The Effect of Twitter Cancel Culture on the Music Industry

Nikitha Gopal, Abigail Velasquez, Peter Wu

March 16, 2022

1 Abstract
Musicians often trend on social media for various reasons but in recent years, there has been a rise
in musicians being “canceled” for committing offensive or socially unacceptable behavior. Due to
the wide accessibility of social media, the masses are able to hold accountable musicians for their
actions through “cancel culture”, a form of modern ostracism. Twitter has become a well-known
platform for “cancel culture” as users can easily spread hashtags and see what’s trending, which
also has the potential to facilitate the spread of toxicity. We analyze how public sentiment towards
canceled musicians on Twitter changes in respect to the type of issue they were canceled for, their
background, and the strength of their parasocial relationship with their fans. Through our research,
we aim to determine whether cancel culture on Twitter could lead to an increase in toxicity towards
a canceled individual.

2 Introduction
Every so often, a public figure or celebrity would trend on Twitter over behavior that the general
public would consider offensive or socially unacceptable by today’s standards. From politics to
Hollywood to the music industry, all kinds of public figures have been called out and criticized for
such behavior by the masses. This form of modern ostracism is known as “cancel culture”, in which
the masses attempt to call out and boycott an individual in response to problematic remarks or
ideologies that they have expressed. This online phenomenon has resulted in celebrities facing real
life repercussions. For example, the world renown rapper, Travis Scott recently faced backlash for
lack of crowd control which results in multiple deaths and injuries (Lamarre). Due to the public
uproar, multiple lawsuits were filed against Scott and many organizers canceled his upcoming
performances. His music also struggled to find air time on the radio. Another example is R. Kelly,
an American singer and songwriter, who was accused for sexually assualting minors. In fact, R.
Kelly was arrested for 10 counts for sexual asualting four women. The hashtag “#MuteRKelly”,
however, began trending on Twitter when “Survingn R. Kelly”, an investigative series began airing
on the US network Lifetime. The hashtag was pushing for R. Kelly songs to be removed from radio
and music applications (Associated Press).

3 Problem Statement and Objectives

To investigate how cancel culture affects the overall public perception of musicians on Twitter,
our main research question is the following: among English-speaking Twitter users, how does their
sentiment towards canceled musicians change over time? As a result of getting canceled, we believe
that there will be an overall increase in toxicity and negativity in terms of public sentiment on

1
Twitter towards the canceled individual. To understand the nuances and quantify the effect of
cancel culture on public figures, we will be conducting case studies on six individuals from three
different genres within the music industry, using Twitter as our main source and platform. To assist
our main investigation, we ask three sub-research questions:
• RQ1: How does sentiment differ over time based on the type of issue the individual was
canceled for?
• RQ2: How does sentiment differ based on the individual’s gender and the genre they cater
to?
• RQ3: How does sentiment differ for individuals with and without fans that have a strong
parasocial connection to the individual?

4 Literature Review
Many social media platforms have been criticized for not doing enough to mitigate or reduce the
spread of misinformation and hate speech within online communities. In response to such criticism,
social media platforms employed the idea of deplatforming, where the platform temporarily or
permanently bans an influential figure’s account for not complying with their guidelines (Fiesler
et al.). However, deplatforming or banning moderate entire communities on Reddit, Facebook or
Twitter, however, is not a widespread tool for content moderation. This is because such action
begins to raise questions about freedom of expression and speech. Apart from the human rights
perspective, deplatforming influential figures could result in a monetary deficit because, as a result,
supporters of the figure may be inclined to stop using the platform. Unfortunately, due to this,
the method of boycotting an influential figure via deplatforming only occurs in extreme cases.
However, the rise of social media has given the opportunity to many marginalized groups to make
an impact by shedding light on issues that are overlooked by the majority. This ability propelled the
phenomenon of “cancel culture”, where the public boycotts an individual figure with the intention
of ending the public figure’s career or pressing for disciplinary action (Velasco).
Although “cancel culture” is a relatively new phenomenon, the concept has been employed by the
public for years. In the 2010s, the term “call-out culture” had roots in online platforms when it was
used to describe Tumblr fandoms calling out blogs (Romano). Prior to that, “call-out culture” was a
technique commonly utilized through other mediums of free speech during the civil rights movement
(Clark). However, due to the global nature of social media platforms, like Twitter, cancel culture has
become more prominent and impactful than before. As mentioned above, many entertainers, such
as R.Kelly, Kanye West, Scarlett Johanssen, have been targeted and held accountable through this
process. Entertainers usually face criticism from the public for issues that are “fueled by politically
progressive social media” (Romano). Some notable examples are statements discriminating against
African Americans or the LGBTQ+ community or actions violating women’s rights.
Although more marginalized groups are being heard, the main question remains: is cancel culture
effective? The public strives to hold a public figure accountable through pushing for a tangible
effect on a public figure’s career. However, as indicated in the Travis Scott example, there are
some immediate setbacks that occur but it is not detrimental. In fact, at the very least, the
artists will continue to collect revenue from streaming platforms. Even in the most extreme cases,
like Louis C.K., who faced sexual misconduct allegations, only had a 10 month career hiatus
(Romano). Moreover, he returned to sold-out stand-up comedy shows. Similarly, R. Kelly saw
increase in number of music streams despite facing sexual assault cases (Romano). Due to the
possible controversies this online phenomenon may cause, it is important to analyze and understand

2
the impact or nuances of “cancel culture” public figures. We used the paper ‘Evaluating the
Effectiveness of Deplatforming as a Moderation Strategy on Twitter, as a guide for determining the
effectiveness of cancel culture (Jhaver et al). We will specifically focus on exploring how the public
sentiment is altered towards an entertainer, specifically musicians, in response to their controversy.

