how many extremists a network has is important
regardless of the use of anonymity or not.
The second decision that has to be made after establishing a sufficient
knowledge of the population is the necessity to use anonymity in the
network. After deciding that anonymity has to be used because of certain
benefits that it might bring to the community, the third and final step is
considering the use of pseudonyms or complete anonymity.
In communities where highly controversial topics are being discussed
and a limited amount of the population feels that has an extremely strong
opinion about the topics, pseudonymity can be used, while taking into
account that a certain part of the population might be susceptible to
the aggressive effect that was shown in this study. On the other hand,
complete anonymity is recommended for communities where debates take
place regularly, and users that feel they have strong opinions about topics
can be found in a big percentage of the population. However, determining
how strong someone’s opinion is about a variety of topics might pose a
challenge. In this study a three point likert scale was used but larger
scales may work too.
In any case, due to the results that were brought to light by this ex-
ploratory study, caution is advised for the use of pseudonymity. Software
engineers that maintain online communities where aggressive incidents are
common might want to consider investigating if pseudonymity might be
the cause.
5.2 Limitations
There are certain limitations that may come with this study and caution
is advised when results have to be generalized for the Facebook popula-
tion. There might be a nonresponse bias from the participants that quit
in the first page of the survey even though that could still be attributed to
initial curiosity and afterwards unwillingness to participate in the survey
especially because of the lack of any incentives for the participants. An-
other thing to consider is that even though still under debate, convenience
sampling is not as exact as random. In a comparison study of convenience
and random samples of older adults differences were larger in some do-
mains than others but remained small to moderate in magnitude [15]. In
addition even though through the six degrees of separation all individuals
had a non-zero chance for being invited to the survey, certain individuals
closer to the first participants had higher chances than the ones further
apart in the chain.
Another thing that was not tested and might have produced some in-
teresting results, is the numerical association of a specific message with the
user. In forums where discussions might occur with complete anonymity,
users would be unable to follow if one message was written by the same
17
Michail Tsikerdekis
person, and could be confusing. One solution to counter this could be
to assign each individual with a random number id which could not be
traced back to the individual but the individual could submit messages
with the same numeric identifier as many times as they want. The im-
mediate question that arises is if that numeric identifier would have the
same effect as a nickname affecting the outcome of the message because
of dissociative imagination, or would it be perceived by the user at the
same level as complete anonymity?
5.3 Final thoughts
The concepts discussed in this paper show the potential benefits of under-
standing the effects of the design on the user interaction especially for the
case of anonymity and aggression. While these questions have been an-
swered about anonymity and aggression in this exploratory study, further
investigation of the phenomenon is needed. As demonstrated in this pa-
per reducing aggression is not impossible and it could be achieved with a
simple software alteration, changing pseudonymity over anonymity. Fur-
ther research and aspects that affect aggression online have to be explored
in order to enhance our understanding of the digital environment and its
effects on the user interactions. Since the goal of the social networking
software is to serve the interactions of users, we should not just focus on
understanding how these interactions relate to the interface and the ar-
chitectural design but also find ways to alter the environment in order to
make the future of online communication safer, more pleasant and more
effective.
References
[1] S. Ainsworth, G. Gelmini-Hornsbya, K. Threapletona, C. Crooka,
C. O’Malleya, and M. Budaa. Anonymity in classroom voting and de-
bating. Learning and Instruction, 21(3):365–378, 2010. doi:10.1016/j.
Learninstruc.2010.05.001.
[2] B. Al-Ani, G. Mark, and B. Semaan. Blogging in a region of conflict:
supporting transition to recovery. In Proceedings of the 28th inter-
national conference on Human factors in computing systems, CHI
’10, pages 1069–1078, New York, NY, USA, 2010. ACM. ISBN 978-
1-60558-929-9. doi: http://doi.acm.org/10.1145/1753326.1753485.
URL http://doi.acm.org/10.1145/1753326.1753485.
[3] R. Atkinson. Societal effects of cyberbullying: the dark side of build-
ing bridges with technology. In South Atlantic Philoshopy of Educa-
tion Society Yearbook., page 0, 2008.
