Content uploaded by Senuri Wijenayake
Author content
All content in this area was uploaded by Senuri Wijenayake on Jun 23, 2020
Content may be subject to copyright.
55
antifying the Eect of Social Presence on
Online Social Conformity
SENURI WIJENAYAKE, The University of Melbourne, Australia
NIELS VAN BERKEL, Aalborg University, Denmark
VASSILIS KOSTAKOS, The University of Melbourne, Australia
JORGE GONCALVES, The University of Melbourne, Australia
Social conformity occurs when individuals in group settings change their personal opinion to be in agreement
with the majority’s position. While recent literature frequently reports on conformity in online group settings, the
causes for online conformity are yet to be fully understood. This study aims to understand how social presence i.e.,
the sense of being connected to others via mediated communication, inuences conformity among individuals
placed in online groups while answering subjective and objective questions. Acknowledging its multifaceted
nature, we investigate three aspects of online social presence: user representation (generic vs. user-specic
avatars), interactivity (discussion vs. no discussion), and response visibility (public vs. private). Our results show an
overall conformity rate of 30% and main eects from task objectivity, group size dierence between the majority
and the minority, and self-condence on personal answer. Furthermore, we observe an interaction eect between
interactivity and response visibility, such that conformity is highest in the presence of peer discussion and public
responses, and lowest when these two elements are absent. We conclude with a discussion on the implications
of our ndings in designing online group settings, accounting for the eects of social presence on conformity.
CCS Concepts:
•Human-centered computing →Empirical studies in collaborative and social com-
puting;Empirical studies in HCI.
Additional Key Words and Phrases: Online Social Conformity; Online Social Presence; Interactivity; Response
Visibility; User Representation; Majority Size; Task Objectivity; Self-condence
ACM Reference Format:
Senuri Wijenayake, Niels van Berkel, Vassilis Kostakos, and Jorge Goncalves. 2020. Quantifying the Eect of So-
cial Presence on Online Social Conformity. Proc. ACM Hum.-Comput. Interact. 4, CSCW1, Article 55 (May 2020),
22 pages. https://doi.org/10.1145/3392863
1 INTRODUCTION
Conformity is a widely observed social phenomenon where individuals adjust their personal opinions
to be in line with the group’s expectations in an attempt to be ‘liked’ within the group (normative
inuence) or to be considered ‘right’ (informational inuence) [
2
,
16
]. While conformity was initially
studied in face-to-face settings, understanding its mechanisms in online settings is becoming in-
creasingly important, primarily given that a growing range of societal interactions are now mediated
through online technologies [10,51].
Moreover, there is an increasing interest in online social conformity in the recent HCI/CSCW
literature. Individuals are seen to conform to contradicting opinions and norms set by majorities when
Authors’ addresses: Senuri Wijenayake, swijenayake@student.unimelb.edu.au, The University of Melbourne, Parkville,
VIC, 3010, Australia; Niels van Berkel, nielsvanberkel@cs.aau.dk, Aalborg University, Fredrik Bajers Vej 5, Aalborg, 9100,
Denmark; Vassilis Kostakos, vassilis.kostakos@unimelb.edu.au, The University of Melbourne, Parkville, VIC, 3010, Australia;
Jorge Goncalves, jorge.goncalves@unimelb.edu.au, The University of Melbourne, Parkville, VIC, 3010, Australia.
Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee
provided that copies are not made or distributed for prot or commercial advantage and that copies bear this notice and the
full citation on the rst page. Copyrights for components of this work owned by others than the author(s) must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires
prior specic permission and/or a fee. Request permissions from permissions@acm.org.
©2020 Copyright held by the owner/author(s). Publication rights licensed to ACM.
2573-0142/2020/5-ART55 $15.00
https://doi.org/10.1145/3392863
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:2 Senuri Wijenayake et al.
discussing political and social issues on social media [
43
,
44
], when completing online quizzes [
6
,
67
]
and visual judgement tasks [
29
], and when commenting on online news websites [
60
]. However,
limited research has been undertaken to understand what triggers conformity in online settings.
Furthermore, as online groups are inherently dissimilar to face-to-face groups due to reduced social
contextual cues and the lack of physicality in online communication [
45
], previous work on well-
understood determinants of face-to-face conformity (such as majority group size [
3
,
30
,
54
], task
objectivity (objective or subjective nature of the task) [
2
,
7
,
17
], and self-condence [
11
,
61
]), may not
be of relevance in online group settings. Thus, further work is required to systematically investigate
whether and how aforementioned determinants of face-to-face conformity manifest in online group
settings, as an initial step towards fully understanding online social conformity.
Furthermore, as conformity is a form of social inuence, the eect of social presence i.e., the
awareness of and the sense of being connected to others via mediated communication [
58
], on
online conformity behaviour has been an interest in recent literature [
33
,
37
,
47
]. However, the
aforementioned studies limit their focus to only one aspect of online social presence (such as user
representation [
47
], level of interactivity [
33
] or response visibility [
37
]), where as in realistic online
settings these aspects are more likely to manifest together (e.g., online discussions usually involve
certain user representations). Thus, in this study, we contribute to the existing literature by investi-
gating the compound eects of three aspects of online social presence – user representation (generic
vs. user-specic avatars), interactivity (discussion vs. no discussion) and response visibility (public
vs. private responses) – on online conformity behaviour. Moreover, we investigate the eect of social
presence in both objective and subjective quiz tasks, as individuals are challenged by majorities
of dierent group sizes, while also accounting for their level of self-condence to provide a wider
understanding of online conformity determinants in comparison to existing work.
Our results reveal main eects from task objectivity, group size dierence (dierence between
the majority group size and the minority group size), and initial self-condence of the participant.
We observe higher conformity in objective questions than in subjective questions, contradicting the
ndings of oine social conformity [
7
], likely due to the inherently lower social presence in online
settings. We also note that larger majorities are more inuential than smaller ones as previously
suggested by both oine [
3
,
20
,
54
] and online [
42
,
65
,
67
] conformity literature. Moreover, partic-
ipants are more likely to conform to the majority when they are less condent or uncertain of their
initial answers, expanding the implications of existing literature on self-condence and conformity
in face-to-face groups [11,61] to online settings.
While we do not observe main eects for the three aspects of social presence manipulated in this
study, we note a statistically signicant interaction eect between interactivity and response visibility.
Conformity peaks in the presence of peer discussion and public responses, and remains at nominal lev-
els in their absence. Thus, our ndings imply that in addition to well-known conformity determinants
such as majority size, task objectivity, and self-condence, the level of social presence facilitated in an
online setting plays a major role in users’ conformity behaviour, making them especially susceptible to
normative inuences. We further discuss how this may be desired in settings where encouraging group
norms is important (e.g., online support groups [
56
,
57
]), while it may have detrimental eects when
conformity is not desired (e.g., online learning environments [
6
]). Thus, our ndings should be taken
into consideration when designing for online group settings in order to promote positive interactions.
2 RELATED WORK
In 1955, Asch set out to investigate the eect of peer opinions on individual judgements in face-to-
face group settings [
3
]. He employed a simple ‘line matching’ experiment, where for a signicant
number of responses (36.8%) participants conformed to a clearly incorrect yet unanimous majority,
establishing the susceptibility of individuals in group settings to conformity inuences. Subsequently,
Deutsch and Gerard [
16
] replicated Asch’s experiment to reveal two motives behind group conformity:
normative and informational inuences. They described that individuals conform to the majority
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:3
either with the intention of agreeing to their positive expectations (normative inuence) or because
they perceive information obtained from others to be more accurate evidence of a given situation
than their own knowledge (informational inuence). The notion of normative inuence was further
explained in [
14
,
38
] as the inclination to ensure one’s membership in a group, and seeking direction
from the majority in uncertain situations as a manifestation of informational inuences [13,38].
Even though seminal social conformity literature is based on oine groups, owing to the recent
movement of human societal interactions towards online platforms (e.g., social networks, learning
platforms, discussion forums, support groups) [
10
,
21
,
40
,
51
,
63
], a growing number of HCI/CSCW
researchers have sought to investigate the manifestation and implications of social conformity on
online group settings as described below.
2.1 Online Social Conformity
Recent literature suggest that social conformity has diverse manifestations in a wide variety of
online settings [
6
,
29
,
43
,
44
,
56
,
60
,
69
]. For instance, a study by Zhu et al. [
69
] investigated how
individuals adjust their online choices when challenged by recommendations from other users. The
study required participants to chose between two options, with and without the knowledge of others’
preferences. They observed that an individual’s choices in an online setting can be signicantly
swayed by others’ opposing recommendations. Similarly, Maruyama et al. [
43
] note that people tend
to adopt the majority’s opinions when discussing social issues on online social networks, even when
they are unaware of the users who are posting the content. Another study highlighted that users
who were actively involved in Twitter during a televised political debate were more inclined to adjust
their voting choice to support the majority sentiment on Twitter, further establishing the presence
of social conformity in online settings [44].
Literature also provides evidence of the negative consequences of social conformity in certain on-
line settings. For instance, Beran et al. [
6
] note that students who saw peer answers when completing
an online quiz frequently conformed to the majority’s answers, and obtained fewer correct answers
compared to students who answered the quiz independently. A similar study by Hullman et al. [
29
]
further emphasised the eect of social information (i.e., the notion that we tend to believe things
more when we see others doing them) on the accuracy of a simple visual judgement task completed
by Mechanical Turk users. They highlight that seeing biased and incorrect social information led
to more errors, which could eventually nullify the expected benets of collective intelligence.
