Journal of Computer-Mediated Communication

FULL-LENGTH RESEARCH ARTICLE

Humor Reduces Online Incivility

Yomna Elsayed1 & Andrea B. Hollingshead2

1
Department of Communication, University of Utah, Salt Lake City, UT 84112, USA
2
Annenberg School for Communication and Journalism, University of Southern California, Los Angeles, CA 90089, USA

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Online incivility is a persistent issue facing many news and social media platforms. To better un-
derstand it, we examined whether humorous content reduced online incivility and whether the
(in)civility of another user might mitigate or amplify this effect in two experiments (Study 1,
N ¼ 122; Study 2, N ¼ 208). Participants in both experiments read an online opinion article
about an instance of negative stereotyping and provided a comment on a simulated online news
forum. The first study manipulated article humor (humorous vs. not humorous). The second
study manipulated humor and the (in)civility of a previous user’s comment (civil vs. uncivil). In
both studies, humor-reduced incivility. Anger mediated this effect. Source liking mediated the
positive effect of humor on reducing anger. The (in)civility of a previous comment did not affect
these results. Overall, the findings point to the value of humor and anger reduction in managing
online incivility.

Lay Summary
Online incivility is a common occurrence in online public forums. This article investigated factors
that may reduce incivility in two online experiments. The first experiment tested whether humor
reduced anger and subsequently the incivility of participants’ comments towards a negative
stereotype-challenging op-ed article. The second experiment tested whether exposure to civil and
uncivil previous user comments influenced the civility of participants’ comments. Humor was
found to reduce online incivility by reducing feelings of anger and increasing liking towards the
author in both experiments. Exposure to another user’s comment, whether civil or uncivil, did
not appear to influence the civility of participants’ comments, suggesting that article content is a
more likely trigger for online incivility.

Keywords: Online Comments, Incivility, Toxicity, Humor, Anger, Source Liking, Machine-learning

https://doi.org/10.1093/jcmc/zmac005

Corresponding author: Yomna Elsayed; e-mail: [email protected]

Associate Editor: Nicole Kraemer; Received: 4 May 2021; Revisions received: 18 December 2021; Accepted: 26 January
2022.

Journal of Computer-Mediated Communication 00 (2022) 1–19 V C The Author(s) 2022. Published by Oxford University Press 1
on behalf of International Communication Association.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/
by/4.0/), which permits unrestricted reuse, distribution, and reproduction in any medium, provided the original work is properly cited.
Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

Online incivility can hinder democratic civil discourse (Papacharissi, 2004) by discouraging individu-
als from participation (Chua, 2009) and from seeking more information (Minich et al., 2018).
Incivility is also referred to as toxicity,1 which is “anything that is rude, disrespectful, or unreasonable
that would make someone want to leave a conversation” (Borkan et al., 2019). Online incivility is dis-
proportionately directed at minority groups (Santana, 2012; Rheault et al., 2019) and can be harmful
to a multicultural society whose democracy is premised on diverse perspectives and civic engagement.
Researchers have also pushed for a more nuanced treatment of incivility as a concept that can be
culturally specific and sometimes beneficial in drawing attention to the plight of marginalized groups

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

(Masullo et al., 2019). This alternative view presents a conundrum where on one hand marginalized
groups are one of the main targets of incivility, and at the same time, incivility could be one of their
very few tools for speaking truth to power. One way of striking this balance is to distinguish between
incivility as “profanity or crude language” and intolerance which “targets categories of people for dis-
crimination, hate, abuse, etc.” (Tromble, 2018).
While journalistic moderation has been suggested as a way for controlling incivility, issues of
scale have driven some news sites to discontinue the practice of allowing users to comment altogether
(Goujard, 2016). Social media companies in particular are constantly looking for ways to make their
platforms less toxic and more encouraging for conversations (Aten, 2020; Tromble, 2018), such as
nudging people before posting toxic content. For example, Twitter found that 34% of people revised
their initial reply or chose not to send the reply at all after seeing a prompt (Perez, 2021). These devel-
opments have led researchers to investigate factors that predict online incivility as a way to manage it
and possibly prevent uncivil remarks from turning into intolerant or hateful speech.

Predictors of Online Incivility

Understanding antecedents is a key step towards managing online incivility. Consequently, predictors
of incivility have been studied in political communication and digital journalism. Researchers have in-
vestigated two general classes of predictors: Those that relate to article content and those that relate to
audience and conversation characteristics. We investigated examples of both predictor classes in the
present research.
Content features include news article topics, authors and quoted sources that may elicit online in-
civility. Content analysis of articles and comments on a local news website revealed that serious or
“hard news” (articles about the economy, politics, immigration, or taxes) were associated with in-
creased incivility in user comments as opposed to “soft news” (articles about lifestyle, journalism, or
technology) (Coe et al., 2014; Brückner & Schweiger, 2017). Additionally, comments on articles with
contentious topics were significantly less civil than comments on more balanced or neutral articles
(Brückner & Schweiger, 2017). Article authors’ status (local vs. national and news vs. opinion), how-
ever, did not seem to have a significant effect on incivility, while sources quoted in the article espe-
cially ones with identifiable partisan leaning (e.g. President Obama) seemed to generate the most
incivility (Coe et al., 2014).
Audience characteristics which include personal attributes of commenters, such as frequency of
commenting and political identity, and conversational patterns, such as discussion volume and cyber-
balkanization, have also been investigated. Frequent commenters, for example, exhibit less incivility
than infrequent commenters (Coe, Kenski & Rains, 2014). Content analysis of an online edition of a
daily newspaper showed that the political identity of audience members affects the discussion in un-
expected ways (Rains et al, 2017). Conservatives, for example, were less likely to be uncivil in the

2 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

presence of other conservatives while non-conservatives were more likely to be uncivil as the presence
of conservatives increased. Conversational patterns such as discussion volume defined by the total
number of posts and comments, and the cyberbalkanization of Facebook pages characterized by in-
creased sharing of pages from within the same community as opposed to other communities, were
positively associated with comment incivility (Lee et al., 2019).
Because incivility is an emotional experience, it is also important to examine the social and emo-
tional cues that surround it (Itzkovich et al., 2020). Young (2019), for example, examined the role of
affect in politics drawing attention to how psychological traits, such as the need for humor, need for

