According to the latest data on suicide by the World Health Organization [1], nearly 800,000

people die every year due to suicide, meaning one person dies every 90 seconds. Suicide can

urgent issue following the death of a British girl, Molly Russel, whose father, Ian Russel, stated

in an interview with BBC that Instagram encouraged his daughter to commit suicide [2]. In

2017, Molly Russel, who was known to have been posting and searching for keywords related to

suicide and self-harm, such as “cutting,” “biting,” and “burning,” ended her own life. The posts

that she “liked” were identified to be images that glorified suicide. This prompted the discussion

on the need for an advanced platform policy to prevent such incidents from happening again.

Furthermore, there have been several incidents of self-expressive YouTubers ending their

own lives. Jamey Rodemeyer, a 14-year-old YouTuber who actively expressed himself through

posts. He appeared to be strong as he shared videos about the “It Gets Better” project, which

aimed to address prevention of teen suicide. His suicide was attributed to excessive hostile

comments. This case shows that social media sites are becoming venues to share personal

opinions and to express oneself—even painful thoughts [3]—but at the same time, are making

it easier for cyber-bullies to target their victims [4].

The influence of social media on concerning behaviors is not limited to children and teen-

agers alone. The debate lies in how media function—whether as a facilitator or as an intervenor

of such behaviors. Considering the debate, this study aims to examine how deliverers of sui-

cide-themed contents discuss suicide and to examine what factors, among content provider

characteristics, story characteristics, and content expression characteristics, predict viewer

Suicide is defined as a “conscious act of self-induced annihilation” [5: p. 203] in the current

Western society. A review of comparable concepts suggests that society has historically con-

demned the act of killing oneself. Synonyms of suicide, such as self-killing, self-disembodiment,

and self-murder, have shared stigmatizing connotations. This is because individuals are the con-

stituents of society, where the sanity of one represents the degree of social health and well-being

of the society as a whole. Suicidal thoughts and behaviors have also been considered pathologi-

cal in the context of religion or morality. For example, the Protestants attribute melancholic

self-disintegration to the temptation of Satan or a diabolical entity, which is distinguished from

the inner self [6]. The concept evolved in the eighteenth century, encompassing terms from vol-

untary death to involuntary self-killing. Since the term “suicide” indicates a change in attitude,

Suicide pertains to not only the self but also society, as society is regarded as an aggregate of

many selves. There have been longstanding concerns over the diffusive nature of suicide. The

diffusion process involves successful or unsuccessful suicide attempts that lead to serious sui-

cidal ideations among others, and some of those contemplators make successful or unsuccessful

attempts [9,10]. The diffusion of suicide in relation to the influence of media was studied follow-

ing the widespread imitation of Werther’s suicide, as described in the novel The Sorrows of the

and how the media augment or intervene in the diffusion of suicidal thoughts and behaviors.

Studies of the potential influence of media-publicized suicide stories of actual suicide have

behaviors is complex. Social media can be hazardous to the vulnerable, as some online com-

munities advocate extreme beliefs and behaviors, such as anorexia, suicide, and deliberate

amputation, which are otherwise considered socially unacceptable [12]. Online discussion

forums and social media chatrooms may facilitate socially undesirable behaviors as a result of

peer pressure [13].

Recent studies that aimed to replicate and extend Phillips’ imitation theorem suggest that

widely publicized suicide stories trigger copycat suicides [10,11]. News or television coverage

of suicide stories may provide role models for individuals at risk, which is related to a social

learning theory of deviant behavior [14]. From this perspective, publicized suicide stories may

encourage suicides in the real world, which makes it imperative to develop guidelines on how

a character in Mozart’s opera, The Magic Flute. Papageno becomes suicidal upon the loss of his

beloved Papagena; however, he refrains from committing suicide thanks to a hopeful song by

three elves. Media intervention effect suggests that media has a preventive function through

[12].

The advancement of media has enabled speedy diffusion of information without bound-

aries. The current study aimed to analyze suicide-themed content in a dominant visual media

service platform, considering its reach and potential influence on the users.

with different technological affordances. They provide an array of features including profile-

works among them are given the power; users have richer and more complex experiences; con-

tent distribution is not limited to content creators alone; and the boundaries of the devices are

blurred [20]. With higher bandwidth, faster and more interactive experiences have been real-

ized, providing users with rich visual media content such as audio and video streaming.

