Efficiency Tool or Academic Threat? Mapping China’s AI Controversy in Higher Education with Mixed Methods

Efficiency Tool or Academic Threat? Mapping China’s AI Controversy in Higher Education with Mixed Methods | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Efficiency Tool or Academic Threat? Mapping China’s AI Controversy in Higher Education with Mixed Methods Mingyi Yang, Qi Song, Minghui Zou This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7171423/v1 This work is licensed under a CC BY 4.0 License Status: Posted Version 1 posted You are reading this latest preprint version Abstract Global higher education is experiencing dual pressures from rapid rise of generative AI tools such as ChatGPT, which drive pedagogical innovation while triggering global concerns over academic integrity and skill degradation. This study addresses a key research gap by investigating how China’s distinct context, marked by government influence, market dynamics, and achievement pressure, influences conflicting understandings of AI in academia. Employing a mixed-methods approach combining Latent Dirichlet Allocation (LDA) topic modeling and Social Construction of Technology (SCOT) framework-guided content analysis, discourse dynamics among key social groups was mapped: producers, advocates, users, and bystanders. Thematic modeling identified five key areas of controversy: AI-enabled academic practices, ethical concerns, efficiency evaluations, policy debates, and the perceived inevitability of AI adoption. Content analysis revealed clear differences in perspectives among key groups: producers highlighted efficiency benefits in their communications; advocates stressed academic integrity risks and endorsed regulatory measures; users explained their reliance on AI as necessary due to academic pressures and curricular dissatisfaction; while bystanders expressed diverse but often practical views on AI’s broader societal impact. Crucially, our analysis indicates that China’s closure mechanism will likely be predominantly policy-driven, advocating tiered governance and pedagogical reforms to mitigate structural pressures fueling student dependency. This research contributes a non-Western perspective to SCOT theory, revealing how administrative-market dynamics shape technology negotiation. Furthermore, it proposes practical implementation pathways for adaptable AI governance frameworks applicable across diverse global higher education contexts.​ Social science/Education Business and commerce/Information systems and information technology Social science/Politics and international relations Social science/Science technology and society Generative AI Chinese higher education LDA topic modeling Social Construction of Technology Figures Figure 1 1. Introduction Generative AI is profoundly reshaping the global higher education landscape. Tools such as ChatGPT drive teaching paradigm shifts through adaptive content generation (Kasneci et al., 2023), while simultaneously triggering concerns over academic integrity and skill degradation worldwide. Notably, distinct national cultural contexts have fostered differentiated governance paths. In Western higher education systems, AI governance typically emphasizes ethical oversight and multi-stakeholder participation. For example, the European Union’s Artificial Intelligence Act (Regulation (EU) 2024/1689) classifies AI application in educational assessments as high-risk (Annex III), mandating transparency disclosures, human oversight, and GDPR-compliant data governance (Article 13–14). In parallel, U.S. universities seek to balance innovation with risk by cultivating comprehensive AI literacy among faculty and students. Stanford University (2025), for instance, introduced its AI Literacy Guide , which outlines a multidimensional framework encompassing functional, ethical, rhetorical, and pedagogical competencies. This encourages technological exploration (e.g., customized AI tool development) while managing ethical risks (e.g., copyright issues, data bias), teaching instructors and students to use AI autonomously and wisely. Such governance logic reflects the Western tradition of academic autonomy and positions AI as a tamable instrumental entity, emphasizing that technology should serve human agency. China’s AI integration reveals a distinct sociotechnical dynamic. In the Chinese context, AI application in higher education is shaped by a triad of administrative intervention, market momentum, and meritocratic anxiety. AI producers employ marketing rhetoric like “super study tool” to promote efficiency worship, while university administrators impose bans due to academic ethics crises, exemplified by Fudan University (2024)’s Regulations on the use of AI tools in undergraduate theses (trial implementation). This contradiction highlights the cognitive tensions distinctive to East Asian education. Technology is simultaneously sanctified as a competitive symbol that fuels public anxiety about “not using AI meaning obsolescence”, and alienated as a defensive target due to deep vigilance against skill replacement. When the same technology carries opposing meanings as both an “efficiency tool” and an “academic threat,” how can we decode its conflicting nature and reach a stable consensus?​​ The social construction of technology (SCOT) framework offers key insights to this question. It posits that technological meaning is not fixed, but rather emergent, from negotiation practices among diverse groups (Pinch & Bijker, 1984). This framework is especially valuable for unpacking how meanings of AI are contested in China’s higher education system, where bureaucratic governance coexists with intense market pressure and competition. However, existing research on relevant social groups’ views about AI in higher education primarily focuses on Western platforms such as X (Feng et al., 2023). Consequently ,critical gaps remain in understanding how Chinese stakeholders negotiate the meanings, risks, and boundaries of generative AI technologies. It also fails to address how closure mechanisms unique to the Chinese context reshape technological trajectories through provisional consensus. This Western-centric paradigm systematically obscures China’s distinctive phenomena of technological construction.