However, online social conformity is not without its potential benets. For example, Sukumaran
et al. [
60
] highlight how normative inuences can be used in online news websites to encourage
high quality and ‘thoughtful’ contributions from its users. They note that when initial comments
are of high quality, subsequent participants were also encouraged to contribute with similar eort.
Similarly, previous work has shown that conforming to the acceptable conventions of behaviour
and linguistic norms improved the sense of belonging and security within an online mental health
support group, so that sensitive issues could be openly discussed [56].
The aforementioned literature indicates both positive and negative implications of online social
conformity. We therefore argue that a thorough understanding of factors inuencing this social
phenomenon is critical to ensure that future online platforms facilitating social interactions are
designed accounting for possible conformity eects. Therefore, we next summarise the existing
literature on major determinants of social conformity as seen in both oine and online settings.
2.2 Determinants of Social Conformity
2.2.1 Majority Size. The majority group size (or the number of inuential sources) on conformity
behaviour has been popularly researched in oine groups and several theories have been put forward.
For instance, Asch [
3
] observed that against a minority of one, the inuential power of the majority
increased until its third member, while adding a fourth member to the majority did not increase its
conformity inuence. Alternatively, Latané [
34
] proposed that while larger majorities have greater
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:4 Senuri Wijenayake et al.
impact, the added impact is smaller for each additional group member. Similarly, the notion that
larger majorities lead to higher conformity was further conrmed in subsequent studies [
30
,
46
,
54
].
However, the above studies considered unanimous majorities against a minority of one (participant)
and did not account for the possibility of having larger minorities – which is typically the case in
realistic online group settings.
Furthermore, a study by Lowry et al. [
42
] compared conformity behaviour across two group sizes
(groups of three and six members) in both face-to-face and computer-mediated communication
(CMC). They highlight that while conformity eects heightened in both conditions as the majority
group size increased, the eect was lowered in the computer-mediated condition. Moreover, Walther
et al. [
65
] investigated the eects of majority and minority group sizes on conformity in CMC group
settings. Interestingly, their results show that the presence of minorities disturbing the unanimity of
the group reduce the impact of the majority group size on conformity. However, while the signicance
of majority and minority group sizes on conformity behaviour has been suggested in the current
literature, it is yet to be thoroughly investigated in an online setting.
2.2.2 Task Objectivity. Literature suggests that task objectivity can also play a signicant impact
on conformity behaviour. While conformity was initially tested in tasks of objective nature with
an obvious correct answer [
3
,
16
], researchers were later interested in investigating how conformity
manifests in tasks of subjective nature. For instance, Ferguson [
17
] observed conformity in tasks of
attitudinal nature. Subsequently, Blake et al. [
7
] compared conformity eects across objective and sub-
jective tasks. The authors note that higher conformity was observed in subjective tasks in comparison
to objective tasks, as a result of higher normative inuences in physical groups. They highlight that the
motivation to achieve correct answers may have outweighed the appeal of conforming to an incorrect
majority in objective questions. However, further work is required to understand whether and how
objective and subjective questions may dier in eliciting conformity behaviour in online settings.
2.2.3 Self-confidence. Several studies have also investigated the eect of participant condence
on conformity behaviour [
11
,
61
]. Their results in unison emphasise a negative relationship between
self-condence and conformity [
11
,
61
]. This notion is in line with Deutsch and Gerard’s view of in-
formational inuences, which suggests that individuals conform to the majority seeking the ‘correct’
response in uncertain situations [
16
]. However, the impact of self-condence on social conformity
is yet to be understood in online group settings.
2.2.4 Social Presence. Social presence has been described in the literature as the the awareness of
and the sense of being connected to others via mediated communication [
58
]. It has also been dened
as the ability of individuals to project themselves socially and emotionally as ‘real’ people in mediated
communication [
19
,
23
]. Furthermore, Short et al. [
58
] explained that social presence facilitated by
a medium depends on its ability to convey the presence of the communicators through verbal and
non-verbal cues. Despite the early perception that CMC is impersonal due to the absence of social
context cues, results of subsequent studies contradicted this notion [
23
,
24
,
62
]. These studies further
emphasised that online social presence is a complex, multifaceted concept which manifests itself
across multiple dimensions (e.g., social context cues, interactivity and privacy), each with its own
variables (e.g., interactivity could be measured in terms of timely responses, communication style,
formality of language etc.) [24,62].
Subsequently, a study exploring the impact of dierent communication mediums (e.g., face-to-face,
telephone, chat, and email) on perceived social presence and interpersonal perceptions, identied that
people are more likely to behave in a manner to be liked by others in a richer communication medium
eliciting higher social presence [
15
]. More recent literature has built on this notion to understand the
impact of dierent aspects of online social presence (such as user representation, interactivity, and
response visibility) on social conformity. For example, previous work has observed that online user
representations (a social context cue) with more anthropomorphic (human-like) features encouraged
higher social attractiveness and trustworthiness of online interaction partners, leading to higher
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:5
conformity in online group settings [
22
,
37
]. This behaviour was further explained in work by Nowak
and Biocca [
47
], describing how the agency (the perception of communicating with a computer agent
or a human being) and anthropomorphism oered by online user representations could signicantly
inuence users’ perceptions of social presence. Researchers highlight that participants reported
higher social presence when representations were more realistic, human-like (high anthropomorphic)
and perceived to represent a real human being. However, the online user representations used in
the above studies were stereotypically gendered (which may also pose a signicant impact on online
conformity behaviour [
35
,
36
,
67
]) and are outdated (e.g., [
37
] compared text boxes against, stick
gures and animated characters). Thus, further work is needed to explore the generalisability of
these observations to more modern user representations.
Interactivity is another major dimension of online social presence [
62
], and has been investigated
with regard to online social conformity. For example, Laporte et al. [
33
] compared conformity be-
haviour in participants answering an online quiz under two conditions: in a unidirectional setting
where participants could see the answers of others alongside their prole pictures (no interaction),
and in a bidirectional setting where participants were connected through a live video chat, capable of
freely communicating with each other. They note higher conformity among participants in the live
video condition, conrming the notion that individuals are more likely to conform in online settings
with higher interactivity. Moreover, literature also supports the notion that online discussions could
elicit both normative and informational inuences leading to changes in an individual’s opinions [
49
].
Furthermore, the impact of response visibility (i.e., whether user responses are visible to others
or not) has been an interest within conformity research. The literature notes that conformity is
considerably higher when users are informed that their responses are visible to the group (public) in
comparison to situations where user responses are private [
37
]. Deutsch and Gerard [
16
] explained
that being aware of others in public conditions leads to participants being susceptible to normative
inuences, in addition to the informational inuences observed in private conditions.
While the impact of user representation, interactivity, and response visibility on online social
conformity has been investigated independently, existing work does not account for their interac-
tion eects, despite their likelihood to manifest simultaneously in realistic online settings. Thus,
further work is required to investigate the impact of their interaction eects on online conformity.
Moreover, it is possible that the eect of social presence on online conformity also depends on other
well-known determinants such as majority group size, task objectivity and self-condence, which
is yet to be thoroughly investigated. Thus, in this study we aim to systematically examine how the
aforementioned aspects of social presence impact online conformity behaviour in both objective
and subjective tasks, with dierent majority – minority group compositions, while also accounting
for eects of self-condence.
3 METHOD
To investigate the eect of social presence on conformity behaviour in an online setting, we manip-
ulate the level of social presence in terms of user representation (generic vs. user-specic avatars),
interactivity (discussion vs. no discussion), and response visibility (public vs. private responses). To
control these variables of online social presence, while simulating a plausible online group setting
where users are required to make judgements, we deployed an online multiple choice questions
(MCQ) quiz containing both subjective and objective questions. MCQ quizzes are frequently used
in online social conformity experiments [
6
,
33
,
53
,
67
,
68
] as they enable the simulation of a clear
group majority and a minority while placing the participant in both these groups.
3.1 The iz
The online quiz contained 18 multiple choice questions (9 objective and 9 subjective questions). The
objective questions were extracted from both existing literature [
67
] and popular online question
repositories such as Sporcle and Britannica, which have been previously used in [
67
,
68
] to extract quiz
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:6 Senuri Wijenayake et al.
Fig. 1. Steps followed during the quiz by participants. Step 1: Initial answer and confidence, Step 2: Peer
answers, Step 3: Peer discussion, Step 4: Update answer and confidence, Step 5: Response visibility
questions. This ensured that questions were only included if they were based on topics considered as
general knowledge within the community considered for this experiment. Previous work investigat-
ingonlinesocialconformitythrough quizzes have alsoutilisedgeneralknowledgetopics in their exper-
iments [
33
,
53
,
67
,
68
]. The subjective questions were extracted from an article in ThoughtCo. outlining
debating topics for high school students. We ensured that no overly sensitive topics were discussed
in the quiz. A complete list of questions utilised in the quiz is included as supplementary material.