[4] A. Barak, M. Boniel-Nissim, and J. Suler. Fostering empower-
ment in online support groups. Computers in Human Behavior, 24
18
The choice of complete anonymity versus pseudonymity for
aggression online
(5):1867 – 1883, 2008. ISSN 0747-5632. doi: 10.1016/j.chb.2008.
02.004. URL http://www.sciencedirect.com/science/article/
pii/S0747563208000198. ¡ce:title¿Including the Special Issue: In-
ternet Empowerment¡/ce:title¿.
[5] R. Barker, T. Dembo, and K. Lewin. Frustration and aggression: An
experiment with young children. University of Iowa Studies in child
Welfare, 18:1–314, 1941.
[6] H. Bechar-Israeli. From bonehead to clonehead : Nicknames, play,
and identity on internet relay chat. Journal of Computer-Mediated
Communication, 1:0, 1995. doi: 10.1111/j.1083-6101.1995.tb00325.x.
[7] L. Berkowitz and L. A. Weapons as aggression-eliciting stimuli. Jour-
nal of Personality and Social Psychology, 7(2):202–207, 1967.
[8] T. Castle and C. Hensley. Serial killers with military experience:
Applying learning theory to serial murder. International Journal of
Offender Therapy and Comparative Criminology, 46(4):453–65, 2002.
[9] D. Davenport. Anonymity on the internet: why the price may be too
high. Commun. ACM, 45:33–35, April 2002. ISSN 0001-0782. doi:
http://doi.acm.org/10.1145/505248.505267. URL http://doi.acm.
org/10.1145/505248.505267.
[10] J. P. Davis, S. Farnham, and C. Jensen. Decreasing online ’bad’ be-
havior. In CHI ’02 extended abstracts on Human factors in comput-
ing systems, CHI EA ’02, pages 718–719, New York, NY, USA, 2002.
ACM. ISBN 1-58113-454-1. doi: http://doi.acm.org/10.1145/506443.
506563. URL http://doi.acm.org/10.1145/506443.506563.
[11] C. Farkas, G. Ziegler, A. Meretei, and A. L¨orincz. Anonymity and
accountability in self-organizing electronic communities. In Proceed-
ings of the 2002 ACM workshop on Privacy in the Electronic Society,
WPES ’02, pages 81–90, New York, NY, USA, 2002. ACM. ISBN 1-
58113-633-1. doi: http://doi.acm.org/10.1145/644527.644536. URL
http://doi.acm.org/10.1145/644527.644536.
[12] R. G. Geen. Human aggression. Open University Press, Buckingham
- Philadelphia, USA, 2001.
[13] J. E. Guevara and H. Umemuro. Unobtrusive estimation of psycho-
logical states based on human movement observation. e-Minds, 2(6),
2010.
[14] S. C. Herring. The rhetorical dynamics of gender harassment on-
line. The Information Society, 15(3):151–167, 1999. doi: 10.1080/
019722499128466. URL http://www.tandfonline.com/doi/abs/
10.1080/019722499128466.
19
Michail Tsikerdekis
[15] D. F. Hultsch, S. W. MacDonald, M. A. Hunter, S. B. Maitland,
and R. A. Dixon. Sampling and generalisability in developmental
research: Comparison of random and convenience samples of older
adults. International Journal of Behavioral Development, 26(4):345–
359, 2002. doi: 10.1080/01650250143000247. URL http://jbd.
sagepub.com/content/26/4/345.abstract.
[16] A. Joinson. Disinhibition and the internet. Psychology and the In-
ternet, 2nd Edition:75–92, 2006.
[17] K. A. Keesom. Anonymity, accountability & john doe. In Proceed-
ings of the 1st annual conference on Information security curriculum
development, InfoSecCD ’04, pages 115–118, New York, NY, USA,
2004. ACM. ISBN 1-59593-048-5. doi: http://doi.acm.org/10.1145/
1059524.1059550. URL http://doi.acm.org/10.1145/1059524.
1059550.
[18] S. Kiesler and L. Sproull. Group decision making and commu-
nication technology. Organizational Behavior and Human Deci-
sion Processes, 52(1):96 – 123, 1992. ISSN 0749-5978. doi:
10.1016/0749-5978(92)90047-B. URL http://www.sciencedirect.
com/science/article/pii/074959789290047B. ¡ce:title¿Group De-
cision Making¡/ce:title¿.