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

cognition and tolerance for ambiguity, as opposed to political beliefs or worldviews, can predetermine
a person’s ideological affinity. She noted how the difference in these traits accounts for liberals’ prefer-
ence for the genre of irony and conservatives’ preference for the genre of outrage in political enter-
tainment. Young’s work, however, is more concerned with trait emotions than state emotions, where
the former are seen as habitual tendencies whereas the latter are emotions experienced in a specific
situation (Bieg, 2013).
The present research on the other hand is more concerned with state emotions experienced in a
situation of online incivility. It examines how humor, specifically self-deprecating humor, may work
to reduce anger and subsequently incivility in these situations. Members of marginalized groups, who
are often targets of incivility, have historically resorted to self-deprecating humor as a means for pre-
emptive self-defense (Juni & Katz, 2001). Incivility exists on a spectrum (Chen, 2017), and while hu-
mor may not be a solution to the far end of the spectrum represented by intolerant speech, it may be
one factor that contributes to a more civil and encouraging discussion atmosphere. Several studies in
persuasion and psychology suggest that non-hostile humor can inhibit aggression (e.g., Baron & Ball,
1974), increase liking (Nabi et al., 2007), and encourage civility (Landreville et al., 2010). Humor,
however, has not been sufficiently studied in digital contexts (Holton & Lewis, 2011) and/or in rela-
tion to comment incivility.
In two online experiments, we contribute to the scholarly conversation on online incivility by ex-
ploring the effect of content, i.e., the article’s use of humor, and of the audience, i.e., other users’ inci-
vility, respectively on the incivility of participants’ responses. Our hypotheses are derived from the
Relief Theory of Humor and from previous research on anger and incivility. The general hypothesis
tested was: humor reduces anger that users experience in response to online content, which in turn
reduces their likelihood of posting an uncivil comment.

Theoretical Framework and Hypotheses

One probable cause of online incivility is the anger experienced by users when engaging with specific con-
tent. It is well established that anger is positively associated with expressions of incivility across different
settings. For example, results from a nationally representative survey showed that anger was positively as-
sociated with the sanction of incivility towards partisan opponents (Miller & Conover, 2015). Similarly,
surveys of employees showed a positive relation between anger and incivility in the workplace leading to
what is referred to as a spiral of incivility (Meier & Semmer, 2013). Anger also mediated the relation be-
tween exposure to elite incivility and uncivil responses (Gervais, 2017). In the context of online discus-
sions, Masullo Chen and Lu (2017) noted that while disagreements generally led to negative emotions and
aggressive intentions, only uncivil disagreements led people to respond uncivilly. Similarly, Masullo, Lu
and Fadnis (2020) found that exposure to incivility increased people’s feelings of anger. When those feel-
ings of anger were intense, people were more likely to speak out regardless of the opinion climate

Journal of Computer-Mediated Communication 00 (2022) 1–19 3

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

(Masullo et al., 2020). Thus, we expect participants who report being angered by online content to respond
with a more uncivil comment than participants who were not angered.
H1: Anger is positively associated with online incivility.

Relief Theory of Humor

The Relief Theory of Humor explains the positive effects of humor in situations where anger is

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

evoked. It is based on the idea that humor can distract from or counteract feelings of anger by creat-
ing a state of counter-arousal that releases “tensions, increases liking, and enhances source attraction”
(Walter et al., 2018). The state of amusement and attentional shifts created by humor may explain
why humorous appeals evoked less anger about climate change than fear-based or informational
appeals (Skurka et al., 2018) and why humorous messages were associated with increased positive
emotions, and reduced reactance (Skalski et al., 2009). We expected these findings would extend to
the reduction of online incivility.
H2: Humorous content will be negatively associated with anger.
H3: Humorous content will be negatively associated with comment incivility.
H4: Reduction of anger will mediate the relation between humor and comment incivility.
Studies have also pointed to the role of source liking in mediating humor effects on negative reac-
tions such as counter argument, and perceptions of bias (Nabi et al, 2007). According to Relief
Theory, humor works to release tensions by increasing liking and attraction (Walter et al, 2018).
Indeed, humor was found to increase source liking for political comedians (Nabi et al., 2007). We ex-
plored whether the use of humor might increase author liking, and in turn, reduce anger through the
research question:
RQ1: Does source (author) liking mediate the relation between humor and anger?
The variables and their hypothesized relations are presented in Figure 1.

Figure 1 Conceptual model of main variables and hypotheses.

4 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

Study 1
Study 1 tested Hypotheses 1–4 by manipulating the author’s use of humor in an op-ed article and ex-
amining the impact on the (in)civility of participants’ comments in a simulated online news forum.
Anger and source liking were measured to determine the indirect pathways through which humor
affects incivility.

Method

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Experimental Design
The experimental design was a one factor between-subjects design with two conditions (Humor vs.
No Humor). Participants were randomly assigned to one condition.

Participants
One hundred and twenty-two participants were recruited using Amazon’s Mechanical Turk. Participation
was voluntary and included only adult US citizens. The stimulus article discussed a US-specific issue. Each
participant was paid $2.50 USD for their participation, which lasted 12 minutes on average.
Of the resulting sample, 62% were male. Participants indicated their age using preset age catego-
ries (43% were between 18 and 24, 24% between 25 and 34, 19% between 35 and 44, 13% between 45
and 55, while none were above 55). Eighty-six percent had some college education. Forty-two percent
self-identified as Christian, 4% as Jewish, 3% as Buddhist, while the remaining 50% reported being
atheists or unaffiliated (none of our participants were Muslim). Fifty-two percent self-identified as lib-
eral while 48% leaned conservative.

Procedure and Stimulus

Participants were invited to take part in a study about “Journalism and Social Media” where they
would read, comment, and answer questions about two op-ed articles. We did not allude to humor,
incivility, or the fact that this is an experimental study in the title or description to avoid priming par-
ticipants with the study’s purpose. We used two actual op-ed articles from the New York Times: one
as a distractor article and the other as the experimental stimulus. The first article, titled “Can Students
Have Too Much Tech?” was a non-humorous article that described research findings regarding the
effects of technology on students (Pinker, 2015). It was chosen for its lack of ideological controversy.
It served as a distractor and remained the same across the experimental conditions.
The second article, and our main stimulus, was a humorous article titled “Inshallah is Good for
Everyone” by Wajahat Ali, a Muslim award-winning playwright. The article used the incident of a Muslim
college student taken off a flight for uttering the Arabic word “inshallah,” or God willing, to introduce the
non-threatening, humorous side of its meaning. It specifically utilized self-deprecating humor with regards
to American Muslims (Ali, 2016). We chose this article because it utilized humor around a serious topic
and could thus be easily modified to reflect a more serious tone without losing its informational value.
For the humor condition, humorous adjectives and jokes from the original article were kept in
the stimulus article. For the non-humor condition, humorous adjectives and jokes were removed or
paraphrased in a more serious tone. Table 1 presents an example of an article modification across the
two experimental conditions. Neither the author nor the articles’ publication information was readily
visible to participants to control for their possible effects.
After reading each article, participants were given a comment text box that resembled the one on
the New York Times website and were required to provide an anonymous comment2 about the

Journal of Computer-Mediated Communication 00 (2022) 1–19 5

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

Table 1. Example of Stimulus Article Excerpt Manipulated to Reflect the Humor  No-humor
Conditions

Humor No Humor

Escorted off an airline flight. Escorted off an airline flight.