5 Data
We collected data from three different music genres: K-pop, Hip-Hop, and Pop. For each music
genre, we chose one male and one female music artist who has gotten canceled as well as one male
and female artist who have not had any controversies that they got canceled for to act as our control
group. Each individual that we have chosen also has a Twitter account that is run by themselves
and/or their social media team.

5.1 Background on Case Studies

For K-pop, our two canceled individuals were NCT and WayV’s Lucas and Aespa’s Giselle. After
allegations from alleged former girlfriends of gaslighting and cheating, Lucas was called out and
then apologized for such behavior two days after the allegations appeared (Kim). As of March 2022,
Lucas still remains on hiatus. Giselle was called out on Twitter for lip-syncing a racial slur from
American R&B artist SZA’s track “Love Galore” in a behind-the-scenes video (Kim). Giselle also
apologized a few days after the incident occured. For our control individuals for the K-pop genre,
we chose NCT and NCT Dream’s Jaemin and ITZY’s Ryujin. In choosing our control individuals
for this genre, we chose individuals that were part of groups that debuted or were formed within
the same time frame as the groups that the canceled individuals are part of and their respective
groups have at least over a million Twitter followers on their official Twitter accounts. It is also
important to note that although Lucas and Jaemin are both members of boy group NCT, they
work under different sub-units or groups as Lucas is a member of WayV and Jaemin is a member
of NCT Dream. This means that they are not directly associated with each other as they do not
work together on a regular basis as part of the same sub-unit/group. Therefore, it is highly unlikely
that the reputation of one would affect the other’s reputation.
The next group of artists we analyzed were Western hip hop artists. In this group, the two
canceled artists were DaBaby and Nicki Minaj. During a festival concert in July of 2021, DaBaby
made some remarks to the crowd that were homophobic. After that concert, the controversy went
mainstream while many festivals and brands dropped DaBaby (Chan). Nicki Minaj faced backlash
after she tweeted vaccine misinformation after being asked her vaccination status (Romano). This
controversy occurred in September of 2021 while the COVID-19 pandemic was still affecting many
around the world. For the control artists in the hip hop group, we chose Lil Baby and Saweetie.
The main criteria for these artists were if they had a similar following or presence as the canceled
individuals and didn’t experience cancellation from controversies. Lil Baby and DaBaby have a
similar number of Twitter followers at 6.7 and 4.4 million respectively. However, Nicki Minaj is the
most followed female hip hop artist on Twitter by a large margin at 23.7 million followers. For our
female control artist, we chose Saweetie as we felt she had cultivated a similar following to Nicki
Minaj through her music and online presence.
The third group of artists we analyzed were Western pop artists. Two individuals who faced
controversy in this genre were Doja Cat and Zayn Malik. Around the end of May 2020, the hashtag
“#DojaCatIsOverParty” began circulating on Twitter since Doja Cat’s 2015 song “Dindu Nuffin”
resurfaced (France). The term, “Dindu Nuffin”, is recognized as ‘a racial slur used to mock victims

3
of police brutality’ (France). Following this, more racist allegations emerged suggesting that Doja
Cat was contributing to racist conservations on online chat rooms years prior. In response to the
controversy, Doja Cat apologized and explained the reasoning behind her actions in an Instagram
post, which is now deleted. Zayn Malik recently faced some backlash due to physical assault
allegations made by his mother-in-law, Yolanda Hadid. On October 28 2021, TMZ reported that
Yolanda Hadid filed a police report against Zayn for domestic violence (Pham). As a result, Zayn
was required to pay fines, complete 90 day probation, and attend anger management classes (Pham).
For the Western Pop control artists, we have chosen Adele and Harry Styles. As mentioned above,
the main criteria for our control artists is to have a similar following or presence to our canceled
artists. Adele is a strong female pop artist with 27.2 million followers on Twitter. Although
her career has been longer and her following is significantly greater than Doja Cat’s 4.2 million
followers, both artists hold a similar presence in the media through their music today. Both Doja
and Adele have achieved chart topper hits in this year, making them both equally relevant in the
current pop industry. In regards to Harry Styles and Zayn Malik, both have a similar number
of followers on Twitter (Harry Styles - 37.4 million and Zayn Malik - 31 million). In addition,
their industry presence is also identical since they started their careers together in boy group One
Direction. Moreover, Harry and Zayn were both equally influential and popular artists through
their independent music post One Direction, making them ideal complements.

