Exploring Selective Exposure and Selective Avoidance Behavior in Social Media

Abstract

This study investigates social media users' preferences of encountering or actively avoiding undesired content and conflicts in social interaction with others. Based on a nationwide survey (N=3706) conducted in Finland and using principal component analysis, we identify three different types of social media use in relation to online information sharing and social interaction: conformist, provocative and protective. We then modelled those variations according to demographic variables and subjective life satisfaction. We found that women are more likely to use social media in a conformist and protective way whereas men have a higher probability to be provocative. We also found that younger and more educated people have a higher probability to use social media in a conformist and protective way. Finally, we suggest that subjective life satisfaction more powerfully predicts provocative use compared to age or education.

Full text

Exploring Selective Exposure and Selective Avoidance
Behavior in Social Media
Sanna Malinen
University of Turku
Department of Social Research
Turku, Finland
sanna.malinen@utu.fi
Aki Koivula
University of Turku
Department of Social Research
Turku, Finland
aki.j.koivula@utu.fi
Teo Keipi
University of Turku
Department of Social Research
Turku, Finland
teo.a.keipi@utu.fi
Ilkka Koiranen
University of Turku
Department of Social Research
Turku, Finland
ilkka.a.koiranen@utu.fi
ABSTRACT
This study investigates social media users’ preferences of
encountering or actively avoiding undesired content and
conflicts in social interaction with others. Based on a
nationwide survey (N=3706) conducted in Finland and using
principal component analysis, we identify three different
types of social media use in relation to online information
sharing and social interaction: conformist, provocative and
protective. We then modelled those variations according to
demographic variables and subjective life satisfaction. We
found that women are more likely to use social media in a
conformist and protective way whereas men have a higher
probability to be provocative. We also found that younger
and more educated people have a higher probability to use
social media in a conformist and protective way. Finally, we
suggest that subjective life satisfaction more powerfully
predicts provocative use compared to age or education.
CCS CONCEPTS
• Human-centered computing → Human computer
interaction (HCI); User studies, → Collaborative and
social computing; Social media
KEYWORDS
selective exposure, selective avoidance, social media, social
networking
1 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 profit or commercial advantage and that copies
bear this notice and the full citation on the first page. Copyrights for
components of this work owned by others than ACM must be honored.
Abstracting with credit is permitted. To copy otherwise, or republish, to post
on servers or to redistribute to lists, requires prior specific permission and/or
a fee. Request permissions from Permissions@acm.org.
SMSociety '18, July 18–20, 2018, Copenhagen, Denmark
© 2018 Association for Computing Machinery.
ACM ISBN 978-1-4503-6334-1/18/07…$15.00
https://doi.org/10.1145/3217804.3217943
ACM Reference format:
Sanna Malinen, Aki Koivula, Teo Keipi, and Ilkka Koiranen. 2018.
Exploring Selective Exposure and Selective Avoidance Behavior
in Social Media. In Proceedings of the International Conference on
Social Media & Society, Copenhagen, Denmark (SMSociety). 1
DOI: 10.1145/3217804.3217943
1 INTRODUCTION
The connection between an individual’s psychological well-
being and social media use has received plenty of scholarly
attention. There is evidence that online social networking
can increase people’s social capital and improve their well-
being [1]. There is also a positive relationship between an
individual’s life satisfaction and intensity of Facebook use,
which has been explained by users’ engagement in behaviors
that contribute to their social capital [2]. Particularly, the
quality of interaction on social media has been found to
matter for psychological well-being. Social media provides
users with many supportive elements, which are important
for experienced life satisfaction [3].
Recently, there has been discussion on how social media
can reinforce people’s existing beliefs and biases. By
providing people with information they prefer and similarly
preventing them from exposure to contradicting views,
social media is suspected to facilitate the emergence of
350

SMSociety, July 2018, Copenhagen, Denmark
Malinen, Koivula, Keipi, & Koiranen
groups with high agreement and non-tolerance of
challenging views, often referred to as “echo chambers” [4].