Mr Makhzoomi wasn’t ranting about death, ter- Mr Makhzoomi wasn’t ranting about death, ter-
ror, Trump or artisanal mayonnaise—any of ror, Trump—any of which might warrant such
which might warrant such a drastic response. a drastic response.

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

article.3 They were informed their comments might be seen by participants in a future study. They
then responded to questions about each article corresponding to our dependent measures. In the
study debriefing, participants were informed about the goals of the study and were given a link to the
original articles on the New York Times website.

Dependent Measures
Source liking, humor, and anger were measured using pre-existing scales. The (in)civility of partici-
pants’ comments was scored using a machine learning sentiment analysis model.

Source Liking
Participants were asked to rate their liking for the author (Nabi et al., 2007) using three 7-point bipolar
adjectives: “0 ¼ Unfriendly, 6 ¼ Friendly”, “0 ¼ Unlikeable, 6¼ Likeable”, “0 ¼ Unpleasant, 6 ¼ Pleasant”.
A composite was created from the average of the three items (M ¼ 4.47, SD ¼1.58, a ¼ .98).

Humor
We used the three-question perceived humor scale by Duncan et al. (1984), which asked participants
to rate on a 7-point Likert scale (0 ¼ Strongly disagree, 6 ¼ Strongly agree) whether “the article was
funny,” “the article was more serious than it was funny (reversed),” and “most people would not find
the article to be humorous.” A composite was created from the average of the three items (M ¼ 2.63,
SD ¼1.68, a ¼ .92).

Anger
We incorporated three of Dillard and Shen’s (2005) questions for measuring emotional reactions per-
taining to anger, annoyance, and aggravation (0 ¼ definitely not, 6 ¼ definitely yes) by asking “did
you feel angry while reading this article?,” “did you feel annoyed while reading this article?” and “did
you feel aggravated while reading this article?” A composite was created from the average of the three
items (M ¼ 2.29, SD ¼ 1.86, a ¼ .91).

Online Incivility
Participants were asked to provide a comment about the stimulus article. To achieve a more objective
and systematic analysis, we utilized a machine-learning model rather than human raters to predict
the incivility of participants’ comments.

6 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

Predicting Incivility using Machine Learning

For this study, we developed a regression prediction model using machine learning to automate
the content analysis process by predicting the incivility of comments. Machine-learning (ML)
models can save time and effort with comparable accuracy to human raters, especially with larger
amounts of data (Weismayer et al., 2018). Using ML for content analysis can provide a more con-
sistent and objective way for analyzing content, as well as the ability to recognize new features in
the input. Unlike dictionary approaches which start with a predefined set of features of incivility
(Muddiman et al., 2019), deep neural network models recognize the unique features of uncivil

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

and civil comments through a process of supervised learning from labeled data and ultimately
learn to identify these features in unlabeled comments. We chose a regression prediction model
to capture the degree of predicted toxicity within the comments on a spectrum, and assess the ac-
curacy of the predicted values as opposed to a 0/1 classification model which may mask a partially
toxic comment as non-toxic.

Toxic Comments Data Set

We trained our model on the Toxic Comments Dataset ( 2 million pre-labeled comments), made
freely available through the Conversation AI team, a research initiative founded by Jigsaw and
Google. Due to the imbalance between toxic and civil comments (civil: 1,660,540, toxic: 144,334), we
sampled from the civil comments a randomized sample matching the size of toxic comments, so the
total training size was  290,000 comments.4

Incivility Prediction Model

We developed a regression prediction model that was trained on the pre-labeled toxic comments
data set to predict the incivility of an unlabeled set of comments. Incivility labels ranged from 0
to 1, where a comment with an incivility label  0.5 was considered uncivil (Borkan et al., 2019).
The model in turn learned the features distinguishing civil from uncivil comments (M ¼ .31, SD
¼ .22, MAE ¼ .16).
Our regression prediction model was a Feed-forward deep neural network,5 composed of four
dense layers: an input layer, two hidden layers and one output layer. We fed the model embed-
dings of training comment data; embeddings are numerical representations of words that carry
semantic relations in their encodings; hence, words with similar meanings would have similar nu-
merical representations. Because comments are usually longer than word length, we employed
Universal Sentence encoder (USE) which is “optimized for greater-than-word length text, such as
sentences, phrases or short paragraphs” (Universal Sentence, 2020). The model performed well in
terms of the mean squared error (MSE) and mean absolute error (MAE) which captures the dif-
ference between predictions and actual observations (see Table 2).

Results
Manipulation Check
Humor
Participants in the humor condition (M ¼ 2.90, SD ¼ 1.67) rated the article as more humorous than
participants in the no humor condition (M ¼ 1.69, SD ¼ 1.55), F (1,119) ¼ 47.12, p < .001, g2 ¼ .28.

Journal of Computer-Mediated Communication 00 (2022) 1–19 7

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

Table 2 MSE and MAE of Training and Test Data Sets using the Universal Sentence Encoder (USE)
Model

Training MSE Training MAE Test MSE Test MAE

.05 .16 0.06 .18

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Table 3 Means and Standard Deviations across all Conditions for Study 1

Condition All Conditions Humor No Humor

N 122 61 62
Anger 2.29 (1.86) 1.76 (1.74) 2.82 (1.84)
Liking 4.47 (1.58) 4.79 (1.61) 4.16 (1.49)
Comment incivility .31 (.22) .27 (.22) .35 (.21)

Note: Standard deviations are in parentheses. A scale of (0–6) was used for all measures except com-
ment incivility.

Descriptive Statistics
Table 3 reports means and standard deviations for the dependent variables across all conditions and
for each condition individually. Correlations are reported in Table 4.