5.2 Methodology for Tweet Collection

We used the Twitter API to scrape tweets from Twitter relating to each individual. In order
to observe the change in public sentiment over time, we decided to collect tweets six months
before and after the date of an artist’s cancellation. For our control artists, we used the date of
cancellation from their coressponding canceled counterpart to gather tweets for them. For each
artist, we generated a list of related hashtags and keywords that was used to collect tweets relating
to the artist and their controversy. For each tweet, we gathered information on the date the tweet
was created, author id, tweet id, and the text of the tweet itself. Below we have presented a table
displaying the hashtags and keywords we included for each query along with the date of cancellation
and the number of initial tweets we collated.
Table 1. Canceled Artists’ Dataset Overview

4
 Hashtags +
Initial Number of Keywords Used for
Artist Name Date of Cancellation Tweets Query
Lucas (NCT/WayV) 8/24/21 117,654 “lucas”,
“bubble”,“bbl”,
“scandal”, “xuxi”,
“yukhei”,
“nct”,“czennie”,
“czennies”, “nctzen”,
“lumis”,
“weishennie”,
“nctzens”,
“weishennies”,
“wayv”, “weishenv”,
“ot23”, “ot22”, “ot6”,
“ot7”, “#NCT”,
“#LUCAS”,
“#WayV”,
“WeiShenV”,
“#WELOVEY-
OULUCAS”,
“#LUCAS_OUT”,
“#WAYV_is_7”,
“#LUCASBEST-
BOY”, “#WayVis7”,
“#LU-
CAS_GETS_OUT_OF_NCT”,
“#LUCAS_GETS_OUT_OF_WayV”
Giselle (Aespa) 10/23/21 36,176 “giselle”, “uchinaga”,
“kpop”, “k-pop”,
“bbl”, “bubble”,
“MYs”, “MY”,
“aeri”, “n word”,
“racist”, “aespa”,
“#aespa”,
“#gisellenword”,
“giselleaespa”,
“#Giselle_OUT”,
“#gisellen-
wordspace”,
“#giselleapologize”,
“#aespa_is_FOUR”
DaBaby 7/25/21 100,810 “dababy”, #DaBaby-
isOverParty,
#DaBookings,
#DaApology,
#Dababy

5
 Hashtags +
Initial Number of Keywords Used for
Artist Name Date of Cancellation Tweets Query
Nicki Minaj 9/13/21 93,332 ’”nicki minaj,
#NickiMinaj,
#Nicki, #MyCousin-
TookTheVaccine,
#NickiMinajs-
CousinsFriendsBalls,
#NickisCousins-
FriendsBalls,
#nickiminajdrag-
gingparty,
#iStandWithNicki,
#istandwithnickimi-
naj, #barbz,
barbz
Zayn Malik 10/28/2021 53,472 “zayn malik”,
“zayn”, “zayn hits
yolanda”, “zayn and
gigi”, “zayn malik
and gigi hadid”,
“zayn cheated”,
#WeAreWithZayn,
#ZaynMalik,
#FreeZayn,
#welovezayn, #Al-
waysWithYouZayn,
#respectzayn,
#WeSupportZayn,
#cancelzayn,
#WeHateZayn
Doja Cat 5/25/2020 387,572 “doja cat”, “doja”,
“dindu nuffin song”,
#dojawasinnocent,
#dojacatisNOTover-
party,
“WeAreSorryDoja”,
#dojacatisoverparty,
#DojaIsOverParty,
#DojaCat

Table 2. Control Artists’ Dataset Overview

6
 Hashtags + Keywords Used
Artist Name Initial Number of Tweets for Query
Jaemin (NCT/NCT Dream) 117,654 “na jaemin”, “jaemin”,
“bbl”, “bubble”, “nct”, “nct
dream”, “nana”, “nctzens”,
“czennie”, “czennies”,
“nctzen”, “#NCT”,
“#NCTDREAM”
Ryujin (ITZY) 36,176 “ryujin”, “shin”, “itzy”,
“#ITZY”, “midzy”,
“midzys”, “#RYUJIN”
Lil Baby 146,198 “lil baby”, “#lilbaby”
Saweetie 70,904 ’“saweetie” , “#saweetie”
Harry Styles 437,775 “harry styles”, “harry”,
“styles”, #harry,
#harrystyles
Adele 34,282 “adele”, #adele, #adele25,
#adele21, #adelelove

6 Methods
6.1 Sentiment Analysis APIs
Initially, we looked at three sentiment analysis tools: Google Perspective, Vader, and TextBlob, in
order to analyze the sentiment in our tweet datasets. However, upon conducting our preliminary
analysis and EDA, we were not able to gather any meaningful results from Vader and Textblob
that were not already evident through our results with the Google Perspective API. Therefore, we
eventually decided to only use Google Perspective for our final analysis.
Using the Google Perspective API, we calculated the probability that each tweet in our datasets
would be perceived as being the following attributes: toxic, severely toxic, an insult, and containing
profanity. The difference between toxic and severely toxic is that the severe toxicity attribute is
less sensitive to milder forms of toxicity, such as text that contains profanity used in a positive
case. Since we are focusing on sentiment among English-speaking Twitter users, we specified that
we would be using the English language while using the Perspective API. Therefore, any tweets
that were not recognized as being primarily written in English by the API were excluded from
our final datasets. In choosing which attributes we would use for our analysis, we chose the
severe toxicity and insult probability attributes. Our preliminary results showed that toxicity and
severe toxicity showed similar trends. Insult and profanity also shared similar trends with each
other. Therefore, we narrowed it down to only using severe toxicity and insult probability scores to
minimize redundancy in our final results and because we believed that these two attributes would
be the best for analyzing change in toxicity over time.

