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 use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
The new reality of cybersuicide raises challenges to ideologies about the traditional form of suicide that does not involve the internet (offline suicide), which may lead to changes in audience’s attitudes. However, knowledge on whether stigmatizing attitudes differ between cybersuicides and offline suicides remains limited.
This study aims to consider livestreamed suicide as a typical representative of cybersuicide and use social media data (Sina Weibo) to investigate the differences in stigmatizing attitudes across cybersuicides and offline suicides in terms of attitude types and linguistic characteristics.
A total of 4393 cybersuicide-related and 2843 offline suicide-related Weibo posts were collected and analyzed. First, human coders were recruited and trained to perform a content analysis on the collected posts to determine whether each of them reflected stigma. Second, a text analysis tool was used to automatically extract a number of psycholinguistic features from each post. Subsequently, based on the selected features, a series of classification models were constructed for different purposes: differentiating the general stigma of cybersuicide from that of offline suicide and differentiating the negative stereotypes of cybersuicide from that of offline suicide.
In terms of attitude types, cybersuicide was observed to carry more stigma than offline suicide (
The way people perceive cybersuicide differs from how they perceive offline suicide. The results of this study have implications for reducing the stigma against suicide.
Suicide remains one of the leading causes of death worldwide according to the latest estimates released by the World Health Organization [
The internet facilitates self-disclosure and social connection, giving rise to an emerging form of suicide (ie, cybersuicide). Unlike the traditional form of suicide that does not involve the internet (ie, offline suicide), cybersuicide covers a broad range of internet-mediated suicidal behaviors and phenomena, including livestreamed suicide [
The livestreamed suicide is commonly considered as one of the most notable and representative types of cybersuicide, particularly in China [
To address these concerns by analyzing social media data (Sina Weibo, a Chinese social media site that is similar to Twitter), this study attempts to directly and systematically investigate the differences in stigmatizing attitudes across cybersuicides and offline suicides in terms of attitude types and linguistic characteristics, respectively.
It is worth noting that cybersuicide is a new and developing form of suicide. The public may not be equally familiar with different types of cybersuicide. In China, because of the prevalence and media coverage of livestreamed suicide, compared with other types of cybersuicide, the public is expected to be more familiar with and more likely to discuss livestreamed suicide on social media. Therefore, to collect sufficient social media data for further analysis, this study aims to consider livestreamed suicide as a typical representative of cybersuicide and compare it with offline suicide.
The research process included the following three steps: (1) data collection, (2) data preprocessing, and (3) data analysis. The data collection and preprocessing procedures are shown in
Procedures of data collection and preprocessing.
First, a participant pool of active Weibo users was created. According to a previous study, 1,953,485 active Weibo users were identified as potential participants [
Second, a database of Weibo posts was constructed. In May 2020, using API, a vast amount of publicly available Weibo posts from 1.06 million active users in the participant pool since their registration (the 2020 official numbers of monthly and daily active users: 511 million and 224 million, respectively [
Third, several relevant Weibo posts were identified through database searches. To obtain the posts reflecting attitudes toward cybersuicides and offline suicides, two sets of keywords were used to search the database, including
The process of database searches included the following three steps: (1) a total of 4460 posts with keywords
After data collection, preprocessing was performed on the raw data to prepare them for further analysis.
First, to exclude irrelevant posts and reclassify misclassified posts, manual scrutiny of the collected data was conducted.
In this study, irrelevant posts were considered as (1) posts that depicted suicides in fictional works (eg, movies, television programs, and novels), (2) posts that focused on suicides in nonhuman animals (eg, dogs), and (3) posts that used suicide-related keywords for nonsuicidal purposes (eg, making a bet). After the removal of irrelevant posts, 7244 posts (cybersuicide: 4460 – 136 = 4324 and offline suicide: 4500 − 1580 = 2920) remained.
Subsequently, 77 posts in the offline suicide group were reclassified as posts related to cybersuicide (livestreamed suicide: n=69, 90% posts; suicide
Therefore, the final sample of this study included 7236 posts (cybersuicide: 4324 + 69 = 4393; offline suicide: 2920 – 69 – 8 = 2843). The demographic characteristics of the participants in the final sample are presented in
Demographics of participants.