Participants followed the structure illustrated in Fig. 1to complete the quiz. First, the participant
is instructed to select their personal answer and rate their condence on the chosen answer as
illustrated in Step 1 of Fig. 1. Self-reported condence levels were denoted using a scale ranging
from 0 – 100 with higher values representing higher levels of condence. Subsequent to submitting
their initial answer and condence, the participant is shown a fabricated list of peer answers as
chosen by four other participants completing the quiz (along with either generic or user-specic
avatars as shown in Step 2(a) and Step 2(b) in Fig. 1). The fabricated peer answers were dynamically
generated by our software to show the distribution of votes from other participants across a clear
majority and a minority, while placing the participant in either group. The participant’s answer was
highlighted for convenience. This notion of using fabricated peer answers to investigate online social
conformity was motivated by recent literature on conformity [
53
,
67
,
68
]. The next step determined
the level of interactivity (discussion vs. no discussion) facilitated during the study. Participants in
discussion conditions were given two minutes to discuss the group answers and their justications
with ‘peers’ through a real-time and text-based group chat as shown in Step 3 of Fig. 1. In reality,
confederates of the researchers were used to simulate ‘peer’ discussions based on a predetermined
script. Subsequently, all users were given the opportunity to change their answer option and/or
condence to nalise their answer (Step 4). Next, the response visibility (i.e., whether or not the nal
answers are visible to the rest of the group) was manipulated as shown in Step 5 of Fig. 1. Participants
in public conditions were shown the nal answers of the group with changed answers highlighted
in red. After viewing the group’s nal answers, the user is taken to the next question.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:7
3.2 Group Composition
We chose an overall group size of ve users (i.e., the size of the majority and the minority sums up
to ve) to investigate the eect of dierent majority – minority group distributions as illustrated
in Fig. 2. While participants were informed that they will be connected with four peers to complete
the quiz, there was only one real participant completing the quiz at any given time. Previous work
investigating social conformity in both face-to-face and online groups has employed a similar group
size [7,16,30,37].
During the quiz, group answers were fabricated such that the participant was evenly placed in
the majority (see group compositions (b) and (c) in Fig. 2) and in the minority group (see group
compositions (d) and (e) in Fig. 2). We also simulated group consensus (in two questions) to provide
a sense of authenticity to the peer answers. Moreover, we ensured that each group combination was
equally tested against topics of both objective and subjective nature.
Fig. 2. Overview of group compositions investigated in the quiz (participant highlighted in red).
3.3 Social Presence
The main objective of this study was to investigate the compound eects of online social presence
amidst other well-known determinants of conformity. Thus, we manipulated online social presence
using three variables – user representation (generic vs. user-specic avatars), interactivity (discussion
vs. no discussion) and response visibility (public vs. private responses) – which resulted in a 2 x 2 x
2 experimental design (8 conditions). Participants of each experimental condition interacted with a
unique combination of the aforementioned interface elements. For example, participants assigned to
the generic xdiscussion xpublic condition were represented using a single generic avatar, discussed
answers with peers via a group chat, and saw nal group answers, whereas participants in the
user-specic xno discussion xprivate condition were denoted using user-specic avatars and did not
see a group chat or nal group answers. We now describe these experimental conditions in detail.
3.3.1 User Representation. User representations are important social context cues contributing
to the perceived social presence in online settings [
62
]. In this study we investigate the impact of
two commonly used online user representations: generic and user-specic avatars. In generic avatar
conditions, a commonly used gender-neutral avatar is assigned to all the ve users along with generic
usernames such as "User 1" and "User 2" to dierentiate between participants (see Step 2(a) in Fig. 1).
Alternatively, in the user-specic avatar conditions, users are assigned dynamically generated avatars
including the rst letters of their rst and last names (e.g., John Doe is represented by
JD
as shown
in Step 2(b) in Fig. 1).
We highlight that our choice of user representations is based on literature explaining how agency
and anthropomorphism associated with dierent online user representations could impact social
presence in virtual group settings [
22
,
37
,
47
]. Based on the evidence provided in literature we
hypothesise that user-specic avatars with user initials convey a stronger sense of being connected
to a ‘real’ human being, than a single generic avatar with computer generated usernames. Moreover,
the practice of assigning a default generic avatar to users is common in online social networks such
as Twitter and YouTube and even in Learning Management Systems such as SAP Litmos and Do-
cebo, where user decisions are likely to be inuenced by others. Alternatively, some popular online
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:8 Senuri Wijenayake et al.
platforms such as Google use user-specic avatars with user initials to represent them. The user
representations used in this study are purposefully devoid of explicit user cues (such as gender, name,
or age) to avoid confounding eects on online social conformity [12,32,35,36,39,67].
3.3.2 Interactivity. We consider two levels of interactivity in this study: discussion and no discussion.
This determines whether participants are given an opportunity to engage in a discussion with peers
after viewing their answers (i.e., discussion) or not (i.e., no discussion).
In discussion conditions, users are informed that after being shown the peer answers, they will
be given two minutes to discuss their answers and rationale behind selecting it, with others in the
group as shown in Step 3 of Fig. 1. However, as we recruited only one participant per session, four
confederates of the researchers pretended to be users participating in the quiz, to simulate a group
discussion. The confederates engaged in the group discussion based on a script, dynamically created
by the software for each question based on the fabricated answer distribution. Confederates have
been used to simulate real users in previous work investigating social conformity in both face-to-
face [
7
,
16
,
30
] and online [
33
] group settings. Alternatively, in no discussion conditions, participants
are only shown the fabricated answer list of the group members. They are not subsequently given
an opportunity to engage in a group discussion, and are directly taken to Step 4 of Fig. 1.
3.3.3 Response Visibility. After the peer answers are shown, participants are given the opportunity
to update their answers and/or condence prior to submitting their nal answers (see Step 4 of Fig. 1).
Response visibility determines whether the nal answers are visible to the group (i.e., public) or
not (i.e., private). Participants in the public conditions are informed prior to the quiz that their nal
answers will be visible to the group. Therefore, they are shown the list of updated answers of the
group before moving to the next question, with updated answers highlighted in red as shown in
Step 5 of Fig. 1. Moreover, to ensure that participants in public conditions are not suspicious of the
authenticity of the answers, in 50% of the questions where at least one other simulated participant
was placed in the minority with or without the user (e.g., see (b), (c) and (d) in Fig. 2), the said simulated
participant changed their answer to that of the majority.
Alternatively, participants in private conditions are told that their nal answers will not be visible
to others in their group, and are taken to the next question upon submitting their nal answers.
3.4 Participants and Procedure
We recruited 64 participants (32 men and 32 women) from dierent educational backgrounds which
included engineering, science, marketing, management, arts and architecture elds. All participants
were between 19 – 36 years of age and were recruited through our university’s online notice board.
Participants were equally distributed among the 8 experimental conditions with an equal number
of men and women participating in each condition.
The study was conducted in a laboratory with one participant per session, under the supervision
of a researcher. Participants were informed that they would be taking part in an online quiz together
with four other participants. As the true purpose of the study could not be disclosed prior to the
quiz as expected in studies investigating conformity behaviour [
59
], we explained that the study
was motivated by the increasing use of online learning platforms and that we intend to investigate
the performance of students in online quizzes. No other information triggering a competing or
cooperative relationship between the participants was provided.
Before starting the quiz, participants completed an online form which collected their self-disclosed
gender, age, and educational background. Upon submitting their demographic details, participants
were randomly assigned to an experimental condition. The steps followed by the participant, and
the interface they interacted with, depended on the condition they were assigned to as illustrated in
Fig. 1. For instance, only participants assigned to discussion conditions, were shown an online group
chat area and were greeted by a conversational agent named ‘QuizBot’. This chat area was not visible
to participants in no discussion conditions. Alternatively, participants in no discussion conditions
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:9
were given textual, step-by-step instructions through the study software (with no involvement of
the researchers) prior to initiating the quiz.
The ‘QuizBot’ was utilised in discussion conditions to regulate group discussions (see Fig. 3) without
the involvement of a researcher. The absence of a researcher closely replicates realistic online group
settings, while also reducing the Hawthorne eect, a crucial aspect in conformity research [
1
,
66
].
The bot also signalled the participants when to start the discussion after viewing the peer answers
and displayed the remaining time left for discussion, to ensure that group discussions were restricted
to the allocated time frame of two minutes per question. The group chat was automatically disabled
by the bot after the allocated time, and the users were prompted to the next step. Furthermore, the
‘QuizBot’ provided confederates regular updates on what the participant was doing before and after
the discussion (i.e., answering the question, changing the answer after the discussion etc.), displayed
the fabricated list of peer answers, and provided them prompts of the discussion points based on
a predetermined script to minimise confusion and error.
Fig. 3. izBot providing initial quiz instructions and informing participant status in the discussion conditions.
Upon completion of the quiz, participants participated in a brief semi-structured interview. The
interview was arranged as follows. First, participants were asked for general thoughts on their
experience participating in the study, to identify if they were suspicious of the authenticity of other
participants. Subsequently, we inquired whether they changed their answers during the quiz and
what factors contributed towards such behaviour. We were also interested in whether they were more
inclined to change their answers in certain types of questions, to understand how task objectivity may
aect their behaviour. Participants were also questioned about the impact of discussion, response
visibility and user representations they were exposed to, to determine whether and how these three
aspects of online social presence aected their conformity behaviour. Subsequently, we debriefed
our participants on the true objective of the study. Participants were then given the opportunity
to withdraw their participation and data collected during the study, if desired. All our participants
consented to the use of the data collected, even upon revealing the study’s true objective.