6.2 RQ1: Type of Issue for Cancellation

The first research question we explored was how sentiment differed over time based on the type of
subject the artist is “canceled” for. The six case studies, we present, fall into three different cate-
gories: discrimination against marginalized groups, abuse allegations, and misinformation. These

7
categories were defined using our domain knowledge and research surrounding the detail of each con-
troversy. In the discrimination against marginalized class, we have Doja Cat, Giselle, and Dababy.
Both Doja Cat and Giselle received harsh criticism for singing terms which are frequently utilized
to degrade the African American community. Similarly, Dababy faced backlash for making homo-
phobic remarks which disrespected the LGBTQ+ community. Next, in the abuse allegation class,
we have Zayn Malik and Lucas. Zayn Malik made headlines for physically assaulting his mother-
in-law and Lucas faced allegations for emotional abuse from former alleged girlfriends. In the final
misinformation class, we have Nicki Minaj’s controversy for COVID-19 vaccine misinformation.
Table 3. Categories of Cancellation for Canceled Artists

Group Artist Names

Discrimination Towards Marginalized Groups Giselle, DaBaby, Doja Cat
Abuse Allegations Lucas, Zayn Malik
Misinformation Nicki Minaj

To test the impact the type of conflict has on public sentiment, we conduct two types of analysis:
qualitative and quantitative. For the qualitative analysis, we compared the words with the highest
frequency before and after the controversy. We started by grouping the data from the artists
in each category and compiled the text for every tweet in each group. To preprocess our text
data, we removed all urls and standardized the text by converting it to lowercase. In addition, we
removed any “stop words” from the tweets. Stop words are the common words, such as prepositions,
conjunctions, and pronouns, that do not add much information or context to a piece of writing.
After cleaning our text data, we were able to generate a table with the number of occurrences
for every word within each category. Based on these results, we will be able to determine which
category promoted more dialogue surrounding the conflict.
In regards to the quantitative analysis, for each category, we compared the severe toxicity and insult
levels before and after the date of the controversy. For the preprocessing of our quantitative data,
we initiated by removing any invalid severe toxicity and insult score that equaled 1000. A score of
1000 indicates that the tweet was not recognized by the API. We then converted all the values with
the ‘created_at’ column to a datetime object and eliminated tweets that did not fall within the
six months of the controversy date. Finally, we generated a ‘days_cancel’ column which presented
the difference in the number of days the tweet was created to the controversy. To proceed with our
analysis, we grouped the processed data by ‘days_cancel’ and computed the rolling average for the
severe toxicity and insult score. In order to recognize any trends, we plotted the final data on a
line plot which shows the change in sentiment for each category over time.

6.3 RQ2: Background of Musician

The background of the artists we researched was another aspect that we were interested in determin-
ing its influence on sentiment after cancellation. The differences between each artists’ background,
we believe, may influence the severity of sentiment, as well as the duration of sentiment during
canceling. In order to determine whether or not the background of an artist has an effect on the
sentiment they receive after being canceled, we created two different groups to look at. The first
grouping of artists is based on the genre of music they’re in. The genres that our artists fall into
are K-pop, hip-hop, and Western pop. The second grouping of artists is based on sex, comparing
female artists with male artists. The metric we used to compare artists in these groupings is de-

8
termining the number of days after cancellation it takes for the median sentiment level to return
to pre-controversy levels.
We dropped all of the control artists from these groups because the control artists weren’t necessary
for the analysis. We were only interested in how background affects the change in sentiment after
getting canceled.
After sorting the artists for the groups in each category, we took the sentiment indicators (severe
toxicity and insult levels) over time for each artist within the group and grouped by the number of
days before and after controversy. The aggregate function we used on the sentiment indicators is
median. In short, we calculated the median severe toxicity and insult levels for each group within
genre and sex over the entire period of tweets we collected.
Next, we plot out the grouped data for the categories to visualize the median indicator levels for
the groups against each other. In order to quantify the effect of the chart, we used the metric
of determining the number of days after controversy until the median sentiment indicator levels
returned to pre-controversy levels. We calculated this metric by taking the sentiment indicator
level at day 0 or the day of the controversy and finding the first instance of the indicator level being
below the day 0 level after the controversy has happened. For each group, we calculated this metric
and compared it to other groups within the same category to determine if there is a difference for
that background category.

6.4 RQ3: Parasocial Relationships

Parasocial relationships are one-sided relationships in which an individual develops illusions of
intimacy with another individual, usually a media personality, as a result of repeated interactions
between the two of them (Wikipedia). For example, a fan who has followed an artist for a period
of time and has learnt enough about them to develop a sense of loyalty might feel that they
have a personal connection with the artist. Depending on the strength of the relationship, media
personalities can have strong influence over their audiences as evident with fandoms or fan bases
like BTS’s ARMYs or Taylor Swift’s Swifties. To determine which artists have fan bases that
share either a strong or weak parasocial relationship with the artist, we used two metrics: total
mean engagement and fandom ratio. For the engagement metric, we collected every tweet an artist
posted 6 months before and after their cancellation date on their official Twitter account. We
then calculated the total mean engagement by taking the sum of the mean number of retweets,
quote tweets, likes, and replies across all artist’s tweets. To calculate the fandom ratio, we used
our previous tweet datasets gathered according to an artist’s related hashtags and keywords. Then
for each artist, we calculated the total amount of tweets that contained at least one of the artist’s
fandom names divided by the total number of tweets in that artist’s tweet dataset.
Table 4. Cancelled Artists’ Fandom Names

