Stance-Taking in Bilibili Bullet-Screen Comments: How Chinese Youth Express Ideological Positions Online

https://doi.org/10.65281/660554

Author:First author:Liu Ru , Ph.D.graduate, School of Marxism, Qufu Normal University, the research direction of youth ideological and political,Rizhao 273165,Shandong,China.Email:LR2027@qfnu.edu.cn

Second author: Huimin Zhang, Female, Master Candidate, School of Marxism, Changchun University of Chinese Medicine, Changchun, Jilin 130000 China

*Corresponding author:Jingjie Pang，associate professor，School of Marxism，Anhui Medical University, Master degree in Ideological and Political Education, specializes in ideological and political education and medical humanities. Email：pangjingjie@ahmu.edu.cn

Acknowledgments:

Research Funds: 1. “Institutionalization of Free Targeted Medical Education in Rural Areas: An Empirical Study on the Four-Dimensional Synergy Model for Ideological and Belief Education” (Project ID: 2023AH050533), a 2023 Key Scientific Research Project in Anhui Province’s Higher Education Sector (Philosophy and Social Sciences);

2. “Empowering Rural Free Targeted Medical Students’ Labor Education with New Quality Productivity: A ‘Four Beauties’ Practice Education Model” (Project ID: 2024jyxm0796), a 2024 Key Teaching and Research Project under Anhui Province’s Quality Engineering Initiative.

ABSTRACT

Bilibili’s bullet-screen comments (弹幕) show the viewpoint of young Chinese users about ideology. In this paper, a total of 15847danmu from the popular video-sharing siteBilibili over the time of 2022 – 2024 is analyzed from the viewpoint of DuBois’s Stance Triangle(2007) and Appraisal theory(Martin & White, 2005). We found out there are 4 main strategies, when it comes to what kind of words young people use: the same emotions, which we also call affective alignment; how they talk about their feelings as young people; showing you are in a group, with “you all” or by speaking to each other; and sarcasm and irony. In the data, affective expression rather than explicitly ideological statements predominates as the patriotism content, and an ironic, humorous stance-taking is more commonly used for social commentary. These are indicative of a type of ideological speech act which is both playfully and seriously engaged with, through the real-time, ephemeral nature of danmu. We are saying that the way you take the stand is following some bigger changes on how younger people say something on the internet.

Keywords: stance-taking; danmu; youth discourse; Bilibili; appraisal; pragmatics

1. INTRODUCTION

Anyone who’s seen videos on Bilibili knows how it works: text rolling past the screen and people reacting in the moment (or at least feeling as though they are reacting in the moment) to the same parts. This bullet-screen comment system, called danmu is now central to how young Chinese viewers watch and engage with videos. with more than 336 million monthly active users and close to 86% of them under 35 years old (Bilibili 2024), the bilibili platform turns into a meeting point for youth culture, new forms of speech and ideology.

What we are interested in here are the stances that users take in these brief comments. The audience who type out ‘破防了'(defense broken)after seeing a documentary on the veterans or ‘YYDS’after seeing an Olympic event is not simply leaving a comment, but also positioning themselves and evaluating, as well as aligning or disaligning with the others. This is stance-taking in Du Bois’s (2007) sense, a public act of evaluation, positioning, and calibration with respect to others.

However, the study of stance-taking in Chinese digital contexts is still patchy. The most work has been on Weibo or WeChat, platforms with other affordances and users (Herring, Stein, and Virtanen 2013 for CMC pragmatics more generally). And Danmu is quite different: comments are fleeting, over the top of what’s visible, and create this strange sensation of being in the same place, even if it has been days or years. How people will show what their thoughts are about certain things.

In terms of the current study of stance-taking of Bilibili danmu is mainly concentrated on how young users cope with ideological land. We have asked the three questions:

•What strategies are used by the users when taking positions on ideological matters in danmu?

•How do these differ across content type (patriotic, social commentary, cultural).

• How much do youth specific linguistic resources play a part in this process?