According to YouTube Press, over 1.9 billion logged-in users visit YouTube monthly, which

account for nearly one-third of the Internet users worldwide. YouTube provides multi-lingual

experiences with a total of 80 different languages, covering about 95% of the entire Internet

population. As it features a variety of video contents, YouTube, as a media-sharing website

diverse range of information at a rapid rate across online networks has not always been dis-

cussed from a positive perspective. Continuous efforts have been made by multiple SNSs to

develop an auto-filtering system to screen for and remove unsafe content. Following the inci-

dent of the live streaming of the New Zealand terror attack by Branton Tarrant, Facebook offi-

cially announced that they would adopt artificial intelligence (AI) technology to automatically

filter harmful information. Twitter announced its plan to filter hate speech and spam tweets,

while Tumbler, an image-based microblogging SNS, censors pornographic or illegal adult con-

tent. All these measures have been enforced, considering the influence of such mainstream

platforms with a large user base.

disinhibited [21,22] when the message is presented with visual effects. Extant studies show that

the inclusion of visual material in a message facilitates longer memory retention [23], more

accurate comprehension of the message [24], greater likelihood of reacting to a call to action

ton, which brings the post to YouTube’s immediate attention, according to Andrea Faville, a

spokesperson for YouTube [28]. The stance of the platform is that “the users should not be

afraid to speak openly about the topics of mental health or self-harm,” and the platform pro-

vides community guidelines, according to which content “promoting or glorifying suicide,

providing instructions on how to self-harm, graphic images of self-harm posted to shock or

disgust viewers” is banned [27]. The platform applies the same policies across all products and

features, such as video posts, content descriptions, comments, and live streams. In instances of

violation, the content is removed, and the creator is sent an email if it is their first time violat-

ing the policy. If it is not the first time, the creator is given a strike; three strikes result in chan-

nel termination.

and become so-called “influencers,” who are considered micro-celebrities. New media scholar

David Marshall [31] identified a transition from “representational” to “presentational” media

and culture. In the age of social media and self-created content, the public self, public-private

self, and transgressive intimate self are presented. This develops into the establishment of trust,

credibility, and a sense of closeness between the creators and audience. The boundary between

legacy media and new media as sources of information has become less meaningful, as the

consumers of media have begun to identify these influencers as new information providers.

Many studies suggest that self-disclosure and honesty tend to increase online when partici-

pants’ identities are hidden. Joinson’s study [34] shows that the visual anonymity in computer-

of content deliverers. There are multiple groups of content uploaders, also known as channel

operators, who are distinct from message deliverers. The group of content uploaders are listed

as clinic and health organizations, news agencies, one-person creators, production organiza-

tions, educational facilities, religious groups, and others, while the group of message deliverers

are listed as survivors of suicide attempt; family members of the deceased; friends; news per-

sonnel; rescuers; third-party narrators; artists, musicians, and film personnel; lecturers and

educators; medical professionals; one-person creators; and others. Some of the message deliv-

erers provide real names, whereas the others are anonymous. Considering the aforementioned

characteristics of the content deliverer, the following hypotheses are proposed.

H1. The characteristics of the deliverer, that is, the one who delivers the suicide-themed

Media contagion effect has been examined from multiple perspectives. A meta-analysis of sui-

type (i.e. newspapers vs. television) [11]. Studies that are based on newspapers compared to

television (TV, 82% less likely), studies that include suicide stories of political/entertainment

celebrity (14.3 times), studies based on real suicides (4.03 times) as opposed to fictional sui-

cides in films and soap operas, and studies based on suicide attempts as an outcome measure

as opposed to completed suicide rates or counts are more apt to investigate copycat effects

[11]. In addition, media influence on suicide has been studied in multiple country-settings.

Regardless of the small effect size compared to other psychosocial risk factors for suicide,

media contagion shows that not only audience characteristics but also media content involve

risk [36,37].

Since the current study did not consider the difference in viewer engagement among differ-

suicide ideation (see Table 2). Considering the aforementioned characteristics of the suicide-

themed content, the following hypotheses are proposed.

H2. The characteristics of the story characteristics, that is, the type of messages delivered,

nity to educate the public about suicide,” “avoid language which sensationalizes or normalizes

suicide, or presents it as a solution to problems,” “avoid prominent placement and undue repe-

trigger suicides. The guidelines are commonly used as educational material for journalists and

editors of traditional media agencies and were developed for traditional news reports of suicide

rather than online news or social media posts. Thus, reporters are advised to refrain from using

customize existing guidelines based on the new media message or channel features.