Artist Name Fandom Name

Lucas (NCT/WayV) “nctzen”, “nctzens”, “wayzennie”,
“wayzennies”, “wayzenni”, “weishennie”,
“czennie”, “czennies”, “#nctzen”,
“#weishennie”, “#nctzens”, “#weishennies”,
“lumi”, “lumis”
Giselle (Aespa) “MY”, “MYs”, “#MYs”, “aerishine”,
“aerishines”

9
 Artist Name Fandom Name
DaBaby NONE or N/A
Nicki Minaj “barb”, “barbz”, “barbs”, “#barbz”, “#barb”
Zayn Malik “directioners”, “directioner”, “zquad”,
“#zquad”
Doja Cat “kittenz”, “#kittenz”

Table 5. Control Artists’ Fandom Names

Artist Name Fandom Name

Jaemin (NCT/NCT Dream) “nctzen”, “nctzens”, “czennie”, “czennies”,
“#nctzen”, “#nctzens”,
“nanadoongie”,“nanadoongies”
Ryujin (ITZY) “midzy”, “midzys”, “#midzy”
Lil Baby NONE or N/A
Saweetie NONE or N/A
Harry Styles “directioners”, “directioner”, “harries”,
“#harries”
Adele “daydreamer”, “daydreamers”,
“#daydreamer”

We then grouped them into either strong or weak parasocial relationships based on whether the
artist reached above or below a certain threshold for each metric, in which our thresholds were
determined based on our metric values and our own domain knowledge. For the total mean en-
gagement, the cutoff threshold was 20,000 and for the fandom ratio, the cutoff threshold was 0.05.
To be categorized as having a strong parasocial relationship, the artist’s total mean engagement
and fandom ratio values would both have to be greater than or equal to its respective threshold.
On the other hand, those who are unable to reach above both thresholds are then categorized as
having a weak parasocial relationship. For example, Zayn Malik meets the threshold for total mean
engagement as ~360,000 > 50,000 but does not meet the threshold for the fandom ratio metric as
0.003 < 0.05, so he was categorized as having a weak parasocial relationship.
Table 6. Engagement Metric Values Used to Determine Parasocial Strength

Artist Name Total Mean Engagement Fandom Ratio Parasocial Strength

Lucas 112,107 0.123 Strong
Giselle 92,741 0.141 Strong
DaBaby 13,226 0.0 Weak
Nicki Minaj 81,179 0.052 Strong
Zayn Malik 359,275 0.003 Weak
Doja Cat 46,636 0.0001 Weak

10
7 Results
7.1 RQ1: Type of Issue for Cancellation
To investigate how sentiment changes over time based on the type of issue an artist was canceled
for, we preprocessed our discrimination, abuse, and misinformation grouped datasets by calculating
the mean severe toxicity and insult score for each date. We then applied a rolling median over an
interval of 14 days on the processed data to smoothen the visualization in order to recognize the
trend.

Figure 1: Most Common Words Before Figure 2: Most Common Words After
Cancellation For Discrimination Group Cancellation For Discrimination Group

The bar charts comparing the word frequency show that discrimination towards marginalized groups
(specifically towards the African American community) promotes more conservation or dialogue
surrounding the controversy in comparison to the abuse and misinformation category. This is evi-
dent since the “Most Common Words After Cancellation For Discrimination Issue Group” presents
“black” at the 7th most frequent word and the word ‘racist’ also appears within the top 50 most
common words post cancellation. However, both the abuse and misinformation graphs did not
display any common words related to the conflict in question following the controversy.

Figure 3: Change in Median Severe Toxicity Figure 4: Change in Median Insult Score By
Score By Type of Issue Type of Issue

11
Table 7. Median Severe Toxicity Score By Type of Issue

Misinformation Discrimination Abuse

Median Severe Toxicity Score Before Cancellation 0.1789 0.1486 0.1298
Median Severe Toxicity Score After Cancellation 0.1817 0.1717 0.1277

Table 8. Median Insult Score By Type of Issue

Misinformation Discrimination Abuse

Median Insult Score Before Cancellation 0.2327 0.1781 0.1728
Median Insult Score After Cancellation 0.2390 0.204 0.1712

When looking at how the subject of the controversy affects sentiment, the artists we present who
are canceled for discrimination towards marginalized groups promote higher negative sentiment
compared to our artists with issues relating to abuse and misinformation. All groups have an initial
spike immediately after the controversy. However, the discrimination category has a slightly large
spike in severe toxicity and two more spikes are seen around the 40 and 90 day mark. In addition,
when comparing the change in median severe toxicity score before and after the controversy, we
observe the following: +1.6% increase for misinformation, +15.6% increase for discrimination, and
-1.6% increase for abuse.
The trend is not as distinct when looking at the insult level plot, however, it does provide a
similar result. Once again, discrimination exhibits the largest spike after the controversy. Although
misinformation and discrimination follow a similar trend in insults levels post the controversy,
discrimination still displays a slight increase in negative sentiment as another spike is recognized
around the 100 day mark. Moreover, when comparing the change in median insult score before
and after the controversy, we observe the following: +2.7% increase for misinformation, +14.5%
increase for discrimination, and -0.93% increase for abuse.