[19] S. Kiesler, J. Siegel, and T. W. McGuire. Social psychological aspects
of computer-mediated communication. American Psychologist, 39
(10):1123–1134, 1984. URL http://content.apa.org/journals/
amp/39/10/1123.
[20] M. Lea, R. Spears, and D. de Groot. Knowing me, knowing
you: Anonymity effects on social identity processes within groups.
Pers Soc Psychol Bull, Vol. 27, No. 5(5):526–537, 2001. doi:
10.1177/0146167201275002.
[21] G. B. Lee. Addressing anonymous messages in cyberspace.
Journal of Computer-Mediated Communication, 2(1):0, 1996.
doi:10.1111/j.1083-6101.1996.tb00183.x.
[22] P. Marsden and J. Wright. Handbook of survey research. Emerald,
2010. ISBN 9781848552241. URL http://books.google.cz/books?
id=mMPDPXpTP-0C.
[23] A. Pfitzmann and M. Hansen. Anonymity, unlinkability, unde-
tectability, unobservability, pseudonymity, and identity management
(a consolidated proposal for terminology)., 2010. http://dud.inf.tu-
dresden.de/Anon Terminology.shtml. Version 0.34.
[24] A. Pinsonneault and N. Heppel. Anonymity in group support
systems research: a new conceptualization, measure, and contin-
gency framework. Journal of Management Information Systems,
20
The choice of complete anonymity versus pseudonymity for
aggression online
Vol. 14, No. 3:89–108, December 1997. ISSN 0742-1222. URL
http://dl.acm.org/citation.cfm?id=1189513.1189518.
[25] L. Portney and M. Watkins. Foundations of Clinical Research Appli-
cations and practice. Prentice Hall International (UK) Ltd., London.,
1993. ISBN 0-8385-1065-5 p. 430-432.
[26] Qing and Li. New bottle but old wine: A research of cyberbullying
in schools. Computers in Human Behavior, 23(4):1777 – 1791, 2007.
ISSN 0747-5632. doi: 10.1016/j.chb.2005.10.005. URL http://www.
sciencedirect.com/science/article/pii/S0747563205000889.
[27] S. A. Rains. The impact of anonymity on perceptions of source cred-
ibility and influence in computer-mediated group communication: A
test of two competing hypotheses. Communication Research, Vol. 34,
No. 1(1):100–125, 2007. doi: 10.1177/0093650206296084.
[28] R. W. A. Rawn and D. R. Brodbeck. Examining the relationship be-
tween game type, player disposition and aggression. In Proceedings of
the 2008 Conference on Future Play: Research, Play, Share, Future
Play ’08, pages 208–211, New York, NY, USA, 2008. ACM. ISBN 978-
1-60558-218-4. doi: http://doi.acm.org/10.1145/1496984.1497024.
URL http://doi.acm.org/10.1145/1496984.1497024.
[29] L. P. Sheridan and T. Grant. Is cyberstalking different? Psy-
chology, Crime & Law, 13(6):627–640, 2007. doi: 10.1080/
10683160701340528.
[30] S. Siegal and N. Castellan. Nonparametric Statistics for the Be-
havioral Sciences. McGraw Hill Book Company NY, 1988. ISBN
0-07-057357-3 p. 320.
[31] J. Suler. The online disinhibition effect. CyberPsychology & Behavior,
7(3):321–326, 2004.
[32] J. R. Suler. Online counseling: A handbook for mental health profes-
sionals, chapter The psychology of text relationships, pages 19–50.
CA: Elsevier, San Diego, 2004.
[33] M. Tanis and T. Postmes. Two faces of anonymity: Paradoxical
effects of cues to identity in CMC. Computers in Human Behavior, 23
(2):955–970, Mar. 2007. doi: 10.1016/j.chb.2005.08.004. URL http:
//dx.doi.org/10.1016/j.chb.2005.08.004.
[34] P. A. Thompsen and D. A. Foulger. Effects of pictographs and quot-
ing on flaming in electronic mail. Computers in Human Behavior, 12
(2):225–243, 1996. doi:10.1016/0747-5632(96)00004-0.
21
Michail Tsikerdekis
[35] M. Tsikerdekis. Engineering anonymity to reduce aggression online.