Reflecting the reporting guidelines for suicide stories, the current analysis categorized con-

tent expression characteristics into five groups: existence of advertisement, expression of

hotlines were listed in the description, the first-half of the video clip, and in the second-half of

the video clip. YouTube platform provides a special warning function. Only those who have

clicked on the “I understand and wish to proceed” option after being shown the YouTube

community warning for inappropriate or offensive content warranting viewer discretion are

allowed to view the content. The genre categories were listed as Entertainment, People &

Blogs, News & Politics, Music, Film & Animation, Nonprofits & Activism, and Education (see

Table 3). Considering the aforementioned characteristics of expressive methods, the following

hypotheses are proposed.

H3. The characteristics of content expression, that is, how the message is delivered, would

determine the degree of viewer engagement.

The current study employed a quantitative content analysis method to identify the factors that

The preliminary analysis employed a bottom-up grounded theory approach [40], which is a

qualitative method, to examine the characteristics of suicide-themed videos. The sample

word search results with the term “suicide” in English on YouTube. The keyword search was

done in Seoul, South Korea at one point in time, September 2019, by the researchers using

Incognito Window on Google Chrome browser. The search result was ranked by the default

method that YouTube provides which is ‘relevance.’ Neither specific inclusion nor exclusion

criteria was set in the sample selection process with the purpose of extracting all possible codes

and anonymity. The characteristics of the stories included four factors: whether the story

involves a public or non-public figure, whether the story is real or fictional, whether the story

is about a suicide attempt, completed suicide, or suicide ideation, and the number of suicide

stories. The characteristics of story expressions were observed using 7 factors: the existence of

advertisement, illustration of method, existence of a warning sign, placement of the warning

sign, existence of hotlines, placement of hotlines, and genre category. Every newly observed

item for each factor was recorded. The list of the items for each factor was built upon after sev-

eral iterative processes until saturation had been reached. The iterative process continued until

only redundant instances were observed and until no new codes occurred to the degree in

which the researchers have agreed that further data collection or data coding is counter-pro-

search results from search engines, YouTube has no clear distinction in terms of pages. After

several videos, a swipe up motion leads to the loading of more videos. Considering that the

number of videos presented before the first swipe up motion was 20, it can be regarded that a

related issue of suicide, and added 7 other videos to make 100. Videos categorized as ‘music’ or

‘film’ were not related to ‘Suicide Squad’ but they were videos created by individuals who

express suicide-related information through the form of music or film.

The sample size (n = 100) was chosen because the purpose of the research was to reflect a

basic query of what general people would likely to be exposed to with the keyword search. The

first several pages of search result presented to the person searching the keyword engage most

viewers whereas the following pages are less attended [42,43]. Multiple health information-

related studies included the first several pages of search results in the sample, implying that

people tend to select the information they are provided with first rather than the information

they are provided later [44–49]. The first two to three pages were the most commonly observed

a public or non-public figure, whether the story is real or fictional, whether the story is about a

suicide attempt, completed suicide, or suicide ideation, and the number of suicide stories; and

(4) characteristics of story expressions, which included the existence of advertisement, illustra-

tion of method, existence of a warning sign, placement of the warning sign, existence of hot-

lines, placement of hotlines, and genre category. A total of 14 variables were examined. The

characteristics of content deliverer, stories, and expressions were dummy coded.

Furthermore, YouTube’s user engagement metrics, including the view count, the number

of likes, and the number of comments of the selected 100 videos, were retrieved using You-

Tube Statistics, which is a free application that tracks the statistics for YouTube videos [52]. It

is assumed that the degree of viewer engagement increases in the following order: view count,

categorical factors of suicide-themed video contents and the amount of attention or

popularity.

Hierarchical multiple regression is a method that considers the relative effect of more than

viewer engagement. The analysis was completed on IBM SPSS (Statistical Package for the

Social Sciences) Statistics software [53].

themed videos. Among 14 variables, 5 variables including message deliverer category, whether

the story is about a suicide attempt, completed suicide, or suicide ideation, illustration of

The results of the hierarchical multiple regression are illustrated in Table 4. Deliverer char-

acteristics were the only predictor that was significant across all three viewer responses (p =

.036, adjusted R