The terms “selective exposure” and “selective avoidance”
are used to describe the behavior in which a person actively
seeks for information that supports their views and avoids
information that challenges them [5]. Consistently,
experimental studies have shown that selective exposure to
attitude-consistent messages strengthens related attitudes
and selective exposure to attitude-discrepant messages
weakens related attitudes [6].
In social media, selective avoidance can be easily
performed by removing or hiding unwanted content or
persons. Social media users can add individuals from
different social contexts to their friends list. This
characteristic, which brings people from diverse contexts
together in a single location, is referred to as “context
collapse” [7]. Context collapse is likely to create tensions
when someone attempts to maintain a consistent
presentation of self for these fragmented social media
audiences [7]. However, also increased and repeated
exposure to dissonant information and perspectives can
motivate selective avoidance and use of boundary regulation
tools, such as hiding and unfriending, to control the exposure
to unwanted content and connections that transmit this
content [8, 9]. The exposure to unpleasant or inappropriate
content and attempts to manage it have been named as one
of the main stressors in social media interaction [10].
Previous research indicates that there are substantial differences
in access to online media, use purposes, skills and benefits gained
from its use [11, 12]. In Finland, context of the present study, recent
statistics show that highly educated and wealthier users are utilizing
social media more actively [13]. Gender has also become a
prominent factor in Finns’ social media use during this decade, as
women are more generally logged on to social media sites [14].
Although the majority of research has focused on younger
demographics, some studies on the variety of social media use in
different age groups have surfaced. Older people have been found
to be more conventional and restricted in their social media
participation, while younger adults and especially teenagers have a
much more extensive selection of different behavior models and
roles when using Facebook [15]. There are also age differences in
how users experience privacy on Facebook. For teenagers and
younger adults, having multiple audiences in the same place
disrupts the content sharing process and causes experiences of
social surveillance and social control [16]. As a reaction to this,
they use conformity as a strategy and avoid sharing anything too
private and personal [16]. According to the same study, older adults
over age 40 were less aware of their privacy settings on Facebook,
and overall, they found the privacy tools too difficult to use.
There is still not much work investigating the variety of
people’s preferences regarding social media exposure.
Munson and Resnick [17] found that Internet users vary
greatly in their attitudes regarding diversity and conformity
of information, as some of them prefer a greater spectrum of
views when reading political content more than others. They
argue that none of these behaviors are a fundamental trait of
human behavior that describes all people but instead, they
describe different preferences of different groups of people,
and should be better considered when designing websites
and aggregating content. In this study, we focus on social
media users’ social action and choices in the case of
unwanted content. Our goal is to provide a new frame to
understand how people are dealing with unwanted content
and information. We form a new typology for social
networking site users by means of selective exposure and
selective avoidance. We also assess how sociodemographic
factors and life satisfaction affect various behavior models
on these platforms. The majority of the research
investigating selective exposure and selective avoidance on
social media has been focusing on single platforms, such as
Twitter or Facebook. In this work, we approach social media
more extensively, covering discussion forums, social
networking sites and online news sites with comment
sections.
This work-in-progress paper is based on empirical data
collected via population-wide survey in Finland between
December 2017 and January 2018. With this data we will
answer the following research questions:
1. Is there variety in the respondents’ behavior in
confronting undesired information and social
interactions?
2. To what extent is online behavior associated with
the demographic background of respondents?
3. To what extent are online behavior and an
individual’s overall life satisfaction associated?
2 DATA AND METHODS
Our analyses are based on a survey, which was collected from
two different sources. The first part was distributed by mail to a
simple random sample of 8000 15–74-year-olds who live in
Finland and speak Finnish. A total of 2452 Finns responded to this
collection, which amounted to a 31 percent response rate as those
who could not be reached were omitted from the sample. Secondly,
we improved the data by collecting a sample of 1200 respondents
aged 18-74 from an online panel of volunteer respondents
administrated by a market research company. The final data
included a total of 3706 respondents of which 66% are based on
probability sampling and 34% are based on nonprobability
sampling.