7.2 RQ2: Background of Musician

For the second subquestion measuring the effect of background on sentiment, we grouped artists
into different genres and sex, and compared the differences in severity and length of sentiment by
taking the median sentiment level for each group.
Table 9. Days After Canceling Until Severe Toxicity Scores Returns To Pre-
Controversy Levels By Genre

Artist Name K-Pop Hip-Hop Pop

# Days to Return to Pre-Controversy Levels 26 35 18
Severe Toxicity Levels at Day 0 0.1577 0.1764 0.2245

Table 10. Days After Canceling Until Insult Scores Returns To Pre-Controversy Levels
By Genre

12
 Artist Name K-Pop Hip-Hop Pop
# Days to Return to Pre-Controversy Levels 18 30 1
Insult Levels at Day 0 0.1279 0.1449 0.1897

Figure 5: Genre Comparison for Change in Figure 6: Genre Comparison for Change in
Median Severe Toxicity Median Insults

Comparing different genres to each other, hip-hop stands out. For median severe toxicity levels, hip
hop artists took 30 days for the severe toxicity to return to the level it was at before controversy.
K-pop artists took 18 days and pop artists took 1 day (Table 9). For median insult levels, hip hop
artists took 35 days, K-pop artists took 26 days, and pop artists took 18 days to return to pre-
controversy insult levels (Table 10). From the data, hip-hop artists seem to undergo a prolonged
negative effect from cancellation compared to our other genres. Also, from our visualizations on
genre severe toxicity and insult levels, canceled hip hop artists experience a more severe change in
toxic sentiment compared to other genres as both plots have an extremely sharp peak in sentiment
levels. Median hip-hop severe toxicity levels go from 0.176 to around 0.300 (Figure 5) while median
hip-hop insult levels go from 0.145 to 0.220 immediately after controversy (Figure 6).
Table 11. Days After Canceling Until Severe Toxicity Scores Returns To Pre-
Controversy Levels By Sex

Artist Name Male Female

# Days to Return to Pre-Controversy Levels 71 18
Severe Toxicity Levels at Day 0 0.111 0.1942

Table 12. Days After Canceling Until Insult Scores Returns To Pre-Controversy Levels
By Sex

Artist Name Male Female

# Days to Return to Pre-Controversy Levels 71 13
Insult Levels at Day 0 0.1328 0.2436

13
 Figure 7: Sex Comparison for Change in Figure 8: Sex Comparison for Change in
Median Severe Toxicity Median Insult

Our results for grouping by sex show that canceled female artists experience a greater backlash in
response to controversy but for a much shorter period of time. Canceled male artists’ sentiment
levels returned to normal/pre-controversy levels 71 days after controversy while female artists took
13 days for severe toxicity and 18 days for insult levels to return to normal (Table 11, 12). Male
artists experienced a much longer duration of the effect from cancellation compared to female artists.
However, female artists sustain a greater level of negative sentiment for the entire time period of
tweets we collected and have a sharper and more drastic peak after being canceled (Figure 7, 8).
To conclude, genre and sex seem to make a difference in the length and severity of the sentiment
for canceled artists. Our canceled hip hop artists experienced longer and more drastic differences
in sentiment over the period of being canceled compared to k-pop and pop artists. Our male artists
who were canceled experienced the effects of being canceled for a much longer period of time than
female artists, but female artists experience a more drastic effect and have higher toxic sentiment
overall.

7.3 RQ3: Parasocial Relationships

To investigate how sentiment changes over time based on the strength of a canceled artist’s paraso-
cial relationship with their fans, we preprocessed our strong and weak grouped datasets by first
calculating the mean severe toxicity and insult probability score for each date. We then applied a
rolling median over an interval of 14 days on the transformed data to facilitate the observation of
long-term trends.

Figure 9: Change in Median Severe Toxicity Figure 10: Change in Median Insult Score
Score Over Time Over Time

Based on both figures, we can see that artists that have a strong parasocial relationship with their

14
fans tend to have reduced negative sentiment overall compared to artists with a weak parasocial
relationship. Figure 9 and Figure 10 both show a dramatic increase in median severe toxicity and
insult score directly after the artist’s cancellation. Both strong and weak parasocial groups also
experience an increase in negative sentiment after 100 days, which is possibly due to events such as
the controversy resurfacing. However, the weak parasocial group shows a more significant increase
that is comparable to its initial spike seen directly after cancellation.
Table 13: Median Toxicity Scores Before Cancellation By Parasocial Group

Strong Weak
Median Severe Toxicity Score 0.1323 0.1492
Median Insult Score 0.1658 0.1826

Table 14: Median Toxicity Scores After Cancellation By Parasocial Group

Strong Weak
Median Severe Toxicity Score 0.1388 0.1606
Median Insult Score 0.1744 0.2033

In terms of how sentiment has changed compared to before cancellation, the median severe toxicity
score increased by +4.5% for the strong group and +7.4% for the weak group after cancellation
(Table 13, 14). The median insult score increased by +5.5% for the strong group and +11.5%
for the weak group (Table 13, 14). Therefore, we can see that there is a slight percent increase
in severe toxicity and insult scores for both groups after cancellation. When comparing our weak
parasocial group against our strong parasocial group, there is also a slight positive difference in
percent increase with a +2.9% difference for severe toxicity and +6% for insults. This supports
that artists with a weak parasocial relationship experience slightly increased toxicity overall after
cancellation compared to those with a strong parasocial relationship.