Demographics | All Weibo posts (N=7236), n (%) | Cybersuicide (n=4393), n (%) | Offline suicide (n=2843), n (%) | |
|
||||
|
Male | 4062 (56.14) | 2473 (56.29) | 1589 (55.89) |
|
Female | 3174 (43.86) | 1920 (43.71) | 1254 (44.11) |
|
||||
|
North China | 1312 (18.13) | 812 (18.48) | 500 (17.59) |
|
Northeast China | 299 (4.13) | 191 (4.35) | 108 (3.8) |
|
East China | 2277 (31.47) | 1330 (30.28) | 947 (33.31) |
|
Central China | 374 (5.17) | 233 (5.3) | 141 (4.96) |
|
South China | 1345 (18.59) | 801 (18.23) | 544 (19.13) |
|
Southwest China | 493 (6.81) | 330 (7.51) | 163 (5.73) |
|
Northwest China | 224 (3.1) | 133 (3.03) | 91 (3.2) |
|
International and unspecified | 912 (12.6) | 563 (12.82) | 349 (12.28) |
Second, to extract psycholinguistic features from each of the 7236 posts automatically, the Simplified Chinese version of Linguistic Inquiry and Word Count software was used. The Simplified Chinese version of Linguistic Inquiry and Word Count is a reliable and valid text analysis tool for the automatic estimation of word frequency in multiple psychologically meaningful categories, including linguistic processes (eg, personal pronouns), psychological processes (eg, affective processes), personal concerns (eg, achievement), spoken categories (eg, assent), and punctuation categories (eg, periods) [
To explore the differences in types of stigmatizing attitudes across cybersuicides and offline suicides, a content analysis was performed on all 7236 posts to determine whether each of them reflected stigma. The coding framework was developed based on expert consensus and available evidence. Specifically, in this study, a researcher (AL) reviewed relevant studies [
To explore the linguistic differences in stigmatizing expressions across cybersuicides and offline suicides, 2 groups of classification models were built using Waikato Environment for Knowledge Analysis (version 3.9.4; University of Waikato) software. Waikato Environment for Knowledge Analysis provides tools for developing machine learning techniques and applying them to practical data mining problems.
The first group of classification models was built to investigate whether linguistic differences existed in the expression of stigma in general (ie, cybersuicide-related or offline suicide-related stigma as a whole) between cybersuicides and offline suicides. The human coding results were considered as the ground truth for the validation of the classification models. In this study, an imbalanced data problem existed. For example, the number of stigmatizing posts belonging to the offline suicide class (minority class: 588 posts) was obviously lower than those belonging to the cybersuicide class (majority class: 1556 posts). Imbalanced data sets pose a challenge for machine learning modeling, as this problem may result in models with poor predictive performance, especially for the minority class. To handle this problem, using simple random sampling, a certain number of posts were randomly selected from the majority class to obtain a well-balanced data set. Subsequently, to improve classification accuracy, a subset of psycholinguistic features was selected for use in model construction. Specifically, a series of 2-tailed independent
It is worth noting that the good classification performance of models in the first group may be attributed to the existence of differences in the amount and distribution of negative stereotypes across cybersuicides and offline suicides rather than the existence of linguistic differences in stigmatizing expressions across cybersuicides and offline suicides. To clarify this issue, the second group of classification models was built to investigate whether linguistic differences existed in the expression of certain negative stereotypes across cybersuicides and offline suicides. To obtain sufficient data for further analysis, in this study, posts reflecting two negative stereotypes (ie,
The study protocol was reviewed and approved by the institutional review board of the Institute of Psychology, Chinese Academy of Sciences (protocol number: H15009). Informed consent was not obtained, as this study was based on publicly available data and involved no personally identifiable data collection or analysis.
The Cohen
Results of human coding (N=7236).