In K. Blashki, editor, Proceedings of the IADIS International Con-
ference - Interfaces and Human Computer Interaction 2011, pages
500–504. IADIS - International association for development of the
information society, Rome, 2011 2011. ISBN: 978-972-8939-52-6.
[36] A. Vasalou, A. N. Joinson, and J. Pitt. Constructing my online self:
avatars that increase self-focused attention. In Proceedings of the
SIGCHI conference on Human factors in computing systems, CHI
’07, pages 445–448, New York, NY, USA, 2007. ACM. ISBN 978-
1-59593-593-9. doi: http://doi.acm.org/10.1145/1240624.1240696.
URL http://doi.acm.org/10.1145/1240624.1240696.
[37] D. Villani, M. Lucchetta, A. Preziosa, and G. Riva. The
role of interactive media features on the affective response
: a virtual reality study. eMinds International Journal on
HumanComputer Interaction, I(5):35–55, 2009. URL http:
//www.martinweb.zobyhost.com/eminds/index.php?journal=
eminds&page=article&op=viewFile&path[]=53&path[]=38.
[38] P. Wallace. The psychology of the internet. Cambridge University
Press, 1999.
[39] D. J. Watts, P. S. Dodds, and M. E. J. Newman. Identity and search
in social networks. Science, 296(5571):1302–1305, 2002. doi: 10.1126/
science.1070120. URL http://www.sciencemag.org/content/296/
5571/1302.abstract.
[40] M. L. Ybarra and K. J. Mitchell. Online aggressor/targets, ag-
gressors, and targets: a comparison of associated youth character-
istics. Journal of Child Psychology and Psychiatry, 45(7):1308–1316,
2004. ISSN 1469-7610. doi: 10.1111/j.1469-7610.2004.00328.x. URL
http://dx.doi.org/10.1111/j.1469-7610.2004.00328.x.
[41] L. Zhang and W. Tu. Six degrees of separation in online society.
Distribution, 3(12):1–5, 2009. URL http://journal.webscience.
org/147/.
[42] P. G. Zimbardo. The Lucifer Effect: How Good People Turn Evil.
Rider & Co, Mar. 2008. ISBN 1846041031.
22
The choice of complete anonymity versus pseudonymity for
aggression online
Appendix
Figure 5: Both sides of the abortion scenario
23
Michail Tsikerdekis
Figure 6: Both sides of the same scenario through different anonymity
states
Figure 7: Animal rights scenario for participants that were pro animal
rights under the state where participants used their real names
24
... Bajo el lema de democratizar y reforzar la credibilidad del medio, generar un sentido de comunidad e intercambio con la audiencia y, de ese modo, "calibrar el peso de la opinión pública" (Mancera Rueda en prensa), comenzaron a circular, dentro del espacio público, intercambios verbales marcados por la agresión y la violencia más explícita. Sin una motivación clara ni ámbito que quedara ajeno a estas descargas violentas, que la corrección política había eliminado de los medios (Noblía 2000a, 2000b, 2004, 2012, los comentarios en los diarios digitales comenzaron a plantear una agenda pública definida por el lector en su diálogo con otros, en la que la confrontación, muy excepcionalmente, toma la forma de un debate real, de naturaleza argumentativa. La participación activa de los lectores, en realidad, se convierte en una actividad eminentemente evaluativa, que se articula a partir de contados patrones ideológicos, que se reiteran de manera sistemática en noticias de cualquier índole o tema. ...
... Por ello, es difícil juzgar cuándo se trata de un comentarista que manifiesta entusiasmo y está amablemente comprometido y cuándo se encuentra acosando. En segundo lugar, el anonimato y la pseudonimia, en sus diferentes grados (Tsikerdekis 2012), permiten que los usuarios puedan encontrar el modo de volver a ingresar a estos espacios, incluso después de ser bloqueados. ...
Article
Full-text available
Este trabajo tiene como objetivo presentar los resultados parciales de un proyecto de investigación en desarrollo que aborda el tratamiento de las noticias vinculadas con la pobreza en medios argentinos. En este caso particular, pondremos el foco en el análisis sobre los comentarios digitales publicados por los lectores en noticias que tienen como eje un tema político y social relevante, como es la pobreza, con el fin de identificar los modos en los que se articula el diálogo entre la noticia, el lector y el diario, en una dimensión que, si bien no se encuentra legitimada por el medio, forma parte de la publicación y, por tanto, es pública (Noblía 2015). El marco teórico es el del Análisis Crítico del Discurso (Fairclough 1995, 2003, 2006) y la metodología es cualitativa. Para el análisis lingüístico, esta investigación adopta el método sincrónico-diacrónico de análisis lingüístico de textos (Pardo 2011, 2014). El corpus está conformado por cuarenta noticias, relevadas entre julio de 2014 y junio de 2015, en el diario La Nación.com.