Hypothesis Tests
A one-way Analysis of Variance (ANOVA) was conducted on the dependent variables. Levene’s test
of homogeneity of variances was not statistically significant (p > .82), so equal variances were
assumed.
To test H1, we conducted a one-way analysis of variance test. Participants were binned based on
the scale midpoint (3) into two groups based on their anger score (high vs. low anger).6 The two
groups differed in terms of their perceived incivility, where participants reporting high anger posted
comments that were significantly more uncivil (M ¼ .36, SD ¼ .21) than participants reporting low
anger (M ¼ .27, SD ¼ .22), F (1, 118) ¼ 4.12, p < .05, g2 ¼ .03, which supports H1.
In support of H2, participants in the humor condition also reported less anger in response to the
article (M ¼ 1.76, SD ¼ 1.74) than in the non-humor condition (M ¼ 2.83, SD ¼ 1.84), F (1, 118) ¼
10.78, p < .001, g2 ¼ .08, (Cohen’s d ¼ .60).
To test H3, we conducted a one-way analysis of variance (ANOVA) on comment incivility. As
expected, ANOVA revealed a significant main effect of humor on comment incivility. Participants’
comments in the humor condition (M ¼ .27, SD ¼ .22) scored significantly lower on incivility than
in the non-humor condition (M ¼ .35, SD ¼ .21), F(1, 119) ¼ 4.30, p < .05, g2 ¼ .03 (Cohen’s
d ¼ .38). This supports H3.7
To test H4, we performed a mediation analysis as proposed by Baron and Kenny (1986). Humor
predicted a significant reduction in comment incivility, b ¼ .19, t (119) ¼ 2.07, p < .05, with an
Adjusted R2 of .03, and a significant reduction in anger, b ¼ .56, t (118) ¼ 3.28, p < .001, with an

8 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

Table 4 Correlations among Dependent Variables for Study 1

1 2 3 4

1. Humor 
2. Anger .42*** 
3. Liking .21* .32** 
4. Comment Incivility .13 .13 .02 

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Note: n ¼ 122.
*p < .05,
**p < .01,
***p < .001.

Adjusted R2 of .08. When controlling for anger, the effect of humor on comment incivility was no
longer statistically significant, b ¼ .07, p ¼ .09, with an adjusted R2 of .03 but neither was the indi-
rect effect of anger, b ¼.01, p ¼ .36, which does not support the mediation hypothesis H4. However,
the small sample size (n ¼ 122) and the collinearity between perception of humor and anger
(r ¼ .42, p < .001) may have affected anger’s indirect effect on incivility (Preacher & Hayes, 2008).
To address RQ1 as to the role of source liking in mediating the relation between humor and an-
ger, we performed another mediation analysis as proposed by Baron and Kenny (1986). The test
revealed a significant effect of humor on source liking, b ¼ .20, t (119) ¼ 2.21, p < .05, with an
Adjusted R2 of .03 and a significant effect of source liking on reducing anger, b ¼ .32, t (119) ¼
3.63, p < .001, with an Adjusted R2 of .09. When controlling for liking, the effect of humor on anger
was reduced, b ¼ .23, t (119) ¼ 2.71, p < .01, while liking remained significant b ¼ .27, t (119)
¼ 3.11, p < .01, which establishes partial mediation.

Discussion
As expected, humor reduced online incivility. Humor also reduced reported feelings of anger.
Specifically, humor increased source liking, which subsequently worked to reduce anger. This may
help explain why humor can relieve tensions in negative situations, thereby adding to the existing lit-
erature on the psychology of humor. While participants reporting lower anger towards the article pro-
duced less uncivil comments than participants reporting higher anger, the mediation hypothesis
between humor and incivility through anger could not be established in Study 1. Causal mediation
effects tend to show more prominently with larger sample sizes (Fritz & MacKinnon, 2007). A post
hoc power analysis revealed that an n of approximately 128 would be needed to obtain statistical
power at the .80 level, and an n of approximately 170 to obtain statistical power at the .90 level.
Therefore, increasing the sample size can help overcome the effects of collinearity and obtain higher
statistical power in mediation analysis (Pieters, 2017), which we did in Study 2.
Study 1 investigated comment incivility in response to a news article. It controlled the impact of
other users’ comments on incivility by not providing any comments to participants. In a natural set-
ting, however, readers’ comments often appear with the article. Study 2 extended the findings of
Study 1 by examining if the incivility of other readers’ comments might reduce the positive effects of

Journal of Computer-Mediated Communication 00 (2022) 1–19 9

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

humor on incivility. Using the same stimulus article, we varied whether participants saw either a pre-
vious civil or uncivil previous comment before commenting.

Study 2
Study 2 investigated other readers’ comments as another possible trigger of anger and online incivil-
ity. Incivility is a dynamic normative phenomenon influenced by community feedback and proximate
comments which affect a person’s decision to respond with incivility (Shmargad et al., 2021).

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Incivility is also associated with partisanship. Rains et al. (2017) noted how the presence of incivility
in readers’ comments can polarize the discussion (Rains et al., 2017). Gervais (2015) found that view-
ing uncivil partisan television increased viewers’ uncivil online communication, especially when like-
minded incivility was present. However, the effect of uncivil messages on group polarization noted by
Thorson et al. (2008) depended on whether participants belonged to the attacked group. Likewise,
Gervais (2017) found that the use of incivility generated more anger when targeting one’s in-group.
Anonymity can also influence incivility outcomes when there is no direct social censure on the
perpetrator. Anonymous forums, for example, exhibited more incivility than non-anonymous forums
especially on the topic of immigration (Santana, 2014). However, recent research seems to downplay
the effect of anonymity in favor of the platform’s social norms (Rösner and Krämer, 2016). When
comments on a blog used aggressive words (aggressive group norm), participants used more aggres-
sive language in their comments as well (Rösner & Krämer, 2016). This can be explained by the phe-
nomenon of social contagion (Shen et al., 2020). In online games, exposure to incivility had a
significant effect on players becoming uncivil themselves (Shen et al., 2020). Hence, we predict:
H5: Participants will be more likely to respond with incivility to a previous uncivil comment
than to a previous civil comment.
Moods can also affect the incivility of discussions. A person’s negative mood combined with ex-
posure to other troll posts increased the likelihood of their engagement in trolling behavior (Cheng
et al., 2017). Exposure to incivility can result in increased negative emotions, such as anger (Phillips &
Smith, 2004) and moral indignation (Hwang et al., 2018). Uncivil comments can also lead to an in-
crease in readers’ hostile cognitions (Rösner et al., 2016). These hostile cognitions can manifest in
behaviors. Coe et al. (2014) noticed for example that incivility resulted in more negative feedback
than did the civil comments. Exposure to incivility increased incivility in participants’ responses as
well, where anger can play a mediating role between exposure and the uncivil response (Gervais,
2017). Hence, we hypothesize that:
H6: Anger will mediate the relation between previous incivility and participants’ incivility.
Additionally, several research studies suggested that the presence of incivility can be overpower-
ing, possibly reducing the effect of other features in messages. For example, people rated uncivil mes-
sages as less important and less informative than civil messages with the same information (Brooks &
Geer, 2007). However, they only found these effects for trait-based messages, suggesting that the pub-
lic generally frowns upon personal uncivil messages as opposed to uncivil issue- or policy-based mes-
sages. Hence, we hypothesize that:
H7: The positive effects of humor on incivility will be reduced when participants are presented
with an uncivil comment from another user. In other words, we expect an interaction effect of
humor and the civility of a previous user on comment incivility.