Let us note from the outset what this study is unable to achieve. We can’t know what a person actually believes or intends—only the public performances of those beliefs and intentions that they give in discourse. And we cannot claim representativeness to all Chinese youth, the users of Bilibili tend to be young, well-educated, and from cities. We can offer a close-up of language use in a corpus of 15,847 comments using the theories of stance and appraisal.

2. BACKGROUND AND PREVIOUS RESEARCH

2.1 Stance in Interaction

There is a long history of discussing stance in linguistics, but Du Bois’s (2007) is especially productive. He argues that stances are constituted of three things going on at the same time: an evaluation (of some object, along some value dimension), a positioning (of oneself as a kind of social actor), and an alignment (with other people). Three of them make his so called “stance triangle”

That is what captures some that a more simple view of attitudes, opinions and so do not. Positions are dialogical by definition. When a Bilibili user posts “这格局太大了”(“this vision is too big”), he or she does not only appreciate the video content, but also claim the identity of a person who can recognize and value such qualities, and they are implicitly aligned with (or invite other viewers to align with) people who share such appreciation.

Martin and White’s (2005) appraisal theory can be used to complement the means by which stance is encoded linguistically. Their framework distinguishes three systems: the affective one (attitude), the judgmental one (judgment), and the appreciative one (appreciation), on the one hand, and how speakers engage dialogically (engagement) in the other. And finally, they also talk about graduation, which is about how evaluative meanings get scaled up and down. It has been used with different genres, but not so much on CMC in Chinese.

2.2 Pragmatic Identity in Discourse

Chen Xinren (2014, 2018)’s works on pragmatic identity could serve as another valuable perspective. Identities for Chen are not attributes but are the resources speakers create and use to interact. Pragmatic identity is “contextualized” and “language-user-chosen”, it arises from discourse and serves communicative purposes (Chen 2018: 35).

This helps to explain how it is possible for the same person to show up very differently in different contexts – earnest, patriotic comments on one video, ironic jokes on another. These are not contradictions, but different pragmatic identities which were constructed for different purposes. What identities are constructed by danmu users and how?

Recently, there are some researches on identity in Chinese digital space. Zhu, Feng, and Chen (2024) are useful collections on social identity in Chinese digital communication with some attention to danmu. In 2025, Yang et al. study pragmatic identity construction on Bilibili danmu among adolescents and find that the participants affiliate with online communities in diverse ways. Our study follows up on this work, but focusing specifically on the taking of ideological stances.

2.3 Danmu as a Genre

Danmu was originally from Japan(on Niconico) and then adopted by Chinese platforms in early 2010s. Johnson (2013) explains this “polyphonic” quality of such comments: many voices overlapped, forming a cacophony that somehow became a shared viewing experience.

What makes danmu a different CMC genre? Several are striking. First, commenters can be temporally linked with specific moments of the video, thereby forming pseudo-synchronous interaction between different viewers with different viewing times (Wu and Ito 2022). Second, they are visually ephemeral, scrolling by and disappearing, so they should be brief and repetitive. Thirdly, overlay format makes comments a part of the viewing experience rather than something separate from it.

These affordances make for certain kinds of discourse. Dong (2023) posits that danmu engenders “light communities” (轻共同体), or provisional, low-commitment affiliations formed around shared linguistic practice rather than sustained social bond. Users may take part in a collective experience even if they do not engage for the entire duration.

Bilibili has documented the progress of these through its yearly top danmu expression reports, “囍” (2017), “真实” (2018), “AWSL” (2019), “爷青回” (2020), “破防了” (2021), “优雅” (2022), “啊？” (2023). These buzzwords form a rough index of changing youth sentiment – but it is important to be careful when reading too much into corporate PR as sociolinguistic data.

2.4 Youth, Language and Ideology Online

The intersection of youth culture, internet language, and ideology in China has garnered a great deal of attention, not all of it academic. Fang and Repnikova (2018) show that “Little Pink” itself was a contested label that different groups used for different ends. Nakahara and Cai (2025) discovered that nationalism on Bilibili interacts with feminism, fandom, and platform conditions in intricate ways.