... Many studies 210 support the antisocial model. For example, users with pseudonyms on the internet had 211 more extreme opinions and were more aggressive toward others than those with their 212 actual names (Tsikerdekis, 2012). Those who believed that they would not be identified 213 on the internet also behaved more aggressively than those who did not (Wright, 2013). ...
Article
Full-text available
This study clarified the advantages of virtual communities on non‐victim experiences among lesbian, gay, bisexual, transgender, questioning, intersex, and asexual (LGBTQIA) individuals in Japan. A total of 3504 Pigg Party users, including 1390 LGBTQIA individuals, reported their experiences of victimization, perceived emotional support, and concealment of their gender/sexual identity in both physical and virtual communities. Japanese individuals with multiple minority statuses had more victim experiences than those with a single or without minority status. Furthermore, differences in victim experiences by gender/sexual minority status were lower in the virtual community than in physical communities. Similar tendencies were also confirmed on perceived emotional support and concealment. Virtual communities provided a more bias‐free social resource to Japanese LGBTQIA individuals than physical communities.
... On social media news streams there is a single, uniform look to all items, hence the platform itself flattens content and removes cues that could help readers differentiate between news items, the way they would do in real life. Furthermore, the anonymity of users on the Internet makes it difficult for users to discern who is spreading fake news, whilst pseudonymity (Tsikerdekis, 2012) allows users to claim expertise or knowledge that they don't possess on certain topics, or by co-ordinating multiple online identities they can make supporting statements to enhance the believability of information presented. ...
Technical Report
Full-text available
[This is NOT a technical report but there is no option for research report.] Since the early 2010s there has been a steady increase in the use of social media by states or terror organisations for the attempted manipulation of opinions and actions among people that are outside their legal control. Examples of this that have been reported in the media include the Russian hybrid warfare in the Ukraine and foreign interference in the US presidential election in 2016. For Norway, as well as other NATO countries, these developments represent a problematic new trend that requires different tools and skills to handle than what one has traditionally associated with influence operations. Although there is a large amount of documentation on these and other social media-based influence operations, little, if anything, has been done to try to explore how such campaigns might have an effect. The aim of this report is to support the Norwegian Armed Forces and the Ministry of Defence to develop a better understanding of issues around social media-based influence operations. This is done by going beyond a mere summary of influence activities that have taken place in social media. Instead this report takes a socio-technical approach to examine various aspects of social media-based influence and contextualise them within studies of online social behaviours and general sociology and ICT-related research. For this purpose, the report uses secondary data from several cases of social media manipulation, both state-organised and smaller, more organic attacks. From this base the report develops a conceptual chain that enables us to understand how an influence operation uses native aspects of social media to achieve its goals. In short, a planned influence operation is executed by active operators and relies on social media affordances (characteristics that facilitate certain activities). These affordances aid influence operations’ amplification and reach so that the content is spread widely and is added to the continuously aggregated and accumulated content stored by social media services. This vast content collection is referred to as the online information sediments. This metaphor is used to emphasise the long-term, cumulative approach of social media where information never disappears but will fade in and out of view depending on what a user is interested in, what they are searching for, and so on. New content is affected by the online information sediments as existing posts will provide material for framing and understanding any new information. Alternatively, new posts may affect existing content by providing new ways of interpreting old posts. Either way, the information from influence operations competes for individuals’ and groups’ attention in an attempt to enter into and manipulate their meaning making processes. The aim would be to get targeted social media users to do something that is beneficial to the actor behind the influence operation.
... Increased Internet use is also correlated with shifting social dynamics whose implications are not yet clear. In addition, anonymity and pseudonymity online seem to increase aggression (Tsikerdekis 2012). However, anonymity may also be useful for those seeking support following stigmatized crimes, like DFV, where victim blaming is common (Stark 2007: 112-3). ...