10 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

Method
Participants and Design
For Study 2, we conducted a 2 (Humor [humor, no humor]  3 (Previous user comment [civil, un-
civil supportive, uncivil unsupportive]) factorial experimental design.
Two hundred and eight participants were recruited using Amazon’s Mechanical Turk.
Participation was voluntary and included only online adult U.S. citizens. Each participant was paid
$2.50 USD for their participation, which lasted 14 minutes on average. Of the resulting sample, 57%
were male. Participants were asked to indicate their age using preset age categories (18% were between

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

18 and 24, 46% between 25 and 34, 23% between 35 and 44, 5% between 45 and 55, and 2% above 55,
and the rest withheld their age). Eighty-nine percent had some degree of college education. Forty-
three percent belonged to Christian denominations, 1% were Jews, 3% Buddhists, while the rest were
either atheists or unaffiliated (none of our participants were Muslim). Like Study 1, 52% of partici-
pants self-categorized as liberal while the rest leaned conservative (center to extremely conservative).
Participants were randomly assigned to experimental conditions.

Procedure and Stimuli

We used the same procedure as Study 1 in Study 2. The only change was that a comment made by a
previous participant appeared underneath the article. The distractor article and its comment8
remained the same across all conditions, while the target article, “Inshallah is Good for Everyone,”
was modified to reflect the different conditions of the two IVs. The manipulation of humor in the tar-
get article was the same as Study 1.
For the Previous User Comment manipulation, we chose actual participants’ comments from
Study 19 and placed one underneath the article. For each condition of social influence, we randomly
picked one of two representative comments from each category to display.
For the civil comment, we chose examples coded in our first study as civil (incivility < .5) (e.g.,
“What a refreshing perspective. Amidst all the fear and noise [sic] it is imperative that we hear the
voices of people affected by Islamophobia”). For uncivil supportive comments, we chose comments
coded as uncivil (incivility > .5) yet supportive of the article (e.g., “Just shows how much the GOP
and the Trumpeters have continued to cause hate in this country. Should be the GONP. Grand Old
Nazi Party”). For uncivil unsupportive, we chose comments coded as uncivil and negative towards the
author/article (incivility > .5) (e.g., “The article is trying to tow the phony liberal logic that Muslims
are not destroying every country that they have been ‘welcomed’ into nor are they responsible for the
9/11 attacks”).

Measures
We used the same machine learning model, scales, and dependent measures from Study 1. All a > .87.

Results
Manipulation Check
The humor manipulation was successful. Participants in the humor condition (M ¼ 2.90, SD ¼ 1.67)
viewed the article as more humorous than participants in the non-humor condition (M ¼ 1.69,
SD ¼ 1.55), F(1, 206) ¼ 29.52, p < .000, g2 ¼ .12. We did not include an incivility manipulation check
to avoid revealing the experiment’s true purpose.

Journal of Computer-Mediated Communication 00 (2022) 1–19 11

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

Descriptive Statistics
Table 5 summarizes means and standard deviations across conditions. Correlations among dependent
variables are reported in Table 6.
Hypothesis Tests
Analysis of Variance (ANOVA) was conducted on the dependent variables. Levene’s tests of homoge-
neity were not significant (all p’s > .13), so equal variances were assumed.
Consistent with Study 1, participants in the humor condition (M ¼ .28, SD ¼ .20) produced
comments significantly lower in incivility than in the non-humor condition (M ¼ .35, SD ¼ .21),

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

F (1, 202) ¼ 5.34, p < .05, g2 ¼ .03. They also reported less anger (M ¼ 2.21, SD ¼ 1.87) than their
counterparts in the non-humor condition (M ¼ 2.93, SD ¼ 1.86), F (1, 204) ¼ 8.02, p < .01, g2 ¼
.04. Source liking mediated the relations between humor and anger, where humor significantly re-
duced anger b ¼ .19, t(206) ¼ 2.82, p < .01, and significantly increased source liking, b ¼ .38,
t(206) ¼ 5.81, p < .001. When controlling for source liking, the effect of humor was no longer sig-
nificant, b ¼ .04, t(206) ¼ .56, p > .57, while effect of liking remained significant b ¼ .41,
t(206) ¼ 6.05, p < .001, establishing full mediation.
Unlike Study 1, anger in Study 2 fully mediated the relation between humor and incivility. Humor
predicted a significant reduction in incivility, b ¼ .16, t(206) ¼ 2.32, p < .05, with an adjusted R2 of
.02, and in anger, b ¼ .19, t(206) ¼ 2.82, p < .001, with an adjusted R2 of .03. Anger predicted com-
ment incivility, b ¼.21, t (206) ¼ 3.07, p < .001, with an adjusted R2 of .04. When controlling for anger,
the effect of humor on incivility was no longer significant, b ¼ .12, p ¼ .07, while the effect of anger
on incivility remained significant, b ¼ .19, t(206) ¼ 2.68, p < .05, establishing full mediation.
With regards to the effect of previous user comments on comment incivility, analyses revealed no
statistically significant main effect on comment incivility, F(2, 204) ¼ 1.29, p ¼ .27, g2 ¼ .01. Hence
H5 was not supported.
H6 predicted that anger will mediate the relation between previous user comments and comment
incivility. Multiple regression analysis indicated no statistically significant effect on anger b ¼ .01,
t(205) ¼ 1.15, p ¼ .28. So, H6 was not supported.
H7 predicted that a previous uncivil comment would reduce the effect of humor on participant
incivility, however, no significant interaction effect was found between humor and the civility of pre-
vious user comments (p > .86). In fact, the effect of humor on reducing comment incivility remained
statistically significant (p < .05); hence, H7 was not supported.

Discussion
The effects of humor on anger, liking, and comment incivility observed in Study 1 were replicated in
Study 2. Additionally, possibly due to the larger sample size, the mediation hypothesis of Study 1 was
supported, where anger fully mediated the relation between humor and comment incivility.
Surprisingly, however, participants did not appear to be influenced by the comment of another user.
This suggests that reactions to the article’s content had a greater impact on user incivility than the
reactions of a previous user.

General Discussion
Taken together, the two studies contribute to our understanding of comment incivility in online news
forums: what triggers it and what factors can be used to reduce it. We identified anger as a trigger and

12 Journal of Computer-Mediated Communication 00 (2022) 1–19

 Y. Elsayed and A. B. Hollingshead

Table 5 Means and Standard Deviations across All Conditions for Study 2

Condition All Humor/ Civil Humor/ Humor/ No Humor/ No Humor/ No Humor/

Conditions Uncivil Uncivil Civil Uncivil Uncivil

Journal of Computer-Mediated Communication 00 (2022) 1–19

Supportive Unsupportive Supportive Unsupportive

N 208 37 32 35 37 38 34
Anger 2.58 (1.89) 1.87 (1.78) 1.88 (1.83) 2.83 (1.88) 3.04 (1.83) 2.94 (1.92) 2.81 (1.89)
Liking 4.45(1.7) 5.43 (1.62) 5.00 (1.53) 4.87 (1.59) 3.79 (1.68) 3.82 (1.57) 3.82 (1.69)
Comment .31(.21) .25(.18) .28(.20) .32(.21) .32(.19) .36(.19) .36(.25)
Incivility

A scale of (0–6) was used in all measurements except comment incivility.