Linguistic research reveals the word-formation of internet neologisms such as homophony, abbreviation, code-mixing, semantic extension(Wang&Liu 2022; Chang 2024). Less attention has been paid to how these resources function in stance-taking – this is where we hope our research will be able to add.

3. ANALYTICAL FRAMEWORK

we combine three perspectives:

Du Bois’ stance triangle is the larger form. We look at how danmu comments make evaluations (what is evaluated and how), position (what position the subject is placed in), and align (how the subject’s position is aligned with others’).

Martin and White’s system provides ways of analyzing the language as it occurs. Attitude subsystem.

• Affect: emotions (being moved, amused, angered)

• Judgment: assessments of behavior against social norms

Appreciation: Evaluations of things, performances, or ideas

Chen’s pragmatic identity theory helps us consider the stance-taking strategy. Users build their identities through their word choices – as fans, critics, patriots, cultural insiders, etc.

One limitation of the framework must be admitted. They were developed for mostly face-to-face interactions, or longer written texts. Danmu is short-lived and brief, some categories will be hard to bear. We treat it as a heuristic rather than a coding scheme.

4. DATA AND METHODS

4.1 Corpus Construction

We have collected a total of 15,847 bullet screen comments from 48 Bilibili videos over the period of January 2022 to December 2024. Videos were chosen to stand for three broad content categories:

• Patriotic / nationalistic (16 videos) – Chinese history, military, national achievements

• Social commentary (16): social commentaries, current affairs, youth life challenges

Cultural content (16 videos): Traditional Chinese culture, Contemporary Youth Culture, Entertainment

The selection criteria include: a total of over 100,000 views; over 5,000 danmu comments; and being distributed over the three years. We use the public API that bilibili provides, and we can get comments texts and some metadata.

Table 1: Corpus overview

Category	Videos	Comments	% of corpus
Patriotic	16	6,234	39.3
Social commentary	16	5,128	32.4
Cultural	16	4,485	28.3
Total	48	15,847	100.0

The imbalance reflects different commenting densities across content types—patriotic content tends to generate more danmu per video.

4.2 Analytical Procedure

The analysis was done in stages. First, we processed the corpus with python(jieba for cutting) and find high frequency item. and compared with the list of Internet neologisms in published sources and Bilibili’s yearly danmu report.

Second, we used an iterative process of pilot coding to develop a coding scheme. Two coders coded 500 comments separately, had discussions on disagreements, and refined the categories. Final interrater reliability, κ = 0.84. Included Categories:

• Stance strategy (affective, evaluative – neologism, group identity, implicit/ironic

• Appraisal type (affect, judgment, appreciation)

• Alignment markers (inclusive pronouns, address terms, shared knowledge claims)

Third, we chose examples to analyze quantitatively, and to study how stance-taking works in context.

4.3 Limitations

Several caveats. Our corpus is not small, but it is still a portion of the Bilibili content — different video types or different times may result in different results. We can’t verify user demogs or tell real sentiment from perf or trolling. coding scheme is bound to simplify the complicated pragmatic phenomenon. We are showing the results to be suggested instead of being definite.

5. FINDINGS

Four stance-taking strategies were found from the analysis: (1) affective alignment, (2) evaluation with neologisms, (3) group identity marking, (4) stance through humor/irony. We discuss each one and then look at distribution by content category.

5.1 Affective Alignment

The most frequent approach included sharing emotions. Users posted responses that were at the same time expressing emotion and asking for agreement from others.

Table 2: Frequent affective expressions

Expression	Count	Gloss	Function
破防了	1,247	‘defense broken’	being overwhelmed
泪目	892	‘tears in eyes’	being moved
爷青回	756	‘my youth returns’	nostalgia
DNA动了	634	‘DNA moved’	deep resonance
暖心	521	‘heartwarming’	positive feeling
笑死	487	‘dying laughing’	amusement

Consider this example from a documentary on veteran reunions:

(1) 破防了，这才是真正的英雄

“defense breached- these are the true heroes”

The user expresses affect (feeling overwhelmed emotionally), makes a judgment (veterans are “real heroes”), and implicitly positions themself as someone who values patriotic sacrifice. The comment assumes and invites alignment: other viewers who feel the same way are invited to see themselves in this response.