The survey included questions of the participants’ basic
demographics, such as gender, education and age. The data
represent both genders well as 50% of the participants were male
and 50% female. The final sample is also relatively representative
in terms of education, as 51% of the sample has secondary level
education and 34% holds master or bachelor degree. Respondent
age ranged from 18 to 74 years, mean being 51 years, which makes
the age distribution of the data slightly skewed towards the older
age groups as the population mean is 46 with respect to applied age
range.
Other questions focused on their media use and attitudes. For
instance, we asked which traditional and online media they used,
351

Exploring Selective Exposure and Selective Avoidance SMSociety, July 2018, Copenhagen, Denmark
the frequency of use and the reasons for use. Usually, older people
might be expected to use the Internet less frequently and with less
variety than younger ones. In general, the Internet and social media
are commonly used in Finland among all age groups, including
older people and unlike in younger age groups, social media use is
expected to grow among people over age 44 [18]. According to the
recent report by Official Statistics of Finland [13], 43% of the age
group of 55-64 years and 25% of the age group of 65-74 years
reported using social networking sites. In this respect, our sample
is bit overrepresented with social media users especially in terms of
older users, as 52% of the age group of 55-64 years and 42% of the
age group of 65-74 years reported using the social networking sites.
We begin our analysis by utilizing principal component analysis
(PCA) for different kind of behavioral variables addressed to social
media use. The main target of PCA is to extract visible features of
how each variables are associated with one another. According to
the PCA solution, we establish dependent variables for
multivariable analysis.
We conduct multivariable analyses separately for different
dependent variables by using ordinary least squared (OLS) models.
The aim of the explanatory analysis is to find the main predictors
for different kind of social media use. In this respect, we test the
extent to which independent variables, namely gender, age,
education and life satisfaction are associated with dependent
variables. We measure the subjects’ experienced life satisfaction
with one question: “How would you rate on a scale from 0 ‘very
unsatisfied’ to 10 ‘very satisfied’ your satisfaction of your life?”
We also control respondents’ social media use frequency.
3 RESULTS
We wanted to find out if there are individual differences in
respondents’ social media behavior, more specifically, in their
willingness to encounter dissonant views, undesirable content and
conflicts, and if demographic factors explain these differences.
Drawing on existing literature [15, 17], we expected that different
user types could be identified in relation to predictability or
diversity of content, and we created statements that would distinct
these individual differences. Principal component analysis was
conducted in order to reveal users’ different behavioral patterns. In
a questionnaire, applied items were presented to respondents as a
set of statements to the main question of “What do you think of the
following statements”. Respondents were asked to choose their
opinion from a five point Likert-scale in which they were given
options such as 1 – “Completely disagree”, 3 – “Do not agree or
disagree”, and 5 – “Completely agree”. A total of nine statements
were presented and they were all employed in PCA. As a result of
PCA, we found three main components measuring respondents’
online behavior from different approaches. These components are
formed on the basis of eight different items as the one item was
excluded from the final solution because of the high uniqueness and
low intercorrelation with any component. The final solution is
presented in Table 1.
The first component, Conformist use, includes items about fear
of hurting others’ feelings, avoidance of conflict, giving a good
impression online and supporting others. The component two,
Provocative use, consists of items about deliberately provoking
others on social media by disagreeing with others and sharing
content that is expected to annoy others. The third component is
named as Protective use, because it describes the aim to protect
oneself from harmful or offensive online content using selective
avoidance strategies. It includes items about hiding undesirable
content and removing or hiding annoying persons from social
networks.
On the basis of the PCA solution, we generated three mean
variables. The descriptive statistics for mean variables is shown in
Table 1. Each of variables are continuous-types and suitable for
parametric tests as interval variables [19]. Next, we run OLS
models to find the main predictors of generated variables. Only
those respondents who used social media at all and had valid scores
on all three dependent variables were included in the analysis.