Humor and Online Incivility

13
Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022
Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

Table 6 Correlations among Dependent Variables for Study 2

1 2 3 4

1. Humor 
2. Anger .31*** 
3. Liking .16* .43*** 
3. Comment Incivility .21** .21** .12 

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Note: n ¼ 208.
*p < .05,
**p < .01,
***p < .001.

humor as a mechanism to reduce it. Source liking explained the relation between humor and anger,
where source liking fully mediated the relation between humor and anger. In both studies, we found
that humorous content was positively associated with reduced comment incivility and reduced anger.
This supports as well as extends the Relief Theory of Humor, where we found that the state of counter
arousal created by humor is facilitated by liking and results in a reduction of anger that subsequently
reduces the likelihood of behaviors such as uncivil public discourse. These findings support the value
of tapping into emotional and affective channels in dealing with behaviors such as incivility, although
whether humor would work on very angry people is still an open question, as participants reported
only moderate levels of anger in this study.10 Study 2 directly replicated the results of Study 1 with a
larger sample, which provides greater validity to the research findings. It also enabled us to use the
same machine learning model which ensured consistency for the content analysis process. Future re-
search should investigate whether the model generalizes across a variety of content genres and topics.
The (in)civility of another user’s comment did not seem to reduce the effect of humor on incivil-
ity, which supports a more nuanced view of audience effects on online discourse (Brooks & Geer,
2007). The identity of the attacked group may have played a role, where the stimulus article described
an episode of discrimination directed at a minority group (Muslims) to which none of the participants
belonged in our studies. Incivility can have differential effects depending on whether participants
belonged to the attacked group (Thorson et al., 2008; Gervais, 2017). Additionally, the content and
previous user comments were not personal attacks, which could have affected our results (Brooks &
Geer, 2007). Finally, we utilized only one previous user comment to create the condition of incivility,
while other studies have mostly relied on many user comments, which may have affected participants’
sense of the prevalence of incivility. According to Lee et al. (2019) discussion volume can influence
comment incivility. Future studies should explore the interplay between group identification, expo-
sure to incivility, and the volume of uncivil comments and how it may affect ensuing incivility.
Our studies also made a methodological contribution. The experimental method used in these
studies was novel: Participants were asked to provide a comment on a news article in a simulated on-
line forum. We required all participants to write a comment to test hypotheses about the relation of
feelings of anger to actual comments. In actual online news forums, users can often choose whether to
comment or not. Future research should examine whether these findings replicate when participants
can choose to comment.
Additionally, the stimulus article utilized self-deprecating humor, which may be viewed as dis-
arming by audiences. Different types of humor such as satire or irony may induce other reactions

14 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

along the political spectrum (Young, 2019). Future experiments should investigate whether our results
replicate with other types of humor. The studies also demonstrated the potential of using machine
learning models in successfully predicting uncivil content. This opens the possibility of using them in
a variety of other experiments due to their systematic performance and comparable accuracy to hu-
man raters (Weismayer et al., 2018). While our incivility-prediction model generally performed well,
achieving very low discrepancy from actual observations in the training data, it still exhibited some in-
accuracies. The toxicity dataset adopts an approach to labeling data that is substantially different from
the extensive codebooks required for more theoretically driven definitions of incivility. Additionally,

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

according to Jigsaw, due to the frequency of co-occurrence of certain identities with incivility, models
may wrongfully learn to associate the mention of these identities with incivility. This is the problem
of unintended bias “Unintended bias,” 2019). This would result in classifying a comment such as, “It’s
sad how misinformed people are. I have relatives who are Muslim and they are the kindest people I
know!” as uncivil. Future research could address this by allowing models to see a more diverse set of
abusive/non-abusive examples mentioning Muslims and other identities.
Our findings have real-world implications that can be helpful to opinion writers struggling to dis-
cuss controversial topics, and to social media platforms battling online toxicity. We acknowledge that
not all incivility is bad, and sometimes it is needed to bring attention to the causes of marginalized
communities (Masullo et al., 2019). This belief is reflected in our choice of a regression model that
predicts a value for toxicity rather than simply classifying content as toxic/not-toxic. In the context of
social media, humor can be used to nudge users into a more civil conversation by releasing the nega-
tive emotions that accompany incivility. It is also less heavy-handed than a ban on incivility, which
can prove counterproductive (Masullo et al., 2019).
Finally, while the stimulus article in this study was about a person from a minority group, the
findings should not be understood as an invitation to tackle the underlying problems of racism or
hate speech using humor. The article should only be understood as a stimulus to learn about the
effects of humor on source liking, anger, and their mediating effect on comment incivility. While the
findings can be helpful in deflecting general situational incivility, racism or hate speech remain far
deeper problems in need of reeducation and structural solutions.

Data Availability Statement

The data underlying this article will be shared on reasonable request to the corresponding author.

Endnotes
1. The use of the terms toxicity and incivility can be dependent on the academic field in which
they are used, and at other times they are used interchangeably (Shen et al., 2020).
2. To create the conditions most likely to promote incivility, respondents’ comments were anony-
mous. Anonymous comments are more likely to be uncivil than non-anonymous comments
(Santana, 2014).
3. To make sure participants read the stimulus article, they were asked two recall questions:
whether the article was about a personal experience or not, and what language the word
“inshallah” was in, according to the article. Hundred percent of study participants responded
correctly to the recall questions.

Journal of Computer-Mediated Communication 00 (2022) 1–19 15

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

4. For details on how Jigsaw labeled the toxic comments dataset, please refer to https://medium.
com/jigsaw/creating-labeled-datasets-and-exploring-the-role-of-human-raters-56367b6db298
5. Model was developed using the Keras framework’s Sequential model powered by Google’s
Tensorflow in Python.
6. There were more participants in low anger (n ¼ 80) than high anger (n ¼ 41), midpoint value 3
¼ low anger.
7. All main effects remained statistically significant even when controlling for demographic varia-
bles such as gender, age, and education.