“DNA动了”(“my DNA moved”) is interesting, because it links personal feeling or affect to some kind of essence that is ethnic or cultural.

(2) DNA动了，这就是文化自信

My DNA moved – this is cultural confidence.

Here, the hyperbolic expression of affect is connected to the official discourse on “cultural confidence” (文化自信). it links a person’s personal feeling to the greater group idea.

We should also note that these affect expressions are not always sincere—the users could be deploying them strategically, performatively, or ironically. But their frequency indicates that they are useful stances in the limited space of danmu.

5.2 Evaluation through Youth Neologisms

The second strategy uses internet slang, neologisms for evaluation indexing youth identity.

Table 3: Evaluative neologisms

Expression	Count	Source	Function
YYDS	1,089	pinyin initials (永远的神)	supreme praise
绝绝子	654	suffix innovation	strong positive
格局	578	semantic extension	praising vision
拿捏	423	semantic extension	praising skill
无语子	312	suffix innovation	negative judgment
离谱	287	semantic extension	negative evaluation

“YYDS” (永远的神, ‘eternal god’) exemplifies pinyin-initial abbreviation:

(3) 中国队YYDS！！！

    ‘Chinese team YYDS!!!’

The choice of YYDS over the likes of 太棒了great) does more than judge-it stakes a claim to membership in youth online culture. And those who are older and the outsiders may not know it exists at all.

“格局”原意是“图案”，后来用来赞扬人胸襟开阔。

(4) 这格局太大了

‘this vision is too big’

The evaluation is positive, but we may not always have been able to distinguish the ironic use of this in context.

5.3 Group Identity Construction

Users actively create group boundaries using pronouns, address terms, and identity claims.

Table 4: Group markers

Expression	Count	Function
我们 ‘we’	2,134	inclusive reference
家人们 ‘family’	876	intimate address
兄弟们 ‘brothers’	445	solidarity
同志们 ‘comrades’	234	political solidarity

Inclusive “我们” constructs an imagined community:

(5) 这是我们的历史，不能忘

This is our history – can’t forget

who is “我们”? The comment makes it unclear whether it refers to Chinese people. Bilibili users. who are watching this video? The ambiguity could be good, so that different people feel included.

“家人们”(‘family members’) imports intimacy of address into the anonymous online.

(6) 家人们懂的都懂

family members — those who know, know

This brings intimate address together with an exclusionary knowledge claim. Understand: become a member of.

5.4 Implicit Stance through Humor and Irony

The fourth strategy conveys stance indirectly. Homophonic substitution is one resource:

Table 5: Indirect expressions

Expression	Standard form	Function
蓝瘦香菇	难受想哭	humorous sadness
杯具	悲剧	ironic tragedy
神马	什么	playful questioning

(7) 看完房价，人生杯具

‘After seeing the price of houses, life is a cup of [tragedy]’

The homophone softens criticism with wordplay. Is this a real lament or an ironic one? Maybe both? The humor provides distance, but there’s still a position taken on housing affordability.

and emoji can indicate irony:

(8) 好好好 [doge emoji]

    ‘Good good good [doge]’

The “doge” emoji (used sarcastically online) is an inversion. This affords a critical stance-taking with plausible deniability..

5.5 Distribution across Content Types

Strategy use varied by content category:

Table 6: Strategy distribution (%)

Strategy	Patriotic	Social	Cultural	Overall
Affective	42.3	28.6	35.7	35.9
Neologism evaluation	31.2	33.4	29.8	31.3
Group identity	18.7	22.1	19.4	19.9
Implicit/ironic	7.8	15.9	15.1	12.9

Affective alignment dominated patriotic content—expressions like “破防了” and “DNA动了” appeared frequently. Social commentary showed more implicit and ironic stance-taking (15.9% vs. 7.8% for patriotic), possibly reflecting the greater sensitivity of social criticism.