Table 1: Three Main Components and Their Loadings
Survey questions
1
2
3
The fear of offending others limits
my posting of my opinions on
social media
.57
I try to give others on social media
an improved image of who I am
.38
I very often “like” other users’
posts in order to show support and
empathy
.38
I purposefully share material on
social media that I believe will
provoke others
.60
I comment on others’ posts on
social media even when I disagree
with them
.62
I share content on social media that
I feel could lead to disputes
.33
I have hidden content that conflicts
with my points of view on social
media
.70
I have hidden or removed annoying
or bothersome users on social
media
.57
Descriptive statistics for mean variables,
Means (Std.dev)
Component 1: Conformist use
2.5
(0.9)
Component 2: Provocative use
2.2
(0.8)
Component 3: Protective use
2.3
(1.2)
Table 2 displays the effects of demographic variables and life
satisfaction on the dependent variables. As seen in the first column
that presents effects on the conformist use, gender was a crucial
predictor as females differ significantly from males. Also there can
be seen a strong effect of age, as the younger people tend to use
352

SMSociety, July 2018, Copenhagen, Denmark
Malinen, Koivula, Keipi, & Koiranen
social media in a more conformist way than older. However,
interestingly, this effect can be seen solely when examining the
three oldest age groups (45-55, 55-64 and 65-74). When it comes
to differences between educational levels, we found that the
respondents with master degree had the highest scores in the
conformist variable. Those with higher life satisfaction tend to be
less conformist.
Table 2: Predicting Three Types of Social Media Behavior
(1)
(2)
(3)
VARIABLES
Conformist
use
Provocative
use
Protective
use
Female
0.155***
-0.323***
0.182***
(0.031)
(0.029)
(0.043)
Age (omitted under 30 years)
31-44
-0.069
0.068
-0.050
(0.049)
(0.045)
(0.068)
45-54
-0.175***
0.076
-0.337***
(0.052)
(0.048)
(0.072)
55-64
-0.304***
-0.022
-0.515***
(0.051)
(0.047)
(0.070)
65 or older
-0.294***
-0.014
-0.610***
(0.053)
(0.049)
(0.073)
Education (omitted primary/secondary)
Tertiary
0.060
-0.011
0.133**
(0.036)
(0.033)
(0.050)
Master
0.170***
0.005
0.192***
(0.040)
(0.037)
(0.056)
Life satisfaction
-0.034***
-0.040***
-0.033**
(0.008)
(0.008)
(0.011)
Constant
1.885***
2.398***
1.790***
(0.096)
(0.088)
(0.133)
Observations
2,716
2,716
2,716
R-squared
0.247
0.074
0.197
Standard errors in parentheses
Models control for social media use frequency
*** p<0.001, ** p<0.01, * p<0.05
When moving to the next column in order to evaluate predictors
of provocative use, we can see that the direction of gender effect
turn around as men reported higher scores than women.
Interestingly, we did not find age or education effects in terms of
provocative use. Instead, the effect of life satisfaction could be seen
to be extremely strong. Those with higher life satisfaction reported
lower scores for provocative use. Finally, we turn to analyze
protective use. Here, we also found a significant difference between
genders, as women reported higher scores. As is the case with
conformist use, protective use is also dependent on users’ age.
Younger users seem to be more protective than the older users.
Again, this is especially the case of the three oldest groups who
reported lower scores than younger. In terms of education, we
found that highly educated users had higher scores. The final
component underlines the effect of life satisfaction, which was
similar in all three behavior models.
4 CONCLUSIONS
In this work-in-progress paper, we have presented the initial
findings from the nationwide survey exploring people’s social
media behavior, our special focus being on selective exposure and
selective avoidance. Using principal component analysis we
identified three types of social media use, which implicate
differences in users’ tolerance of conflicts and exposure to
unwanted content. We generated three mean variables on the basis
of component solution for further analysis.
According to the results of OLR analysis, there are significant
structural differences in how different population groups encounter
dissimilar opinions and conflicts. In terms of gender, women are
more likely to protect themselves or act conservatively on social
media. Men, on the other hand, are more likely to act provocatively.