The experimental design was approved by the Ethics Committee of our university. The experiment
lasted for approximately 60 – 90 minutes per participant depending on the experimental condi-
tion they were assigned to, including brieng, completing the quiz, and the nal interview. Each
participant received a $20 gift voucher for participation.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:10 Senuri Wijenayake et al.
3.5 Preliminary Study
Prior to the experiment, we conducted a preliminary study with 10 men and 10 women, where
participants were asked to answer the same set of MCQs along with justications for their answers.
We obtained a total of 360 responses from the preliminary study. Next, we arranged the four answer
options in each MCQ based on a descending order of the number of votes they received during the pre-
liminary study, to determine the most popularly chosen answers for each question. This orderwas later
considered along with the answers chosen by the participants in the main experiment, to dynamically
determine the positioning of the majority and the minority groups when fabricating peer answers. For
instance, in the question shown in Fig. 1(“What is the largest country in the world (by area)?”), “Russia”
and “Canada” were the top two answers chosen by the preliminary study participants. Thus, when a
participant in the main study selected “Russia” as their initial answer, our software dynamically fabri-
cated the peer answers placing “Canada” in a clear majority. The same approach was used to decide the
minority’s answer, when participants were placed in the majority. This ensured that the majority or
the minority was always placed in a plausible answer option, regardless of being correct or incorrect.
Moreover, the justications provided by participants in the preliminary study were used to create
a script which was used by confederates to support their answers during the main experiment (in
discussion conditions). This ensured that all justications provided by the confederates simulating
participants, closely represented how the considered community perceived the topic in question. This
was especially crucial in the subjective questions. For objective questions, we chose justications
that could be used with all four of the answer options (e.g., “Canada/Russia/China/USA is huge! I also
remember this fact from my geography class in high school.”), and were counterbalanced among
the four options during the experiment. We ensured that the justications used in the nal script
did not include any obvious personal cues (such as gender or age) which could result in confounding
eects on social conformity [
12
,
32
,
55
]. None of these explanations were explicit on the condence
level of the ‘participant’, or displayed uncertainty in the chosen answer. For instance, in the question
shown in Fig. 1the explanations provided by confederates could be as follows:
Confederate 1: “I remember from the world map that
Canada
is the biggest country in the world.”
Confederate 2: “
Canada
is huge! And because it covers the most area as compared to other
countries.”
Confederate 3: “I have read about this. Also by judging from remembering the relative size of
Canada on a map of the globe.”
Confederate 3: “I remember this as
Canada
from geography in high school and a rough memory
of the world map.”
4 RESULTS
All 64 participants answered 18 questions each, which resulted in a total of 1152 responses. Of these,
participants were placed in the minority in 512 responses and in the majority or in a consensus in 640
responses (equally distributed among objective and subjective questions), to avoid drawing suspicion
to the plausibility of the fabricated peer answers. On that note, we emphasise that our intention
was not to compare results between majority and minority responses, but rather to investigate the
impact of social presence on conformity across a diverse range of group compositions.
Upon displaying the fabricated peer feedback, participants were given the opportunity to:
•Change both initial answer option and condence level.
•Change only their initial answer option.
•Change only their initial condence level.
•Make no change to their initial answer.
Participants changed their initial answer and/or condence at least once during the quiz, resulting in
a total 431 changed responses with an average of 6.7 changes (
𝑆
D = 4.1) per participant. Fig. 4illustrates
the distribution of the participants’ post-feedback responses, when their answer was supported by
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:11
a minority, or by a majority during the quiz. Participants were more inclined to change one’s answer
(with or without a change in condence) when placed in a minority (in approximately 34% of the
minority responses). On the other hand, being placed in the group majority was more likely to result
in an increase in participant’s self-condence on selected answer (in approximately 28% responses).
Thus, these results establish that participants changed their answers post feedback, not randomly
but due to the inuence of the predictors we considered, conrming the validity of our results.
0
10
20
30
40
50
60
70
80
90
100
Minority responses Majority responses
% of responses
Changed both
Changed answer only
Changed confidence only
No change
Fig. 4. Distribution of minority and majority responses across the four post-feedback response types.
4.1 Model Construction
For the purpose of this study, we dene conformity as the act of changing one’s answer to that of
the majority. Our results show that 55 participants conformed at least once during the quiz, resulting
in 152 conformity responses (approximately 30% conformity), with an average of 2.4 conformity
responses (
𝑆
D = 1.6) per participant. We then investigated the impact of the following ten predictor
variables on the conformity behaviour of our participants. For our model construction, we only
consider the responses of participants when placed in a minority, as the dependent variable was
determining conformity behaviour.
•Majority size: Size of the majority (possible values: 3 or 4).
•Minority size: Size of the minority (possible values: 2 or 1).
•Group size dierence
: Dierence between the majority group size and the minority group size
(possible values : 1 or 3).
•User representation
: The avatar used to represent users in the online platform (possible values:
generic or user-specic).
•Interactivity
: Whether or not users were given an opportunity to discuss peer answers with the
group (possible values: discussion or no discussion).
•Response visibility
: Whether or not the nal group answers were visible to others (possible
values: public or private).
•Task objectivity: Subjective or objective nature of the question.
•Initial condence: Participant’s condence in answer prior to revealing peer answers (ranging
from 0 to 100).
•Gender: Participant’s self-disclosed gender.
•User ID: An unique identier assigned to a given user during the quiz.
We used the R package lme4 [
5
] to perform a generalised linear mixed-eects model (GLMM) anal-
ysis of the relationship between the aforementioned variables and participant conformity. A GLMM
allows us to identify the eect of a set of predictor variables on an outcome variable (conformity)
while following an arbitrary (i.e., possibly non-normal) distribution. We specied participant (User
ID) as a random eect to account for individual dierences in our model.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:12 Senuri Wijenayake et al.
All statistically signicant predictors included in the nal model (following model selection
through incremental removal of variables based on their predictive power) are shown in Table 1.
We perform a likelihood ratio test with the null model [
8
] and nd that our model is statistically
signicant (
𝜒2
= 188.98, p
<
0.001) and explains 41.8% of the variance in accuracy (
R
= 0.65,
R2
= 0.42).
To ensure the validity of the model, we check for the existence of multicollinearity. Our predictors
report a variance ination factor between 1.05 and 1.46, well below the often-used threshold of 5
to detect multicollinearity [25].
Predictor Coecient P-value
Task objectivity (objective) 1.92 < 0.001
Group size dierence 0.69 < 0.001
Initial condence -0.04 < 0.001
No discussion : private -1.15 0.017
No discussion : public -1.13 0.017
Discussion : private -0.97 0.039
Table 1. Eect of statistically significant predictors on participant conformity. The sign of the coeicient (+/-)
denotes the direction of the relationship between the predictor and conformity behaviour. Absolute value
of the coeicient determines the eect size.
We observe statistically signicant main eects of task objectivity (see Fig. 5), group size dierence
(see Fig. 6), and initial condence of participants (see Fig. 7), while the level of interactivity (discussion
vs. no discussion) and response visibility (public vs. private) demonstrate a statistically signicant
interaction eect on conformity behaviour (see Fig. 8). No eect was observed for participant gen-
der or user representations used in this experiment. Next, we present a more detailed look at the
signicant features.
4.2 Task Objectivity, Group Size Dierence and Initial Confidence
Our results show that the task objectivity (objective or subjective nature of the task) had the strongest
eect on conformity behaviour. Out of the 152 conformity responses, 120 responses were related
to objective questions (79%) while only 32 responses were related to subjective questions (21%),
suggesting that participants were more inclined to conform to the majority’s answer in objective
questions than in subjective questions. We illustrate the likelihood of participants conforming to
the majority’s answers in subjective and objective questions in Fig. 5. We note that in subjective
questions, participant conformity ranged between 0 – 50% with a median value of 0%, while in
objective questions the value ranged between 0 – 100% with a higher median value of 50%.
0
10
20
30
40
50
60
70
80
90
100
Objective QuestionsSubjective Questions
Likelihood to conform (%)
Fig. 5. The likelihood of participants conforming to the majority in subjective and objective questions.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:13
Moreover, the group size dierence between the majority and the minority displayed a statistically
signicant main eect on conformity behaviour. We illustrate this relationship in Fig. 6as a density
plot, which shows the distribution of participants’ likelihood of conforming, when the group size
dierence between the majority and the minority was 1 (majority of 3 vs. minority of 2) and 3
(majority of 4 vs. minority of 1). We note that the curve representing conformity behaviour when
the group size dierence is at 1, is left-skewed (with a mean of 24% and a median of 25%), whereas
the curve representing conformity behaviour when the group size dierence is at 3, is right-skewed
(with a mean of 35% and a median of 25%), suggesting that participants are more likely to conform
to the majority when the dierence between the groups is higher.
Fig. 6. The distribution of the likelihood of participants conforming to the majority against the considered
group size dierences. The solid vertical lines denote the mean likelihood to conform in each case.