6. DISCUSSION

6.1 Stance-Taking as Identity Work

These patterns suggest that the data follows a certain trend or behavior. First, danmu users are not simply expressing their pre-existing attitudes but creating identities by staking out a position. You type the YYDS, instead of “very impressive”, is not just evaluation of something, but the claim of one’s own culture, and belongs to which generation.

and is compatible with Chen’s (2018) idea of the pragmatic identity being emergent and purposive. Users select available resources to form a position suitable for the moment. The same person can make serious patriotic remarks on one video and make a joke on another, it is a different identity performance, not a contradiction.

6.2 Affect and Ideology

The widespread affective stances-takings, especially about patriotic contents, are well worth paying attention to. Users express their ideological stances through displaying emotion, not by giving arguments. To be “破防” by patriotism content makes the position appear to be a feeling and not a political statement.

This can be used for a number of things. It might cut down on pushback—emotional responses seem less like propaganda than ideology. It creates possibilities for a collective experience, shared affect makes the group stick together. And it offers safety – feeling expression is less risky than belief statement.

Whether this is true feeling, strategic performance, or some of each is hard to tell from the words alone. Probably all three are true to some degree.

6.3 The Danmu Environment

The medium forms the message. Danmu is brief, thus there are many “破防了” and “YYDS”, which are formulaic; danmu is ephemeral, hence there are a lot of repeatable danmu, knowing that users can swipe down. It has a pseudo-synchronicity that creates a shared viewing even if other commenters are from days or months ago.

And these affordances might help us understand how people stand on things in danmu differently than in other online spaces. The format rewards short recognisable expressions that get many jobs done at once – it’s evaluating, identifying oneself, inviting allies, all in a few characters.

6.4 Heterogeneity and Complexity

We wish to oppose overgeneralizing. We see differences in our data – different strategies are more dominant in certain types of content, and the same expression can behave differently in different contexts. Chinese youth are not a homogenous group, and there is no single online nationalism (or stance).

The greater use of ironic stance-taking on social commentary compared to other topics points to a modulation according to sensitivity. This is consistent with other research on how Chinese internet users navigate censorship and express criticism indirectly (Yang 2016).

6.5 Limitations Revisited

We should say again what this study can’t tell us. We looked at the discourse patterns, not the beliefs or intentions. We can’t tell sincere from performative stance-taking, genuine from ironic affect. Our corpus is specific types of content on one platform for a period of time. Different sampling will probably result in different patterns.

And analytical limitations. Coding necessarily becomes simpler—some comments can be several categories, and irony is famously difficult to find in writing. We have sought, rather than to paper over our difficulties.

7. CONCLUSION

This paper examines how young Chinese people take stances in Bilibili bullet-screen comments. We found four main strategies: a) Affective alignment; b) Evaluation via neologisms; c) Group identity construction; d) Implicit stances via humor/irony. These strategies spread out unevenly by content type, emotional expressions show up the most in patriot stuff and ironic stances happen a lot in social comments.

Based on our study results, it appears danmu constructs an atmosphere of ideological discourse that is brief and follows certain formulas, and could be called “affective indirect.” In this context, users present ideological positions mainly in emotional expressions, using language specific to their generation, instead of direct debate. This mode of staking out a position is both playfully and earnestly so, both together and not.

Some questions are still unanswered. How does it compare to other platforms or other countries with a bullet-screen system? how did they change over the years? Users themselves can only understand their actions based on their own understanding. these will call for different ones—comparative analysis, longitudinal data, user interviews—that go past corpus pragmatics.

Yet what we can say is that the danmu enables a productive site of stance-making online. And because it constrains us in a certain way, this format can show us ways of getting at those things, and seeing the ways is also how bigger questions about language, identity, ideology of the Internet come to the foreground.

REFERENCES

AHMED, Sara. 2014. The Cultural Politics of Emotion (2nd ed.). Edinburgh: Edinburgh University Press.

BILIBILI. 2024. Bilibili Annual Report 2024. Shanghai: Bilibili Inc.