Also, the highly educated respondents are more likely to protect
themselves from dissonant content or act more conciliatory when
exposed to such subject matter. Our findings also show the effect
of age in two behaviors, conformist and protective, as oldest
respondents were less conformist and protective online. This
finding is similar to previous work [16], which indicates that
particularly young people experience social control online, and
therefore, they tend to be more restricted in what they share with
others. This may also indicate that older people are less aware of or
less concerned about “netiquette” and their own privacy. However,
in relation to provocative social media behavior, no age differences
were found.
In terms of life satisfaction, our findings are not completely
uniform with previous research. Surprisingly, those who are
conformist, i.e., engage in supportive behavior and avoid offending
others, did not score highest in life satisfaction. On the other hand,
the effect of life satisfaction was extremely strong in provocative
social media use: those who tend to engage in provocative behavior
were the least satisfied with their lives. This confirms the
assumptions that anti-social online behavior such as trolling and
deliberately offending others reflects an individual’s lower
psychological well-being. These findings also suggest that people
with higher life satisfaction do not need to resort to any of these
different strategies, while those who are less satisfied with their
lives are more dependent on these strategies when they are
confronting undesired content.
Scholars have argued that one of the most harmful
consequences of social media is exposure to antagonist material
[20-22]. Taking this into account, it is not surprising that people are
actively protecting themselves from such content. Especially
women and younger age groups engaged in selective avoidance,
which may indicate that those groups were more aware of harmful
online content than others, or they did not prefer seeing conflicting
views in their social media newsfeeds.
When considering the formation of echo chambers in social
media environment, one of the most crucial factors is people’s
tendency to actively keep one’s social media content preference-
consistent. This can mean that different population groups’ values
and views are becoming more separated. In this sense, those who
try to protect themselves are placing their individual and personal
preferences before the benefit of society as a whole. Given that
353

Exploring Selective Exposure and Selective Avoidance SMSociety, July 2018, Copenhagen, Denmark
isolation can encourage polarization in terms of norms, behaviors
and attitudes, the formation of echo chambers poses a social risk.
Thus, the balance between the desire to protect oneself and the cost
of isolating oneself from opinions or people with whom one
disagrees is an important one; where users seek to minimize
challenging viewpoints that might otherwise widen a worldview
beneficially, for example, a social loss is experienced in the form
of lost opportunities for valuable dialogue. Harmful biases,
prejudices and beliefs in inaccurate information represent the
harmful side of protecting oneself against discomfort. On the other
hand, protecting oneself from harassment and intrusive content or
users is an important aspect of online navigation, one that should
not be eliminated.
ACKNOWLEDGMENTS
We thank the research group of Economic Sociology at
University of Turku for making this survey possible. The research
for this article was funded by the HS Foundation Grant (decision
date 30.3.2017) and the Strategic Research Council of the Academy
of Finland (decision number 314250).
REFERENCES
[1]
Nicole B. Ellison, Charles Steinfield, and Cliff Lampe. 2007. The
benefits of Facebook “friends”: Social capital and college students’
use of online social network sites. Journal of Computer Mediated
Communication 12(4), 1143–1168.
[2]
Sebastián Valenzuela, Namsu Park, Kerk F. Kee; Is There Social
Capital in a Social Network Site?: Facebook Use and College
Students' Life Satisfaction, Trust, and Participation. Journal of
Computer-Mediated Communication 14(4), 875–901.
[3]
Hyun Jung Oh, Elif Ozkaya, and Robert LaRose. 2014. How does
online social networking enhance life satisfaction? The
relationships among online supportive interaction, affect,
perceived social support, sense of community, and life
satisfaction. Computers in Human Behavior 30, 69-78.
[4]
Cass Sunstein. 2001. Echo chambers: Bush v. Gore, impeachment,
and beyond. Princeton University Press, Princeton, NJ.
[5]
Leon Festinger. 1957. A theory of cognitive dissonance. Stanford
University Press, Stanford, CA.