Furthermore, our results show that individuals who display higher condence on their initial
answers are less likely to be inuenced by the majority as demonstrated by the distribution of initial
condence levels of participants across non-conforming and conforming responses in Fig. 7. While
the initial condence level of participants ranges between 0 – 100 in both non-conforming and
conforming responses, the median initial condence values is at 80 and 55 respectively.
0
10
20
30
40
50
60
70
80
90
100
No Conformity Conformed
Initial Confidence
Fig. 7. Distribution of initial confidence values of participants across non-conforming and conforming responses.
4.3 Social Presence
We manipulated the social presence facilitated by the platform using three variables: user repre-
sentation, interactivity and response visibility. While we do not nd statistically signicant main
eects for each aspect, we note a statistically signicant interaction between the level of interactivity
(discussion vs. no discussion) and response visibility (public vs. private). The interaction eect be-
tween interactivity and response visibility results in four levels of social presence: discussion:public,
discussion:private,no discussion:public, and no discussion:private. The eect of these four levels of
social presence on conformity behaviour is illustrated in Fig. 8.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:14 Senuri Wijenayake et al.
Fig. 8. Proportion of conformity responses across the four levels of interactivity and response visibility.
We highlight that highest conformity is observed when participants are provided the opportunity
to discuss answers with peers, while also displaying their nal answers to the rest of the group before
moving to the next question (discussion:public condition). In contrast, participants are least likely to
conform when there is no peer discussion, and the nal responses are private. Moreover, based on our
results in Table 1, we observe that even with private responses, having peer discussion is more likely
to result in conformity when compared to conditions with no discussion, but with public responses.
4.4 alitative Analysis
Based on the transcripts from our semi-structured interviews, we aim to obtain a richer understanding
of the quantitative results presented in the previous section. Two of the paper’s authors individually
transcribed and categorised the interview data, the outcomes of which were subsequently combined
in an online spreadsheet to aid in the discussion and comparison of the categorisations. The two
authors then collaboratively followed a deductive thematic analysis of the participants’ responses [
9
].
In particular, our semi-structured interview aimed to identify the factors which participants believed
to aect their shift in answer choices and condence. We group our qualitative analysis across the
manipulations concerning online social presence (user representation, interactivity, and response
visibility) and factors previously highlighted in the (oine conformity) literature (task objectivity,
majority group size, and self condence).
4.4.1 Online Social Presence and Conformity. We altered the user representation of our participants
(and their peers) using either generic avatars or user-specic avatars (using initials) and subsequently
asked our participants how they perceived their assigned representation condition in contrast to the
alternative. A number of participants mentioned that the use of initials created a more ‘human-like’
experience; “I like the initials better. It is more human and it kind of acknowledges each of us dierently.
We all have our own uniqueness, when compared to the generic avatars.” (P17). Some participants high-
lighted how they subconsciously viewed user-specic avatars as ‘real’ people facing circumstances
similar to them, which resulted in a connection to these participants; “I think in the discussion, when
I saw avatars with initials, it put a person behind it for me rather than seeing "User 1" or "User 2". To
me that would seem a little more automated than having avatars with initials. I think in a glimpse,
I connected more with others and sometimes felt like I could relate to a particular user.” (P64). Also,
participants described how the use of initials supported in recognition of peers’ (perceived) abilities
to answer correctly; “If you have 20 – 30 questions, then we might see who is giving more correct answers
consistently, with the avatars with initials.” (P20).
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:15
However, a widely shared perception among participants is that the level of representation is
limited in both conditions. As described by P01, an individual’s full name would reveal additional
information as opposed to solely using initials; “If I see the full name and not just the initials, we can
kind of guess the background of that person and where they are from. From there, you can guess what
their experience and opinions might be.”. Participants also describe how additional information not
included in our study may be used to judge peers; “Neither option shows you their age, or experience
and there is no way to know that person’s background.” (P32). In addition to the aforementioned factors
of background and age, one participant suggested that the inclusion of academic titles would sway
their judgement; “But if they have a title like "Professor", I will be more convinced.” (P16).
We also asked our participants on the perceived eect of interactivity with their peers. We describe
three distinct behaviours as observed in the interviews. First, participants note that chat-based
interaction allows them to establish a level of condence in the ability of their peers; “If I get a chance
to talk with others, I will be more condent. I can determine whether they have the knowledge required. For
example, if someone tells me ‘I have been there before’ to the capital of Bulgaria question, I will trust them.
Else I will chose the majority.” (P16). Second, participants note that the discussion allows them to obtain
insights into the reasoning behind their peers’ answers. This was particularly mentioned in relation
to subjective questions. For example, “For most of the questions I was very rm about my answers. [...]
But for subjective questions, I am happy to know why they chose their answers and then maybe change
mine accordingly.” (P01). Third, a small number of participants note that a group of participants could
reach new insights through discussion; “Discussion will help people get new ideas as well.” (P06).
However, we also note that a substantial portion of participants expressed that the discussions
did not change their opinion in subjective questions. They emphasised how they have already made
up their mind, and could only be swayed on objective questions if novel factual information would
be shared; “I don’t think that a discussion would make a change in subjective questions. I feel that I am
entitled to have my opinion and they are entitled to theirs. I justied it pretty well in my head. For objective
questions however, if I got new information from the discussion that would have made a dierence.” (P25).
When considering the eect of response visibility, a large portion of participants from the ‘public’
condition expressed concerns about the perception of their peers; “I don’t want others to nd me foolish.
[...] I knew I was more likely to be wrong and I did not want myself to stand out and feel stupid.” (P12).
Participants which reported to be more at ease with the public visibility of their responses frequently
cited the fact that the study was anonymous – thereby reducing the potential feeling of embarrass-
ment; “I won’t care about what others think whether I change or not. It is still anonymous, I don’t think it
has an eect on me.” (P16). Participants did not only consider whether their peers would label them to
be ‘correct’ or ‘incorrect’, but additionally considered how they were perceived. As illustrated by P46,
“I might be a little pressured if they can see the changes. I would feel like they would think ‘I was not good
enough persuading User 3’ so that adds a little pressure, because I am considering others’ feelings. I don’t
want them to feel bad that they could not persuade me”. Similarly, a participant assigned to the ‘private’
condition noted that even if he would not change his answer, he would use alternative signals to com-
municate with his peers (i.e., answer condence); “If the amended answers were shown to others, I may
reduce my condence. Because others have dierent opinions and secondly, I would think others think ‘he
is very stubborn, he won’t change his answer’. I think this would negatively aect my personality.” (P14).
4.4.2 Traditional Factors and Conformity. Task Objectivity (either objective or subjective in nature)
had a signicant eect on participants’ conformity behaviour, with participants more likely to
conform on objective questions. This sentiment was also observed in our interview data, with the
majority of participants describing their reluctance to switch on subjective questions. The fact
that they could not be ‘wrong’ on subjective questions was a widespread belief among participants;
“Because it is all about my feelings. There is no correct or wrong answer.” (P24). Furthermore, participants
note how the distinction between peer and self-expertise is more critical when considering objective
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:16 Senuri Wijenayake et al.
questions; “If it was about opinions, usually I won’t change because I have my own personal values about
things. If it was more about intelligence, I will listen to others because I am not an expert in this area.” (P16).
A number of participants reported that a more in-depth conversation would have the possibility
to sway their mind on subjective questions, whereas an opposing majority on objective questions
was sucient motivation for switching to the majority opinion; “For subjective questions you need
to read more about it. But for objective questions you know there is a correct answer and someone else
can know it. While I wanted to know more about why others were thinking what they were thinking,
two or three lines are not enough to change my opinion.” (P42).
Group size dierences, the size of the majority in comparison to that of the minority, were reported
as having a considerable impact on participant responses. Participants were more condent in ac-
cepting answers from larger majorities than smaller majorities; “I often followed the majority because
the majority may have the right answer. As we were a group of ve, I was more sure of following the
four person majority than a three person majority.” (P18). If confronted with a unanimous majority
of opposing answers, participants report conforming to the majority despite being condent in their
own answer; “Even when I was pretty sure, there were some times where I changed my answer. It was
because everyone else in the group chose something else.” (P07).
Lastly, participants’ initial condence in their answer showed to be a signicant predictor of their
subsequent conforming behaviour. Our interview data conrms that participants were aware of this
behaviour; “If I had a lower condence in the answer, I was more likely to change” (P10). Participants were
unable to see the condence of their ‘peers’, but this did not deter them from changing their answer;
“Yes - Mainly when it was objective questions and in the case I had no idea of the answer, I often followed
the majority. [...] Even if I did not know their level of condence I still followed the majority.” (P18).
5 DISCUSSION
As people continue to utilise online platforms to pursue social connections and support [
4
,
31
,
48
,
56
,
57
], their experiences are susceptible to social conformity inuences as observed in physical
settings [
6
,
43
,
44
,
56
]. Interestingly, the literature suggests that online conformity has both positive
and negative implications. For instance, conformity is not desired when individuals accept incorrect
group judgements over personal decisions [
6
,
29
]. Biased social information leads to frequent errors,
which could even nullify the perceived benets of collective intelligence. However, existing work also
highlights the importance of social conformity in strengthening online group relationships and creat-
ing a sense of belonging [
56
,
57
]. Moreover, conformity is also seen as a means to encourage acceptable
group norms and standards within online communities, improving the quality of their output [
60
].