Article
Full-text available
The use of technology, including smartphones, cameras, Internet-connected devices, computers and platforms such as Facebook, is now an essential part of everyday life. Such technology is used to maintain social networks and carry out daily tasks. However, this technology can also be employed to facilitate domestic and family violence. Drawing on interviews undertaken with 55 domestic and family violence survivors in Brisbane, Australia, this article outlines survivors' experiences of technology-facilitated domestic and family violence. The frequency and nature of abusive behaviours described by the women suggest this is a key form of abuse deserving more signifcant attention.
... First, the anonymity: Anonymity certainly decreases inhibitions (Suler, 2004). This lowering of reticence can easily lead to non-normative behavior, and online anonymity has been clearly shown to be linked to Internet harassment and cyberbullying (for example, Ševčíková & Šmahel, 2009;Tsikerdekis, 2012). Moreover, in one study, participants were asked to use a sexual harassment hashtag on Twitter, either anonymously or with identifiable details. ...
Article
Video game players face a fundamental challenge in managing their competing desires for both privacy and publicity, for being both apart from, and a part of, the communities in which they play. In this paper, we argue that "gamertags" are important tools for protecting gamers' privacy as well as creative outlets for expressing meaningful aspects of identity. Based on 30 semi-structured interviews focused on players' usernames, we find through the pseudonyms under which they play, gamers both hide identifying information such as their offline names and addresses while bringing attention to information that is deeply meaningful to them, such as their family nickname or favorite music. By deemphasizing some parts of their identity and by emphasizing others, players not only shape how they are perceived by other gamers, but they also attempt to preclude accidental disclosure of more identifying information. We argue that gamertag practices thus constitute an important form of boundary work through which gamers actively seek to draw lines between their offline and multiple online worlds in the ways that they wish. We argue that gamers use these names to both protect and project aspects of their identities--at times even seeking protection through projection--as a way of addressing their competing desires to both conceal and reveal different aspects of their identities. As boundary work, players' efforts to carefully protect personally-identifying information and intentionally project personally meaningful information to their communities help them better manage their online identities, relationships with others, and overall data privacy.
Chapter
Demand for online education, which provides students with the ability to study around their work and family commitments, has increased considerably in recent years and is expected to grow further. However, there are key differences between online and on-campus education that give rise to unique and complex challenges for online educators. One potential challenge is apparent greater volatility of online students that can see online educators experience greater levels of instructional dissent. We have termed this phenomenon ‘eRage’—students communicating electronically with staff in a rude, antisocial manner to express disagreement or contradictory opinions regarding classroom issues. This chapter will examine the challenges of online education that could contribute to eRage; briefly examine the literature pertaining to instructional dissent and provide recommendations for online educators to manage this somewhat overlooked and clandestine issue moving forward.
Article
Discussions about recent state-run influence operations in social media often focus only on quantitative elements—the number of people interacting with fake news or how many tweets were sent by bots. This article suggests that understanding how influence operations in social media may affect individuals and groups requires a socio-technical approach to examine what is unique about the social media information environment and people’s interactions in and through these media. A socio-technical understanding emerges through the development of a model based on the Cyber Kill Chain that conceptualises the influence operation process as interlinked stages seeking alternate actions from a target audience.
Article
Deficiency of correctly implemented and robust defence leaves Internet of Things devices vulnerable to cyber threats, such as adversarial attacks. A perpetrator can utilize adversarial examples when attacking Machine Learning models used in a cloud data platform service. Adversarial examples are malicious inputs to ML-models that provide erroneous model outputs while appearing to be unmodified. This kind of attack can fool the classifier and can prevent ML-models from generalizing well and from learning high-level representation; instead, the ML-model learns superficial dataset regularity. This study focuses on investigating, detecting, and preventing adversarial attacks towards a cloud data platform in the cyber-physical context
Preprint
This research examines the role of online anonymity in shaping herding behavior in peer-to-peer (P2P) lending markets. Drawing on theories from social psychology literature, we argue that a lender forms different credibility perceptions toward preceding peers per their anonymity status, and then uses such perceptions to adjust her herding momentum toward them. We collect data from a leading P2P lending platform and classify individuals’ usernames into either anonymous or real-sounding group. A cross-classified multilevel model is developed to explain the variation in individual lending amounts, while accounting for lenders’ choice of loan listings and unobserved heterogeneity at the listing and the lender level. To our surprise, the results show that online anonymity intensifies herding: successors demonstrate a stronger herding magnitude toward predecessors who have anonymous usernames than those who have real-sounding ones. This finding, which we attribute to successors’ perception of high expertise toward online anonymity, challenges a conventional wisdom that considers identity concealment a negative factor for source credibility. An extended analysis further reveals that the uncovered positive effect of anonymity on herding is accentuated in the early stage of the fundraising window but attenuated on borrowers with real-sounding usernames; nevertheless, we find no such discrepancies between listings that are assigned with high-risk and low-risk credit grades. Our work contributes to literatures on the impact of anonymity and herding in online environments.