8 Discussion
In analyzing how sentiment towards canceled musicians changes over time, we can conclude that
there is an overall increase in negative and toxic sentiment after cancellation. However, the duration
of such negative sentiment towards a canceled artist appears to be short-term.
Depending on the severity of the issue, we believe this could affect the intensity of negative sentiment
towards a canceled artist. Our research shows that artists canceled for discrimination against a
marginalized group tended to experience greater backlash compared to artists canceled for abuse or
misinformation. It was unexpected that artists canceled for abuse allegations did not demonstrate
higher levels of toxicity. However, due to the sensitivity of the topic as well as other factors such
as the strength of the artist’s parasocial relationship with their fans, it is possible these reasons are
why conversations about these individuals were less toxic than we expected.
Given that canceled artists’ backgrounds also could affect sentiment, we analyze how the genre
of music they cater to and their sex could affect sentiment after cancellation. We found that
hip-hop artists experienced a higher rise in negative sentiment and took the longest to return to

15
pre-controversy median severe toxicity and insult levels compared to K-pop and Western pop. The
huge spike in negative sentiment for hip-hop artists was unexpected, but is likely influenced by
issues they were canceled for, their target audience, and their general reputation in the eyes of the
public. In terms of sex, male artists took longer to return to pre-controversy median sentiment
levels, but female artists had a more drastic rise and a greater overall average in negative sentiment
after cancellation. This suggests that there might be a gender imbalance with Twitter cancel culture
in which an artist’s sex could affect the intensity of negative sentiment they feel towards a canceled
artist.
The strength of a canceled artists’ parasocial relationship can also affect sentiment as we found
that artists who have a strong parasocial relationship with their fans tend to experience less tox-
icity overall. Fans who have a strong parasocial relationship with an artist might feel a sense of
loyalty towards them. This suggests that if fans feel loyal towards an artist, they might be more
compelled to defend them and suppress as much negativity surrounding them as possible after the
artist’s cancellation. Especially considering how fans of WayV’s Lucas were mass tweeting hashtags
such as #WELOVEYOULUCAS or #LUCASBESTBOY as signs of support after his controversy
(Figure 12), it appears that the strength of an artist’s parasocial relationship with their fanbase
can significantly influence public sentiment surrounding them.
After conducting our research, there are several limitations that we would like to point out. First
of all, we only focused on sentiment among English-speaking Twitter users. Depending on the
language and the social media platform, there is the possibility that our findings could differ by
seeing more intense or reduced toxicity. Also, in categorizing the strength of an artist’s parasocial
relationship, the thresholds we chose for our metrics were dependent on the data we collected as
well as our domain knowledge. This would mean that using our thresholds to categorize other
artists might not accurately represent the true strength of their parasocial relationship with their
fans. Therefore, for future research, we would advise conducting network analysis on a canceled
artist’s Twitter followers to quantify the strength of their parasocial relationship with their fans.
In addition, since we only conducted case studies on six canceled individuals of varying follower
counts, it is possible that our findings may differ if we have chosen different artists or included
more canceled artists. Despite this, all of our canceled artists saw some sort of increase in toxicity
post cancellation, suggesting that cancel culture may lead to an overall rise in negative sentiment
over time. For future research, we would advise seeing if overall sentiment towards a canceled artist
is affected by whether they publicly apologized or not. We suspect that if an artist had publicly
apologized, they may show reduced toxicity and negative sentiment after cancellation compared to
artists who have not publicly apologized.
To summarize, it appears that there is a rise in toxicity towards canceled individuals over time.
However, there are various factors that seem to contribute to the variation in negative sentiment
after cancellation. Cancel culture on Twitter is still an evolving phenomenon that has only begun
to be recognized in recent years. Because of this and the many factors that can affect the amount of
toxicity directed towards canceled individuals, it is difficult to determine whether cancel culture is
effective in ostracizing them from society or to what extent would it affect their lives and reputations.
Regardless, the intense backlash they face in the short term demonstrates that Twitter users can
hold canceled individuals accountable for their actions to some extent. This could put pressure on
canceled individuals to publicly apologize and acknowledge their actions, which is already evident
with some of our case studies such as Aespa’s Giselle (Kim). Cancel culture can emphasize that
certain behavior is not acceptable in today’s socio-political climate by holding accountable canceled
individuals on a public stage. However, future research is needed to assess its effect on canceled

16
individuals’ reputations in the long-term and the extent to which cancel culture can go.