Thus, it is imperative to understand what determines conformity behaviour in online settings in order
to understand how future platforms can be designed to benet from social conformity inuences.
Our results show that previously established determinants of oine social conformity such as task
objectivity, group size dierence, and self-condence pose signicant eects on online conformity
as well. We observe higher conformity in objective questions than in subjective questions, which
contradicts the face-to-face literature on this regard [
7
]. Participants explained that conforming to
the supposedly ‘correct’ answers of the majority in objective questions was more appealing than
conforming to the majority’s opinions in subjective questions to be ‘liked’ with a group. This implies
that they are more susceptible to informational inuences (the need to be ‘right’) than normative
inuences (the need to be ‘liked’) in online settings, possibly due to the anonymity and reduced social
presence in online groups [
16
]. Furthermore, participants were more likely to conform as the distance
between the majority and the minority increased, conrming the existing literature on majority
group size in both oine and online contexts [
3
,
20
,
30
,
42
,
54
,
65
,
67
,
68
]. Participants explained that
larger distances between themselves and the majority exerted a sense of isolation and pressure to t in,
suggesting the existence of normative inuences [
16
]. Moreover, we also note that participants were
less pressured to conform when they were more condent about their initial responses, extending
previous ndings of face-to-face conformity literature [11,61] to online settings.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:17
5.1 Impact of Social Presence on Online Social Conformity
Social presence is an important aspect of online interactions as it determines the degree to which we
feel connected with our online correspondents [
58
,
62
]. While recent studies signify the importance
of social presence on online user behaviour [
18
,
26
–
28
,
64
], its eect on conformity is underexplored.
Moreover, despite being recognised as a multifaceted concept, existing work does not investigate
multiple aspects of social presence simultaneously. Thus, in this study we investigate how social
presence impacts online conformity behaviour across dierent user representations (generic vs.
user-specic avatars), levels of interactivity (discussion vs. no discussion), and response visibility
(public vs. private responses).
Our results show an overall conformity rate of 30%, slightly lower than the conformity rates
reported in face-to-face literature (usually above 33% [
2
,
3
]). Dierences in perceived social presence
in face-to-face and online settings may have been a contributing factor in this regard, as highlighted
by our participants during the interviews. In addition, our results imply that online social presence
itself has a clear eect on conformity behaviour. While no main eects were observed for the three
variables of social presence, higher social presence manipulated through interactivity (peer discus-
sions) and response visibility (public responses) resulted in higher conformity. Our qualitative results
conrm that participants in discussion conditions with public responses felt compelled to conform to
the group answers, as their nal answers were visible to peers subsequent to the discussion. Moreover,
they were also concerned with how others would perceive their non-conforming behaviour despite
the group’s eort to convince them, suggesting susceptibility to normative inuences. Thus, in com-
parison to recent work in online conformity highlighting the predominant eects of informational
inuences [
67
,
68
], our study suggests the presence of both normative and informational inuences,
likely due to the added social presence via discussion and public responses.
Furthermore, we note that peer discussion itself without public responses (i.e.,discussion:private)
yielded more conformity than public responses in the absence of discussion (i.e.,no discussion:public),
suggesting that the level of interactivity imposed a higher contribution towards the perceived social
presence than response visibility. Participants reported that discussion (high interactivity) provided
them with an opportunity to understand the reasoning behind peer decisions, potentially increasing
their condence in the peer answers, resulting in increased conformity. Participants also highlighted
how the discussion was most convincing when they could compare and relate their experiences with
peer arguments, demonstrating high levels of social presence [41,62].
We further note that the chosen user representations (generic vs. user-specic avatars) did not
signicantly dier in their inuence on conformity. While, some participants highlighted during
interviews that user-specic avatars were more eective in indicating the presence of a human user in
comparison to generic avatars, the absence of explicit user cues (such as name, gender, age, etc.) may
have invalidated this dierence. Participants further suggested that real photographs or full names
of their peers would have been more inuential alternatives, as they provide more information about
their peers and their background, which could have impacted their nal decision.
5.2 Implications for the Design of Online Group Seings
Social presence is a crucial element of online platform design as it is seen to contribute towards
platform attractiveness [
26
], user involvement and interaction [
18
,
26
], user satisfaction [
24
,
52
], and
trust [
27
,
28
]. Moreover, ndings from our study acknowledge that social presence may also play
a vital role in online conformity behaviour. We note that social presence could manifest in other
means in online settings (e.g., communication style and strategy [
62
]), a topic beyond the scope of
this paper. Thus, we present the following implications based on how online social presence can be
manipulated in terms of interactivity, response visibility, and user representation.
We note that perceived social presence can be controlled via the level of interactivity and response
visibility provided by an online setting, such that enhancing interactivity and response visibility leads
to higher social presence and vice versa. Higher social presence would also increase the susceptibility
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:18 Senuri Wijenayake et al.
of users to social inuences such as conformity. Therefore, when higher conformity is desired (e.g.,
to encourage group norms and standards in online communities [
56
,
60
]), increasing the level of
interactivity and visibility of user input is recommended. This would be particularly eective when
normative conformity is desired (e.g., online support groups [56]).
While increasing perceived social presence in an online setting may seem desirable in most sit-
uations, we cannot disregard its eect on overall conformity behaviour. For example, the eect of
social presence in online learning platforms has been debated in existing literature. While some
studies suggest that social presence (in terms of peer interactions and feedback) leads to higher
student satisfaction in online learning environments [
23
,
24
,
52
], other work shows that students who
interact with peers are more likely to conform to erroneous judgements of their peers [
6
]. Therefore,
platform designers should be aware of this conundrum when determining an appropriate balance
between social presence and social inuence.
Furthermore, our ndings suggest that the visibility of user actions to others in an online setting
is an important determinant of social presence as well as their conformity behaviour. When partic-
ipants were told that their nal answers would be visible to their peers, they were more inclined to
adjust their answers in favour of the majority, as they were more concerned with being ‘liked’ than
being ‘right’. Thus, response visibility may be particularly important when it is desired to enhance
normative conformity. This may be applicable to online support groups where co-dependency and
group togetherness is more important compared to other online settings [56,57].
While our results do not demonstrate a clear dierence in the impact on conformity between the
generic and user-specic avatars, we highlight that participants preferred user representations with
more information and human-like features, as suggested by [
22
,
37
,
47
]. However, we emphasise
that while more information may improve the perceived social connection between users, having
richer user representations (such as photographs or anthropomorphic avatars) may also generate
stereotypical behaviour as seen in prior literature [
35
,
36
,
50
,
67
]. For example, in [
67
] participants
stereotypically perceived competency of others based on gender of their avatars, which in turn
inuenced their conformity behaviour. Thus, we recommend that future work investigate alternative
online user representations devoid of explicit user cues (e.g., name, gender, age), such as default
site-specic avatars as used in Slack and Snapchat, or animal avatars as used in Google Docs as
shown in Fig. 9, to facilitate social presence without triggering similar stereotypical behaviour.
Fig. 9. Avatars used in Slack, Snapchat and Google Docs
In conclusion, our results acknowledge that social presence is a critical factor to be considered
in online platform design. We emphasise that designers should pay special attention to how the level
of interactivity, response visibility, and user representation could be used to control the perceived
online social presence, thus also manipulating how susceptible users are to conformity inuences.
Future work could extend this work to investigate whether and how dierent platform designs could
be developed, manipulating proven determinants of online social conformity (such as social presence,
task objectivity, group size, self-condence etc.) to encourage or discourage conformity and other
social inuences as required, to capitalise on their potential positive and negative implications.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:19
5.3 Limitations
We note the following limitations in our study. While our participants came from dierent educational
backgrounds and levels, they demonstrated above average computer literacy. Thus, further work
may be required to ensure whether these observations can be generalised to a wider population.
While within the community considered for this study (Australia), the chosen topics were accept-
able as general knowledge questions (as supported by prior literature [
67
,
68
], and our quantitative
and qualitative results), they may not generate similar results cross-culturally. Dierent cultural
backgrounds may in fact restrict the applicability of the current ndings. Therefore, we encourage
future research to extend our work by investigating dierent communities following a similar pilot
test (as explained in Section 3.5) to ensure that chosen topics are acceptable as general knowledge
in the targeted community.
Moreover, for the purpose of this experiment we dened social presence in terms of three aspects:
user representation, interactivity, and response visibility. However, we note that social presence is
a broader concept with other dimensions (e.g., social context, online communication [
62
]), which
we did not consider in this study to avoid overly complicating the experimental setup. In addition,
the aspects we did consider can also manifest in other ways. For example, interactivity can be de-
termined via the communication style and formality of language in addition to what was tested in
this study [
62
]. Therefore, we note that our study provides an initial step for future work that could
further investigate other dimensions and aspects of social presence in online environments.
6 CONCLUSION
Recent literature on conformity has given a signicant emphasis to its diverse manifestations in online
settings [
6
,
43
,
44
,
56
,
60
,
67
–
69
]. However, limited eort has been invested in understanding what
triggers conformity in online settings. Therefore, this study investigated the eect of social presence
(i.e., the awareness of and the sense of being connected to others via mediated communication [
58
])
on online conformity, while also accounting for the eects of majority group size, task objectivity,
and self-condence – well-known determinants of oine conformity. We manipulated the social
presence facilitated by our study setup across three variables – user representation, interactivity,
and response visibility – to investigate their combined eect on social conformity.