CHANG, Muge. 2024. Sociolinguistics of Chinese internet slang: Viral buzzwords (2019–2023). In Y. Wang (ed.), Digital Communication in Chinese Contexts, 234-256. London: Routledge.

CHEN, Xinren 陈新仁. 2014. Yuyongxue shijiao xia shenfen yanjiu de xianzhuang yu qianjing 语用学视角下身份研究的现状与前景 (Current research on identity from the pragmatic perspective). Xiandai Waiyu 现代外语 37(5): 702-710.

CHEN, Xinren 陈新仁. 2018. Yuyong Shenfen Lun 语用身份论 (Pragmatic Identity: How to Do Things with Words of Identity). Beijing: Beijing Normal University Press.

DONG, Jie 董洁. 2023. Xinmeiti yuyan de gongtongti jiangou yu shenfen rentong 新媒体语言的共同体构建与身份认同 (Community construction and identity in new media language). Yuyan Zhengce yu Guihua Yanjiu 语言政策与规划研究 8(2): 45-62.

DU BOIS, John W. 2007. The stance triangle. In R. Englebretson (ed.), Stancetaking in Discourse: Subjectivity, Evaluation, Interaction, 139-182. Amsterdam: John Benjamins.

FANG, Kecheng, and Maria REPNIKOVA. 2018. Demystifying “Little Pink”: The creation and evolution of a gendered label for nationalistic activists in China. New Media & Society 20(7): 2162-2185.

HERRING, Susan C., Dieter STEIN, and Tuija VIRTANEN (eds.). 2013. Pragmatics of Computer-Mediated Communication. Berlin: De Gruyter Mouton.

JOHNSON, Derek. 2013. Polyphonic/pseudo-synchronic: Animated writing in the comment feed of Nicovideo. Japanese Studies 33(3): 297-313.

MARTIN, James R., and Peter R. R. WHITE. 2005. The Language of Evaluation: Appraisal in English. Basingstoke: Palgrave Macmillan.

NAKAHARA, Junki, and Jing CAI. 2025. User-generated “feminism” on BiliBili: Nationalism, Internet culture, and platform dynamics in digital China. New Media & Society 27(2): 234-256.

WANG, Xiaoming, and Yan LIU. 2022. Chinese internet neologisms: Formation, spread, and social functions. Language in Society 51(3): 423-445.

WU, Tong, and Mizuko ITO. 2022. Time, space, and sociality in synchronous and asynchronous danmu communication. Journal of Computer-Mediated Communication 27(2): 1-18.

YANG, Chunsheng. 2024. Chinese Sociolinguistics: Language and Identity in Greater China. Cambridge: Cambridge University Press.

YANG, Guobin. 2016. The Red Guard Generation and Political Activism in China. New York: Columbia University Press.

YANG, Shuang, Ling CHEN, and Mei LIU. 2025. “High energy ahead!”: Exploring Chinese adolescents’ pragmatic identities in bullet-screen discourse. Humanities and Social Sciences Communications 12(1): 1-14.

ZHU, Hongqiang, Debing FENG, and Xinren CHEN (eds.). 2024. Social Identity and Discourses in Chinese Digital Communication. London: Routledge.

APPENDIX: CODING CATEGORIES

Category	Subcategory	Description	Examples
Affective	Positive	Positive emotion display	破防了, 泪目, 暖心
	Negative	Negative emotion display	无语, 难受
	Nostalgia	Nostalgic feeling	爷青回, 童年回忆
Evaluative neologism	Positive	Youth slang for praise	YYDS, 绝绝子, 格局
	Negative	Youth slang for criticism	离谱, 无语子
Group identity	Pronouns	Inclusive we/us	我们, 咱们
	Address	Intimate address terms	家人们, 兄弟们
	Category	Identity labels	中国人, XX粉
Implicit	Irony	Non-literal evaluation	好好好+狗头
	Homophone	Sound-based substitution	杯具, 蓝瘦香菇
	Humor	Playful expression	笑死, 哈哈哈哈