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

8. The comment on the distractor article was “It all makes sense. If a kid is not supervised, it begins
a downward spiral. Whether it’s a lack of cuddling or lack of supervision regarding time spent
in front of a computer; parents need to spend time with their kids.” It was chosen for its agree-
ment with/and summary of the article’s points.
9. Comments from Study 1 were chosen for their representativeness of each category of incivility
(hence they were either very civil or very uncivil but not neutral).
10. Participants do not generally experience high levels of anger in experimental studies (e.g.,
Dillard & Shen, 2005).

References
* [dataset] Conversation AI, 2018, Toxic Comments Dataset, Kaggle. https://www.kaggle.com/c/jig
saw-toxic-comment-classification-challenge/data
Ali, W. (2016, April 22). Inshallah is good for everyone. The New York Times. Retrieved from https://
www.nytimes.com.
Aten, Jason. Twitter Wants to Become Less Toxic by Letting Users Control Who Can Reply to Tweets.
Inc. Retrieved from https://www.inc.com
Baron, R. A., & Ball, R. L. (1974). The aggression-inhibiting influence of non hostile humor. Journal of
Experimental Social Psychology, 10(1), 23–33. https://doi.org/10.1016/0022-1031(74)90054-7
Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psycho-
logical research: Conceptual, strategic, and statistical considerations. Journal of Personality, Social,
and Psychology, 51, 1173–1182. https://doi.org/10.1037/0022-3514.51.6.1173
Bieg, M. (2013). Trait and state academic emotions: Two sides of the same coin? (Doctoral dissertation).
https://d-nb.info/1114396826/34
Borkan, D., Dixon, L., Sorensen, J., Thain, N., & Vasserman, L. (2019, May). Nuanced metrics for mea-
suring unintended bias with real data for text classification. In Companion Proceedings of the 2019
World Wide Web Conference (pp. 491–500). https://doi.org/10.1145/3308560.3317593
Borkan, D., Sorensen, J., & Vasserman, L. (2019). Creating labeled datasets and exploring the role of
human raters. Retrieved fromhttps://medium.com/the-false-positive/creating-labeled-datasets-
and-exploring-the-role-of-human-raters-56367b6db298
Brooks, D. J., & Geer, J. G. (2007). Beyond negativity: The effects of incivility on the electorate.
American Journal of Political Science, 51(1), 1–16. http://www.jstor.org/stable/4122902.
Brückner, L., & Schweiger, W. (2017). Facebook discussions of journalistic news: Investigating article
objectivity, topic, and media brand as influencing factors. SCM Studies in Communication and
Media, 6(4), 365–394. https://doi.org/10.1177/0267323100015001002
Chen, G. M. (2017). Online incivility and public debate: Nasty talk. New York: Palgrave Mcmillan.
Cheng, J., Bernstein, M., Danescu-Niculescu-Mizil, C., & Leskovec, J. (2017, February). Anyone can be-
come a troll: Causes of trolling behavior in online discussions. In Proceedings of the 2017 ACM

16 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

conference on computer supported cooperative work and social computing (pp. 1217–1230). https://
doi.org/10.1145/2998181.2998213
Chua, C. E. H. (2009). Why do virtual communities regulate speech? Communication Monographs
76(2): 234–261. https://doi.org/10.1080/03637750902828420.
Coe, K., Kenski, K., & Rains, S. A. (2014). Online and uncivil? Patterns and determinants of incivility
in newspaper website comments. Journal of Communication, 64(4), 658–679. https://doi.org/10.
1111/jcom.12104
Dillard, J. P., & Shen, L. (2005). On the nature of reactance and its role in persuasive health communi-

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

cation. Communication Monographs, 72, 144–168. https://doi.org/10.1080/03637750500111815
Duncan, C. P., Nelson, J. E., & Frontczak, N. T. (1984). The effect of humor on advertising comprehen-
sion. Advances in Consumer Research, 11, 432–437.
Fritz, M. S., & MacKinnon, D. P. (2007). Required sample size to detect the mediated effect.
Psychological Science, 18(3), 233–239. https://doi.org/10.1111/j.1467-9280.2007.01882.x
Gervais, B. T. (2015). Incivility online: Affective and behavioral reactions to uncivil political posts in a
web-based experiment. Journal of Information Technology & Politics, 12(2), 167–185. https://doi.
org/10.1080/19331681.2014.997416
Gervais, B. T. (2017). More than mimicry? The role of anger in uncivil reactions to elite political inci-
vility. International Journal of Public Opinion Research, 29(3), 384–405. https://doi.org/10.1093/
ijpor/edw010
Goujard, C. (2016, September 8). Why news websites are closing their comments sections. Retrieved
from https://medium.com.
Holton, A. E., & Lewis, S. C. (2011). Journalists, social media, and the use of humor on Twitter.
Electronic Journal of Communication, 21, 1–2. https://doi.org/10.36106/gjra
Hwang, H., Kim, Y., & Kim, Y. (2018). Influence of discussion incivility on deliberation: An examina-
tion of the mediating role of moral indignation. Communication Research, 45(2), 213–240. https://
doi.org/10.1177/0093650215616861
Itzkovich, Y., Alt, D., & Dolev, N. (2020). Personal precursors of academic incivility. In The challenges of ac-
ademic incivility (pp. 35–77). Cham: Springer. https://doi.org/10.1007/978-3-030-46747-0_4
Juni, S., & Katz, B. (2001). Self-effacing wit as a response to oppression: Dynamics in ethnic humour.
The Journal of General Psychology, 128 (2), 119–142. https://doi.org/10.1080/00221300109598903
Landreville, K. D., Holbert, R. L., & LaMarre, H. L. (2010). The influence of late-night TV comedy
viewing on political talk: A moderated-mediation model. The International Journal of
Press/Politics, 15, 482–498. https://doi.org/10.1177/1940161210371506
Lee, F. L., Liang, H., & Tang, G. K. (2019). Online incivility, cyberbalkanization, and the dynamics of
opinion polarization during and after a mass protest event. International Journal of
Communication, 13, 20. https://doi.org/1932–8036/20190005.
Masullo Chen, G. & Lu, S. (2017) Online Political Discourse: Exploring Differences in Effects of Civil
and Uncivil Disagreement in News Website Comments. Journal of Broadcasting & Electronic
Media. 61 (1) 108–125. 10.1080/08838151.2016.1273922.
Masullo Chen, G., Muddiman, A., Wilner, T., Pariser, E., & Stroud, N. J. (2019). We should not get rid
of incivility online. Social Media þ Society, July-September, 1–5. https://doi.org/10.1177/
2056305119862641
Masullo Chen, G. M., Lu, S., & Fadnis, D. (2020). Does online incivility cancel out the spiral of silence?
A moderated mediation model of willingness to speak out. New Media & Society, 23(11), 1–24.
https://doi.org/10.1177/1461444820954194

Journal of Computer-Mediated Communication 00 (2022) 1–19 17

Humor and Online Incivility Y. Elsayed and A. B. Hollingshead

Meier, L. L., & Semmer, N. K. (2013). Lack of reciprocity, narcissism, anger, and instigated workplace
incivility: A moderated mediation model. European Journal of Work and Organizational
Psychology, 22(4), 461–475. https://doi.org/10.1080/1359432X.2012.654605
Miller, P. R., & Conover, P. J. (2015). Red and blue states of mind: Partisan hostility and voting in
the United States. Political Research Quarterly, 68(2), 225–239. https://doi.org/10.1177/
1065912915577208
Minich, A. L., Mendoza, A. T. O., & Brown, E. C. (2018). Effect of incivility on attitudes and informa-
tion seeking. North American Journal of Psychology, 20(1).