[6]
Silvia Knobloch-Westerwick, Cornelia Mothes, Benjamin K.
Johnson, Axel Westerwick, and Wolfgang Donsbach. 2015.
Political Online Information Searching in Germany and the United
States: Confirmation Bias, Source Credibility, and Attitude
Impacts. Journal of Communication 65 (2015), 489–511.
[7]
Alice E. Marwick and danah boyd. 2011. I tweet honestly, I tweet
passionately: Twitter users, context collapse, and the imagined
audience. New media & society 13(1), 114-133.
[8]
Qinfeng Zhu, Marko Skoric, and Fei Shen. 2017. I Shield Myself
From Thee: Selective Avoidance on Social Media During Political
Protests. Political Communication 34(1), 112-131.
[9]
Sanna Malinen. 2015. ‘Unsociability’ as Boundary Regulation on
Social Network Sites. In Proceedings of 23rd European
Conference on Information Systems (ECIS 2015).
[10]
Jesse Fox and Jennifer J. Moreland. 2015. The Dark Side of Social
Networking Sites: An Exploration of the Relational and
Psychological Stressors Associated with Facebook use and
Affordances. Computers in Human Behavior 45(4), 168-176.
[11]
Alexander J. Van Deursen, and Jan A Van Dijk. 2014. The digital
divide shifts to differences in usage. New media & society 16(3),
507-526.
[12]
Jan Schradie. 2011. The digital production gap: The digital divide
and Web 2.0 collide. Poetic 39(2), 145-168.
[13]
Official Statistics of Finland (OSF). 2017. Use of information and
communications technology by individuals [e-publication].
ISSN=2341-8710. Helsinki: Statistics Finland [referred:
29.1.2018].
Access method: http://www.stat.fi/til/sutivi/tup_en.html
[14]
Ilkka Koiranen, Teo Keipi, Teo, Aki Koivula, and Pekka Räsänen.
2017. The different uses of social media – A population-level study
in Finland. Working Papers in Economic Sociology (IX).
University of Turku, Turku.
[15]
Niki Panteli and Ben Marder. 2017. Constructing and enacting
normality online across generations: The case of social networking
sites. Information Technology & People 30(2), 282-300.
[16]
Petter Bae Brandtzæg, Marika Lüders, and Jan Håvard Skjetne.
2010. Too many Facebook “friends”? Content sharing and
sociability versus the need for privacy in social network sites.
International Journal of Human–Computer Interaction 26(11-12),
1006-1030.
[17]
Sean A. Munson and Paul Resnick. 2010. Presenting diverse
political opinions: how and how much. In Proceedings of the
SIGCHI Conference on Human Factors in Computing
Systems (CHI '10). ACM, New York, NY, USA, 1457-1466.
[18]
Aki Koivula, Teo Keipi, Ilkka Koiranen and Pekka Räsänen. 2018
A middle-aged social Internet with a millennial exodus? Changes
in identifications with online communities between 2009 and 2017
in Finland. In Proceedings of International Conference on Social
Computing and Social Media. Springer, Cham.
[19]
James Carifio and Rocco J. Perla. 2007. Ten common
misunderstandings, misconceptions, persistent myths and urban
legends about Likert scales and Likert response formats and their
antidotes. Journal of Social Sciences 3(3), 106-116.
[20]
Teo Keipi, Matti Näsi, Atte Oksanen, and Pekka Räsänen. 2016.
Online hate and harmful content: Cross-national perspectives Vol.
200. Taylor & Francis.
[21]
Kristen Harrison and Veronika Hefner. 2008. Media, body image,
and eating disorders, in Calvert, S. and Wilson, B. (Eds), The
Handbook of Children, Media, and Development. Blackwell,
Malden, MA, 381-406.
[22]
Jane D. Brown and Piotr S. Bobkowski. 2011. Older and newer
media: Patterns of use and effects on adolescents' health and well‐
being. Journal of Research on Adolescence 21(1), 95-113.
354