Our results reveal that in addition to the expected eects from group size dierence, task objectivity,
and self-condence, the interaction among certain aspects of social presence also play a vital role on
online conformity behaviour. We note that higher levels of online social presence (in terms of interac-
tivity and response visibility) heightens the susceptibility of users to normative inuences in addition
to the commonly observed informational inuences, leading to higher conformity in online settings.
We conclude with a discussion on what our ndings imply for the design of online group settings.
First, we emphasise that social presence can indeed be used to control conformity inuences in
online settings. Second, our results suggest that platform designers can manipulate the level of
interactivity facilitated and the visibility of user actions to manage online social presence, which
in turn could help regulate conformity inuences. Thus, designers should pay special attention to
the above aspects of social presence in order to encourage and shape user behaviour as desired
in a given setting. We also highlight that the ‘correct’ level of social presence varies based on the
requirement of the platform and thus should be determined after careful consideration of its pros and
cons. Moreover, our ndings emphasise that online users prefer user representations that can provide
more user-specic information. However, as a substantial amount of literature investigating online
user representations and stereotyping recommends otherwise [
35
,
36
,
67
], we suggest alternatives
that provide a compromise between stronger user cues and stereotypical conformity. We encourage
future work to investigate eects of other dimensions and variables of social presence, such as social
context and communication style [
62
], on online conformity that can expand upon these implications.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:20 Senuri Wijenayake et al.
REFERENCES
[1]
John G Adair. 1972. Demand characteristics or conformity?: Suspiciousness of deception and experimenter bias in confor-
mity research. Canadian Journal of Behavioural Science/Revue canadienne des sciences du comportement 4, 3 (1972), 238.
[2]
Solomon E Asch. 1951. Groups, leadership and men. Carnegie Press, Oxford, England, Chapter: Eects of group pressure
upon the modication and distortion of judgements, 177–190.
[3] Solomon E Asch. 1955. Opinions and social pressure. Scientic American 193, 5 (1955), 31–35.
[4]
Louise Barkhuus and Juliana Tashiro. 2010. Student Socialization in the Age of Facebook. In Proceedings of
the SIGCHI Conference on Human Factors in Computing Systems (CHI ’10). ACM, New York, NY, USA, 133–142.
https://doi.org/10.1145/1753326.1753347
[5]
Douglas Bates, Martin Mächler, Ben Bolker, and Steve Walker. 2015. Fitting Linear Mixed-Eects Models Using lme4.
Journal of Statistical Software 67, 1 (2015), 1–48.
[6]
Tanya Beran, Michelle Drefs, Alyshah Kaba, Noof Al Baz, and Nouf Al Harbi. 2015. Conformity of responses among
graduate students in an online environment. The Internet and Higher Education 25 (2015), 63–69.
[7]
Robert R Blake, Harry Helson, and Jane Srygley Mouton. 1957. The Generality of Conformity Behavior as a Function
of Factual Anchorage. Diculty of Task, and Amount of Social Pressure. Journal of Personality 25, 3 (1957), 294–305.
[8]
Benjamin M. Bolker, Mollie E. Brooks, Connie J. Clark, Shane W. Geange, John R. Poulsen, M. Henry H. Stevens, and
Jada-Simone S. White. 2009. Generalized linear mixed models: a practical guide for ecology and evolution. Trends in
Ecology & Evolution 24, 3 (2009), 127–135.
[9]
Virginia Braun and Victoria Clarke. 2006. Using thematic analysis in psychology. Qualitative Research in Psychology
3, 2 (2006), 77–101. https://doi.org/10.1191/1478088706qp063oa
[10]
Michael J. Brzozowski, Phil Adams, and Ed H. Chi. 2015. Google+ Communities As Plazas and Topic Boards. In
Proceedings of the 33rd Annual ACM Conference on Human Factors in Computing Systems (CHI ’15). ACM, New York,
NY, USA, 3779–3788. https://doi.org/10.1145/2702123.2702600
[11]
Jennifer D Campbell, Abraham Tesser, and Patricia J Fairey. 1986. Conformity and attention to the stimulus: Some
temporal and contextual dynamics. Journal of Personality and Social Psychology 51, 2 (1986), 315–324.
[12]
Emily Christodes, Towhidul Islam, and Serge Desmarais. 2009. Gender stereotyping over instant messenger: The
eects of gender and context. Computers in Human Behavior 25, 4 (2009), 897–901.
[13] Robert B Cialdini. 2001. Harnessing the science of persuasion. Harvard Business Review 79, 9 (2001), 72–81.
[14]
Robert B Cialdini and Noah J Goldstein. 2004. Social inuence: Compliance and conformity. Annu. Rev. Psychol. 55
(2004), 591–621.
[15]
Joanie B Connell, Gerald A Mendelsohn, Richard W Robins, John Canny, and John Canny. 2001. Eects of communication
medium on interpersonal perceptions. In Proceedings of the 2001 International ACM SIGGROUP Conference on Supporting
Group Work. ACM, 117–124.
[16]
Morton Deutsch and Harold B Gerard. 1955. A study of normative and informational social inuences upon individual
judgment. The Journal of Abnormal and Social Psychology 51, 3 (1955), 629–636.
[17]
Leonard W Ferguson. 1944. An analysis of the generality of suggestibility to group opinion. Journal of Personality
12, 3 (1944), 237–243.
[18]
David R Fortin and Ruby Roy Dholakia. 2005. Interactivity and vividness eects on social presence and involvement
with a web-based advertisement. Journal of business research 58, 3 (2005), 387–396.
[19]
D Randy Garrison, Terry Anderson, and Walter Archer. 2001. Critical thinking, cognitive presence, and computer
conferencing in distance education. American Journal of distance education 15, 1 (2001), 7–23.
[20]
Solomon C Goldberg. 1954. Three situational determinants of conformity to social norms. The Journal of Abnormal
and Social Psychology 49, 3 (1954), 325–329.
[21]
Jorge Goncalves, Vassilis Kostakos, and Jayant Venkatanathan. 2013. Narrowcasting in Social Media: Eects and
Perceptions. In Proceedings of the 2013 IEEE/ACM International Conference on Advances in Social Networks Analysis and
Mining (ASONAM ’13). ACM, New York, NY, USA, 502–509.
[22]
Li Gong. 2008. How social is social responses to computers? The function of the degree of anthropomorphism in
computer representations. Computers in Human Behavior 24, 4 (2008), 1494–1509.
[23]
Charlotte N Gunawardena. 1995. Social presence theory and implications for interaction and collaborative learning
in computer conferences. International journal of educational telecommunications 1, 2 (1995), 147–166.
[24]
Charlotte N Gunawardena and Frank J Zittle. 1997. Social presence as a predictor of satisfaction within a
computer-mediated conferencing environment. American journal of distance education 11, 3 (1997), 8–26.
[25]
Joseph F Hair, William C Black, Barry J Babin, Rolph E Anderson, and RL Tatham. 2010. Multivariate Data Analysis.
Pearson, New Jersey, NJ, USA.
[26]
Khaled Hassanein and Milena Head. 2007. Manipulating perceived social presence through the web interface and its
impact on attitude towards online shopping. International Journal of Human-Computer Studies 65, 8 (2007), 689–708.
[27]
Khaled S Hassanein and M Head. 2004. Building online trust through socially rich web interfaces. In Proceedings of
the 2nd annual conference on privacy, security and trust, Fredericton, New Brunswick, Canada. 15–22.
[28]
Traci J Hess, Mark Fuller, and Damon E Campbell. 2009. Designing interfaces with social presence: Using vividness and
extraversion to create social recommendation agents. Journal of the Association for Information Systems 10, 12 (2009), 1.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
antifying the Eect of Social Presence on Online Social Conformity 55:21
[29]
Jessica Hullman, Eytan Adar, and Priti Shah. 2011. The Impact of Social Information on Visual Judgments. In Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’11). ACM, New York, NY, USA, 1461–1470.
https://doi.org/10.1145/1978942.1979157
[30]
Chester A. Insko, Richard H. Smith, Mark D.Alicke, Joel Wade, and Sylvester Taylor. 1985. Conformity and Group Size: The
Concern with Being Right and the Concern with Being Liked. Personality and Social Psychology Bulletin 11, 1 (1985), 41–50.
[31]
Adam N. Joinson. 2008. Looking at, Looking Up or Keeping Up with People?: Motives and Use of Facebook. In Proceedings
of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’08). ACM, New York, NY, USA, 1027–1036.
https://doi.org/10.1145/1357054.1357213
[32]
Jitendra Kumar. 1983. Conformity behavior as a function of confederates’ age and size of the confederate group.
Personality Study & Group Behaviour 3 (1983), 69–73.
[33]
Lieve Laporte, Christof van Nimwegen, and Alex J. Uyttendaele. 2010. Do People Say What They Think: Social
Conformity Behavior in Varying Degrees of Online Social Presence. In Proceedings of the 6th Nordic Conference on
Human-Computer Interaction: Extending Boundaries (NordiCHI ’10). ACM, New York, NY, USA, 305–314.
[34] Bibb Latané. 1981. The psychology of social impact. American psychologist 36, 4 (1981), 343.