Categories | Cybersuicide, n (%) | Offline suicide, n (%) | |||
|
4393 (100) | 2843 (100) | |||
|
Stigmatizing | 1556 (35.4) | 588 (20.7) | ||
|
Nonstigmatizing | 2837 (64.6) | 2255 (79.3) | ||
|
1556 (100) | 588 (100) | |||
|
Weak and pathetic | 114 (7.3) | 80 (13.6) | ||
|
Self-centered | 97 (6.2) | 44 (7.5) | ||
|
Stupid and shallow | 528 (33.9) | 129 (21.9) | ||
|
False representation | 387 (24.9) | 13 (2.2) | ||
|
Glorified and normalized | 148 (9.5) | 195 (33.2) | ||
|
Immoral | 111 (7.1) | 69 (11.7) | ||
|
Strange | 59 (3.8) | 14 (2.4) | ||
|
Embarrassing | 24 (1.5) | 7 (1.2) | ||
|
Vengeful | 40 (2.6) | 9 (1.5) | ||
|
Mad | 48 (3.1) | 28 (4.8) |
For stigma in general, posts on cybersuicide were more likely than posts on offline suicide to contain stigmatizing expressions (
For negative stereotypes, posts on cybersuicide were often coded as
For exploring linguistic differences in the expression of stigma in general between cybersuicides and offline suicides, to achieve a balanced data set, 600 stigmatizing posts were randomly selected from posts in cybersuicide group (cybersuicide: 600 posts and offline suicide: 588 posts). A total of 6 key features were selected for use in the model construction (Table S2 in
Performance of classification models.
Models | Stigma in general | Stupid and shallow | Glorified and normalized | |
|
||||
|
Precision | 0.73 | 0.72 | 0.72 |
|
Recall | 0.73 | 0.72 | 0.72 |
|
0.73 | 0.72 | 0.72 | |
|
||||
|
Precision | 0.83 | 0.80 | 0.79 |
|
Recall | 0.83 | 0.80 | 0.78 |
|
0.83 | 0.79 | 0.78 | |
|
||||
|
Precision | 0.85 | 0.84 | 0.78 |
|
Recall | 0.85 | 0.83 | 0.77 |
|
0.85 | 0.83 | 0.77 | |
|
||||
|
Precision | 0.86 | 0.84 | 0.81 |
|
Recall | 0.85 | 0.84 | 0.81 |
|
0.85 | 0.84 | 0.81 |
For exploring linguistic differences in the expression of
To explore the linguistic differences in the expression of
To our knowledge, this study provides the first systematic analysis of the differences in stigmatizing attitudes toward cybersuicides and offline suicides. The results of this study have implications for reducing stigma against suicide.
First, it is necessary to confront and reduce stigma against cybersuicide. In this study, a large proportion of cybersuicide-related and offline suicide-related posts were coded as
Second, the public reacts differently to cybersuicides and offline suicides. In terms of attitude types, in this study, cybersuicide was observed to carry more stigma than offline suicide (cybersuicide: 1556/4393, 35.42% and offline suicide: 588/2843, 20.68%; χ21=179.8;
Apart from the differences in attitude types, linguistic differences in the expression of stigma between cybersuicides and offline suicides also exist. Such differences existed not only at the level of stigma in general but also at the level of negative stereotypes. In this study, the
Third, the use of linguistic analysis methods can facilitate the identification of suicide-related stigma in mass media. Mass media is a major contributor to the dissemination of incorrect information, which may reinforce negative stereotypes surrounding mental illness [
This study has some limitations. First, this study mainly focused on the stigma against livestreamed suicide. Therefore, it is unclear whether these findings are applicable to other types of cybersuicide. Second, social media users are not representative of all people in China. Therefore, the findings may not be applicable to nonusers. Third, the API of Sina Weibo only allowed us to download posts from a certain number of registered users. Therefore, these findings should be further confirmed on a larger scale and in more diverse populations in the future. Fourth, because of the lack of posts obtained from people with suicidal intentions, this study cannot analyze the stigma that people with suicidal intentions put on themselves (ie, self-stigma). Fifth, because of the lack of information about user types (eg, celebrities and general users), this study cannot investigate the differences in attitudes across different types of users and cannot compare attitude responses elicited by the deaths by suicide of different types of users.
This study used a nonintrusive method to directly and systematically examine the differences in stigmatizing attitudes toward cybersuicides and offline suicides. The results of this study support the conclusion that the way people perceive cybersuicide is very different from the way people perceive offline suicide and offer insight into the reduction of stigma against suicide.
Supplementary Tables S1 and S2.
application programming interface
Central China
East China
North China
random forest
South China
Southwest China
None declared.