9 Conclusion
Our initial hypothesis was that online communities, specifically on the social media platform Twit-
ter, respond to an artist’s controversy by effectively “canceling” them. We defined “canceling”
as a rise in negative sentiment and toxicity towards an individual. From our research into cancel
culture, we determined that cancel culture does negatively impact sentiment towards music artists’
on Twitter. Through our analysis, we found different factors and aspects of artists can have an
effect on the severity and duration of toxicity received after cancellation. We identified these areas
of interest as the type of issue they were canceled for, the background of the artist, and the strength
of their parasocial relationships. To conclude, our investigation through our three sub-questions
shows that cancel culture overall can impact public opinion and sentiment towards controversial
artists on Twitter. We captured the effect of cancel culture before and after artists’ controversies,
but long term effects like complete ostracization from mainstream popular culture or continued
toxicity did not seem to be evident as sentiment eventually returned to normal within the time
frame we looked at. Cancel culture is a topic that is heavily debated, but we believe that our
analysis has provided some context to understand and quantify cancel culture as a whole.

10 References
• Associated Press. “Following Verdict, Will R. Kelly’s Music be Canceled?.” Billboard, Bill-
board, 28 Sep. 2021, https://www.billboard.com/music/music-news/r-kelly-verdict-music-
canceled-9637759/.
• Chan, Anna. “A Timeline of DaBaby’s Homophobic Comments Controversy.” Billboard, Bill-
board, 2 Mar. 2022, https://www.billboard.com/photos/dababy-rolling-loud-homophobic-
comments-controversy-timeline-9608086/.
• D. Clark, Meredith. “DRAG THEM: A Brief Etymology of so-Called ‘Cancel Cul-
ture.’ ” Communication and the Public, vol. 5, no. 3–4, Sept. 2020, pp. 88–92,
doi:10.1177/2057047320961562.
• Fiesler, C., J. Jiang, J. McCann, K. Frye, and J. Brubaker. “Reddit Rules!
Characterizing an Ecosystem of Governance”. Proceedings of the International
AAAI Conference on Web and Social Media, vol. 12, no. 1, June 2018,
https://ojs.aaai.org/index.php/ICWSM/article/view/15033.
• France, Lisa Respers. “Doja Cat Denies Taking Part in Racist Conversations.” CNN, Ca-
ble News Network, 25 May 2020, https://www.cnn.com/2020/05/25/entertainment/doja-cat-
racial-slurs-trnd/index.html.
• Jhaver, Shagun, et al. “Evaluating the Effectiveness of Deplatforming as a Moderation Strat-
egy on Twitter.” Proc. ACM Hum.-Comput. Interact, 5, CSCW2, Article 381, Oct. 2021,
https://doi.org/10.1145/3479525.
• Kim, U. “Aespa’s Giselle Apologizes for Mouthing a Racial Slur from a Song.”
Soompi, Soompi, 25 Oct. 2021, https://www.soompi.com/article/1495215wpp/aespas-giselle-
apologizes-for-mouthing-a-racial-slur-from-a-song.
• Kim, U. “NCT’s Lucas Apologizes for Past Behavior Involving Ex-Girlfriends
+ SM Announces He Will Halt Activities.” Soompi, Soompi, 25 Aug. 2021,
https://www.soompi.com/article/1485476wpp/ncts-lucas-apologizes-for-past-behavior-
involving-ex-girlfriends-sm-announces-he-will-halt-activities.

17
 • Lamarre, Carl. “What Happens to Travis Scott Now?.” Billboard, Billboard,
11 Nov. 2021, https://www.billboard.com/music/music-news/travis-scott-astroworld-what-
happens-next-9657656.
• “Parasocial Interaction.” Wikipedia, Wikimedia Foundation, 9 Feb. 2022,
https://en.wikipedia.org/wiki/Parasocial_interaction.
• Pham, Jason. “Here’s the Alleged Reason Zayn’s Fight with Gigi’s Mom Started-She Took a
‘Hit at His Ego’.” StyleCaster, StyleCaster, 5 Nov. 2021, https://stylecaster.com/why-zayn-
malik-yolanda-hadid-fight/.
• Romano, Aja. “Cancel Culture Why We Can’t Stop Fighting About.” Vox,
Vox, 30 Dec. 2019, http://courses.bowdoin.edu/sociology-1101-spring-2020/wp-
content/uploads/sites/319/2020/05/What-is-cancel-culture_-Why-we-keep-fighting-about-
canceling-people.-Vox.pdf.
• Romano, Aja. “Nicki Minaj Isn’t Anti-Vax, Exactly. That’s Why Her Vaccine Resistance
Is so Concerning.” Vox, Vox, 14 Sept. 2021, https://www.vox.com/22673528/nicki-minaj-
vaccine-tweets-cousins-friend-met-gala.
• Velasco, Joseph. “You Are Cancelled: Virtual Collective Con-
sciousness.” Rupkatha Journal, Vol. 12, No. 5, 2020,
https://pdfs.semanticscholar.org/dd2e/c36189e588a491cff61a0fba26114c6a5ada.pdf.

11 Appendix
11.1 Figures of Most Common Words For Assault and Misinformation Groups

Figure 11: Most Common Words Before Figure 12: Most Common Words After
Cancellation For Assault Allegation Group Cancellation For Assault Allegation Group

18
Figure 13: Most Common Words Before Figure 14: Most Common Words After
Cancellation For Misinformation Group Cancellation For Misinformation Group

19