[35]
Eun-Ju Lee. 2004. Eects of gendered character representation on person perception and informational social inuence
in computer-mediated communication. Computers in Human Behavior 20, 6 (2004), 779–799.
[36]
Eun-Ju Lee. 2007. Wired for gender: Experientiality and gender-stereotyping in computer-mediated communication.
Media Psychology 10, 2 (2007), 182–210.
[37]
Eun-Ju Lee and Cliord Nass. 2002. Experimental tests of normative group inuence and representation eects in
computer-mediated communication: When interacting via computers diers from interacting with computers. Human
Communication Research 28, 3 (2002), 349–381.
[38]
John M Levine. 1999. Solomon Asch’s legacy for group research. Personality and Social Psychology Review 3, 4 (1999),
358–364.
[39]
Wendy Liu and Derek Ruths. 2013. What’s in a name? Using rst names as features for gender inference in twitter,
In 2013 AAAI Spring Symposium Series. AAAI Spring Symposium - Technical Report SS-13-01, 10–16.
[40]
Yong Liu, Jayant Venkatanathan, Jorge Goncalves, Evangelos Karapanos, and Vassilis Kostakos. 2014. Modeling What
Friendship Patterns on Facebook Reveal About Personality and Social Capital. ACM Trans. Comput.-Hum. Interact.
21, 3, Article Article 17 (June 2014), 20 pages. https://doi.org/10.1145/2617572
[41]
Patrick R Lowenthal and Chareen Snelson. 2017. In search of a better understanding of social presence: an investigation
into how researchers dene social presence. Distance Education 38, 2 (2017), 141–159.
[42] Paul Benjamin Lowry, Tom L Roberts, Nicholas C Romano Jr, Paul D Cheney, and Ross T Hightower. 2006. The impact
of group size and social presence on small-group communication: Does computer-mediated communication make
a dierence? Small Group Research 37, 6 (2006), 631–661.
[43]
Misa Maruyama, Scott P. Robertson, Sara Douglas, Roxanne Raine, and Bryan Semaan. 2017. Social Watching a Civic
Broadcast: Understanding the Eects of Positive Feedback and Other Users’ Opinions. In Proceedings of the 2017 ACM Con-
ference on Computer Supported Cooperative Work and Social Computing (CSCW ’17). ACM, New York, NY, USA, 794–807.
[44]
Misa T. Maruyama, Scott P. Robertson, Sara K. Douglas, Bryan C. Semaan, and Heather A. Faucett. 2014. Hybrid Media
Consumption: How Tweeting During a Televised Political Debate Inuences the Vote Decision. In Proceedings of the
17th ACM Conference on Computer Supported Cooperative Work & Social Computing (CSCW ’14). ACM, New York, NY,
USA, 1422–1432.
[45]
Katelyn YA McKenna and Amie S Green. 2002. Virtual group dynamics. Group Dynamics: Theory, Research, and Practice
6, 1 (2002), 116–127.
[46]
Brian Mullen. 1983. Operationalizing the eect of the group on the individual: A self-attention perspective. Journal
of Experimental Social Psychology 19, 4 (1983), 295–322.
[47]
Kristine L Nowak and Frank Biocca. 2003. The eect of the agency and anthropomorphism on users’ sense of
telepresence, copresence, and social presence in virtual environments. Presence: Teleoperators & Virtual Environments
12, 5 (2003), 481–494.
[48]
Julie Prescott, Terry Hanley, and Katalin Ujhelyi Gomez. 2019. Why do young people use online forums for mental
health and emotional support? Benets and challenges. British Journal of Guidance & Counselling (2019), 1–11.
[49]
Vincent Price, Lilach Nir, and Joseph N Cappella. 2006. Normative and informational inuences in online political
discussions. Communication Theory 16, 1 (2006), 47–74.
[50]
Rabindra Ratan, David Beyea, Benjamin J Li, and Luis Graciano. 2019. Avatar characteristics induce users’ behavioral
conformity with small-to-medium eect sizes: a meta-analysis of the proteus eect. Media Psychology (2019), 1–25.
[51]
Bernardo Reynolds, Jayant Venkatanathan, Jorge Gonçalves, and Vassilis Kostakos.2011. Sharing Ephemeral Information
in Online Social Networks: Privacy Perceptions and Behaviours. In Human-Computer Interaction – INTERACT 2011,
Pedro Campos, Nicholas Graham, Joaquim Jorge, Nuno Nunes, Philippe Palanque, and Marco Winckler (Eds.). Springer
Berlin Heidelberg, Berlin, Heidelberg, 204–215.
[52]
Jennifer Richardson and Karen Swan. 2003. Examing social presence in online courses in relation to students’ perceived
learning and satisfaction. Journal of Asynchronous Learning 7, 1 (2003), 68–88.
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
55:22 Senuri Wijenayake et al.
[53]
Michael Rosander and Oskar Eriksson. 2012. Conformity on the Internet–The role of task diculty and gender
dierences. Computers in Human Behavior 28, 5 (2012), 1587–1595.
[54]
Leon Rosenberg. 1961. Group size, prior experience, and conformity. The Journal of Abnormal and Social Psychology
63, 2 (1961), 436–437.
[55]
Victor Savicki, Merle Kelley, and E Oesterreich. 1999. Judgments of gender in computer-mediated communication.
Computers in Human Behavior 15, 2 (1999), 185–194.
[56]
Eva Sharma and Munmun De Choudhury. 2018. Mental Health Support and Its Relationship to Linguistic Accommodation
in Online Communities. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ’18).
ACM, New York, NY, USA, Article 641, 13 pages.
[57]
Eva Sharma and Munmun De Choudhury. 2018. Mental Health Support and Its Relationship to Linguistic Accommodation
in Online Communities. In Proceedings of the 2018 CHI Conference on Human Factors in Computing Systems (CHI ’18).
ACM, New York, NY, USA, Article 641, 13 pages. https://doi.org/10.1145/3173574.3174215
[58]
John Short, Ederyn Williams, and Bruce Christie. 1976. The social psychology of telecommunications. John Wiley & Sons.
[59]
David J Stang. 1976. Ineective deception in conformity research: Some causes and consequences. European Journal
of Social Psychology 6, 3 (1976), 353–367.
[60] Abhay Sukumaran, Stephanie Vezich, Melanie McHugh, and Cliord Nass. 2011. Normative Inuences on Thoughtful
Online Participation. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’11). ACM,
New York, NY, USA, 3401–3410.
[61]
Abraham Tesser, Jennifer Campbell, and Susan Mickler. 1983. The role of social pressure, attention to the stimulus,
and self-doubt in conformity. European Journal of Social Psychology 13, 3 (1983), 217–233.
[62]
Chih-Hsiung Tu and Marina McIsaac. 2002. The relationship of social presence and interaction in online classes. The
American journal of distance education 16, 3 (2002), 131–150.
[63]
Jayant Venkatanathan, Evangelos Karapanos, Vassilis Kostakos, and Jorge Gonçalves. 2012. Network, Personality and
Social Capital. In Proceedings of the 4th Annual ACM Web Science Conference (WebSci ’12). Association for Computing
Machinery, New York, NY, USA, 326–329. https://doi.org/10.1145/2380718.2380760
[64]
Jayant Venkatanathan, Vassilis Kostakos, Evangelos Karapanos, and Jorge Gonçalves. 2013. Online Disclosure
of Personally Identiable Information with Strangers: Eects of Public and Private Sharing. Interacting with
Computers 26, 6 (11 2013), 614–626. https://doi.org/10.1093/iwc/iwt058 arXiv:https://academic.oup.com/iwc/article-
pdf/26/6/614/1962602/iwt058.pdf
[65]
Eva Walther, Herbert Bless, Fritz Strack, Patsy Rackstraw, Doris Wagner, and Lioba Werth. 2002. Conformity eects
in memory as a function of group size, dissenters and uncertainty. Applied Cognitive Psychology: The Ocial Journal
of the Society for Applied Research in Memory and Cognition 16, 7 (2002), 793–810.
[66]
Senuri Wijenayake, Niels van Berkel, and Jorge Goncalves. 2020. Bots for Research: Minimising the Experimenter
Eect. In International Workshop on Detection and Design for Cognitive Biases in People and Computing Systems (CHI’20
Workshop). ACM, 9 pages.
[67]
Senuri Wijenayake, Niels van Berkel, Vassilis Kostakos, and Jorge Goncalves. 2019. Measuring the Eects of
Gender on Online Social Conformity. Proc. ACM Hum.-Comput. Interact. 3, CSCW, Article 145 (Nov. 2019), 24 pages.
https://doi.org/10.1145/3359247
[68]
Senuri Wijenayake, Niels van Berkel, Vassilis Kostakos, and Jorge Goncalves. 2020. Impact of contextual and personal
determinants on online social conformity. Computers in Human Behavior (2020), 106302.
[69]
Haiyi Zhu, Bernardo Huberman, and Yarun Luon. 2012. To Switch or Not to Switch: Understanding Social Inuence
in Online Choices. In Proceedings of the SIGCHI Conference on Human Factors in Computing Systems (CHI ’12). ACM,
New York, NY, USA, 2257–2266. https://doi.org/10.1145/2207676.2208383
Received January 2020; accepted March 2020
Proc. ACM Hum.-Comput. Interact., Vol. 4, No. CSCW1, Article 55. Publication date: May 2020.