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

Muddiman, A., McGregor S., & Stroud J. (2019) (Re)claiming our expertise: parsing large text corpora
with manually validated and organic dictionaries. Political Communication, 36(2), 214–226.
https://doi.org/10.1080/10584609.2018.1517843
Nabi, R. L., Moyer-Guse, E., & Byrne, S. (2007). All joking aside: A serious investigation into the per-
suasive effect of funny social issue messages. Communication Monographs, 74(1), 29–54. https://
doi.org/10.1080/03637750701196896
Papacharissi, Z. (2004). Democracy online: Civility, politeness, and the democratic potential of online polit-
ical discussion groups. New Media & Society, 6, 259–283. https://doi.org/10.1177/1461444804041444
Perez, S. (2021, May 5). Twitter rolls out improved reply prompts to cut down on harmful tweets. Tech
Crunch. Retrieved from https://techcrunch.com.
Phillips, T., & Smith, P. (2004). Emotional and behavioural responses to everyday incivility:
Challenging the fear/avoidance paradigm. Journal of Sociology, 40(4), 378–399. https://doi.org/10.
1177/1440783304048382
Pieters, R. (2017). Mediation analysis: Inferring causal processes in marketing from experiments.
In Advanced Methods for Modeling Markets (pp. 235–263). Cham: Springer. https:
//doi.org/10.1007/978-3-319-53469-5_8
Pinker, S. (2015, January 30). Can students have too much Tech? The New York Times. Retrieved from
https://www.nytimes.com.
Preacher, K. J., & Hayes, A. F. (2008). Asymptotic and resampling strategies for assessing and compar-
ing indirect effects in multiple mediator models. Behavior research methods, 40(3), 879–891.
https://doi.org/10.3758/BRM.40.3.879
Rains, S. A., Kenski, K., Coe, K., & Harwood, J. (2017). Incivility and political identity on the internet:
Intergroup factors as predictors of incivility in discussions of news online. Journal of Computer-Mediated
Communication, 22(4), 163–178. https://doi/org/https://doi.org/10.1111/jcc4.12191
Rheault L.Rayment E.Musulan A. (2019) Politicians in the line of fire: Incivility and the treatment of
women on social media. Research & Politics. 6 (1). https://doi.org/10.1177/2053168018816228.
Rösner L. Krämer N. Verbal venting in the social web: Effects of anonymity and group norms on ag-
gressive language use in online comments. Social Media Society. 2016. 2 (3) 1–13.
Rösner, L., Winter, S., & Krämer, N. C. (2016). Dangerous minds? Effects of uncivil online comments
on aggressive cognitions, emotions, and behavior. Computers in Human Behavior, 58, 461–470.
https://doi.org/10.1016/j.chb.2016.01.022
Santana, A. D. (2012) Civility, anonymity and the breakdown of a new public sphere (Doctoral
Dissertation), University of Oregon, 2012,
Santana, A. D. (2014). Virtuous or vitriolic: The effects of anonymity on civility in online newspapers reader
comment boards. Journalism Practice, 8(1), 18–33. https://doi.org/10.1080/17512786.2013.813194
Shen, C., Sun, Q., Kim, T., Wolff, G., Ratan, R., & Williams, D. (2020). Viral vitriol: Predictors and con-
tagion of online toxicity in World of Tanks. Computers in Human Behavior, 106343. https://doi.
org/10.1016/j.chb.2020.106343

18 Journal of Computer-Mediated Communication 00 (2022) 1–19

Y. Elsayed and A. B. Hollingshead Humor and Online Incivility

Shmargad, Y., Coe, K., Kenski, K., & Rains, S. A. (2021). Social norms and the dynamics of online inci-
vility. Social Science Computer Review. https://doi.org/10.1177/0894439320985527.
Skalski, P., Tamborini, R., Glazer, E., & Smith, S. (2009). Effects of humor on presence and recall of
persuasive messages. Communication Quarterly, 57(2), 136–153. https://doi.org/10.1080/
01463370902881619
Skurka, C., Niederdeppe, J., Romero-Canyas, R., & Acup, D. (2018). Pathways of influence in emo-
tional appeals: Benefits and tradeoffs of using fear or humor to promote climate change-related
intentions and risk perceptions. Journal of Communication, 68(1), 169–193. https://doi.org/10.

Downloaded from https://academic.oup.com/jcmc/article/27/3/zmac005/6552940 by guest on 28 August 2022

1093/joc/jqx008
Tromble, R. (2018). Twitter “healthy conversations”. Retrieved from https://www.rebekahtromble.net/
political-communication
Thorson, K., Fung, T., & Vraga, E. K. (2008, August). How you feel makes you what you are: Partisan
reactions to political incivility online. In Annual Meeting of the Association for Education in
Journalism and Mass Communication, Chicago, IL, August (pp. 6–9).
Universal-Sentence-Encoder (2020). Tensor Flow Hub. Retrieved from https://tfhub.dev/google/univer
sal-sentence-encoder/2
Unintended bias in toxicity classification. (2019). Jigsaw. Retrieved fromhttps://www.kaggle.com/c/jig
saw-unintended-bias-in-toxicity-classification/overview
Walter, N., Cody, M. J., Xu, L. Z., & Murphy, S. T. (2018). A priest, a rabbi, and a minister walk into a
bar: A meta-analysis of humor effects on persuasion. Human Communication Research, 44(4),
343–373. https://doi.org/10.1093/hcr/hqy005
Weismayer, C., Pezenka, I., & Gan, C. H. K. (2018). Aspect-based sentiment detection: Comparing human
versus automated classifications of TripAdvisor reviews. In Information and communication technolo-
gies in tourism 2018 (pp. 365–380). Springer. https://doi.org/10.1007/978-3-319-72923-7_28
Young, D. G. (2019). Irony and outrage: The polarized landscape of rage, fear, and laughter in the
United States. USA: Oxford University Press. https://doi/org/10.1093/oso/9780190913083.001.0001

Journal of Computer-Mediated Communication 00 (2022) 1–19 19