Article
Full-text available
Six degrees of separation is a well-known idea that any two people on this planet can be connected via an average number of six steps. Having succeeded in the real world, the theory even directly or indirectly motivated the invention of online societies. However, no much effort has been paid on checking if the theory really holds for online societies whose connection pattern may not be identical to the real world. This paper tries to give an answer to the question by both mathematical modeling and online measurements. The mathematical approach formulates the problem as a Minimum Diameter Problem in graph theory and evaluates the maximum and average values of the number of connections between any two random-selected community members. Measurements are conducted in three different kinds of online societies, namely ArnetMiner for academic researchers, Facebook for students, and Tencent QQ for teenagers in China. Analysis of these measurements verifies our theoretical findings.
Article
Full-text available
The Social Identity Model of Deindividuation Effects (SIDE) proposes that depersonalization of self and others is responsible for the effects of visual anonymity on group behavior. The authors investigated these mediating processes by assessing the effects of group-based self-categorization and stereotyping of others on group attraction within visually anonymous or video-identifiable groups communicating via computer. Structural equation modeling showed that visual anonymity increased group-based self-categorization, which directly increased attraction to the group and indirectly increased group attraction by enhancing group-based stereotyping of others. Visual anonymity had no effect on self-categorization in terms of a wider social category (nationality). Predictions derived from alternative perspectives that visual anonymity decreases group attraction by increasing impersonal task focus or by attenuating evaluation concerns were not supported.
Article
Cambridge: Cambridge University Press, 1999. 264 pages (£20.00 hardcover) ISBN: 0521632943 This book is a timely examination of web users' behaviour and the ways in which it affects others participating among the different electronic environments available. The Internet is a technology that has expanded rapidly over relatively few years and millions of users are interacting with each other using the new medium without having considered how this communication differs in quantity and quality from more established information channels. Using examples from previously established research in the field of social psychology and more recent studies on various aspects of the Internet phenomenon, the author considers how human behaviour is influenced by the peculiar characteristics of the new 'Web-World'. After a brief introductory chapter where some of the Internet jargon is usefully demystified, there follow two related chapters on how people 'invent' themselves on the Web, discussing role-playing, impression formation and management, and identity experiments. They examine how users attempt to overcome the lack of the usual non-verbal cues in face-to-face communication and add socioemotional expressiveness to their online personas. The author also discusses how this process of online self-projection can become altogether delusional, deceptive and dangerous as participants morph between generational, gender and personality profiles. The following two chapters examine the dynamics of group behaviour online. They illustrate the psychological phenomena of conformity, polarization, conflict and co-operation occurring in mailing lists, e-mail traffic, news and discussion groups and chat rooms. Surprisingly, much of group behaviour online has similar social regulation as their real life counterparts.
Article
This experimental study examined the effects of pictographs (typographic symbols typically used to express emotion) and quoting (responding to a message by including full or partial quotes from that message) on the perception of flaming (hostile verbal behavior) in electronic mail. It was hypothesized that (a) the presence of pictographs would reduce perceptions of flaming and (b) the use of quoting would increase perceptions of flaming. The results were generally supportive of the first hypothesis. A significant moderating effect on perceptions of flaming was observed in messages with pictographs, although that effect diminished as the intensity of hostility increased. Limited support was found for the second hypothesis; results suggest quoting may increase perceptions of flaming only when messages are overtly hostile. In addition to these findings, the study advances a clearer understanding of when a message becomes a flame, as the results suggest flaming is most clearly associated with the expression of antagonism.