​​ This study seeks to address the abovementioned gap by analyzing negotiation dynamics among relevant social groups on Chinese social media. Drawing upon data collected from a Chinese social media platform, Weibo, we examined how contests over technological meaning led to temporary closure. A three-phase mixed methods approach was adopted. First, Laten Dirichlet Allocation (LDA) topic modeling was used to extract key themes from Weibo discussions. Then, directed content coding grounded in the SCOT framework was conducted to categorize producers, advocates, users, and bystanders, while parallel textual analysis examined semantic patterns within these same discussion threads. Finally, we traced the emergence of provisional closure mechanisms by examining how institutional policy interventions resolve disputes, while concurrently assessing their broader social implications. This bottom-up empirical approach moves beyond Western-centric viewpoints while provides actionable insights for adaptive governance with Chinese characteristics. 2. Literature Review 2.1. The Social Construction of Technology (SCOT) The Social Construction of Technology (SCOT), proposed by Pinch and Bijker (1984) within the academic context of critiquing Technological Determinism, fundamentally challenges the notion that technological development follows an inherent, autonomous logic that unidirectionally determines social change (MacKenzie & Wajcman, 1992), as exemplified by McLuhan’s (1964) famous assertion that “the medium is the message”. Rooted in constructivist epistemology, the SCOT framework posits that technological forms emerge as products of negotiations among social groups, highlighting the contingent and multi-directional nature of developmental pathways (Bijker, 1997). This theoretical stance inherits the “Strong Programme” from the Sociology of Scientific Knowledge (Bloor, 1976), asserting that technology is inherently non-neutral, with its design, application, and meanings inextricably shaped by power relations, cultural values, and social conflicts (Winner, 1986). Overall, the SCOT theoretical system comprises four core analytical dimensions: Social Groups, Interpretive Flexibility, Closure, and Stabilization. These concepts systematically reveal the dynamic processes among multiple actors in technological evolution. Within the SCOT framework, social groups are defined as relevant actors with direct/indirect technological connections who share meanings attached to a specific artifact and participate in technological construction. Categorized by their technological relationships and organizational forms, four primary groups emerge: Producers (e.g., engineers, designers, investors, marketers), Advocates (e.g., policymakers, researchers, lobbyists), Users, and some scholars later refer to as Bystanders (Pinch & Bijker, 1984). To begin with, producers shape technological forms directly through design, production and marketing. For example, Dunlop’s pneumatic tire design simultaneously enhanced cycling comfort and speed performance (Clayton, 2002). Beyond direct production dynamics, advocates exert indirect yet systemic influence on technological configurations through regulatory frameworks. A paradigmatic example emerges from New York State’s legislative prohibition against handheld phone operation while driving. This policy intervention effectively institutionalized earphones as mandatory technical prerequisites for mobile device usage in vehicles, fundamentally reconfiguring both material design conventions and sociocultural perceptions of responsible communication practices (Humphreys, 2005a). Completing the triad of relevant social groups, users represent the experiential pole of technological interaction. Users are individuals or groups who interact directly with technological artifacts through practice, assigning diverse meanings to it based on contextualized use. Unlike producers or advocates, users are not formally organized but engage in emergent meaning-making through appropriation and adaptation. Their interpretations are shaped by personal needs, cultural norms, and situational demands, which collectively determine the artifact’s social legitimacy (Pinch & Bijker, 1984; Humphreys, 2005a). While users actively engage with technologies through situated practices, bystanders, though not directly interacting with a technology, nevertheless shape its social acceptance and cultural meaning through moral judgments, public discourse, or normative critiques (Pinch & Bijker, 1984). Unlike users or producers, bystanders lack material or institutional stakes in the technology itself but influence its trajectory by constructing societal perceptions of its appropriateness, risks, or symbolic value, as seen in 19th-century anti-cycling movements that condemned female riders for violating gender norms, forcing advertisers to redefine bicycles’ gendered symbolism (Garvey, 1996). In this sense, these social groups do not operate in isolation, rather, their dynamic relationships manifest through continuous interactions involving language, usage, and technical structures. On the one hand, systematic power struggles between producers and advocates occur. The former manipulate physical structures through design and capital, while the latter establish normative frameworks via policy discourse (Humphreys, 2005b). These interactions drive technological adaptations between institutional constraints and market imperatives, potentially fostering mandatory standards or alternative solutions. On the other hand, user-bystander interactions form dynamic tension of meaning production: embodied practices generate diverse interpretations, while moral judgments construct social regulations, collectively reshaping technologies’ legitimacy boundaries. Interpretive flexibility, SCOT’s core analytical tool, reveals how opposing meanings are inscribed onto a single technological artifact. A classic illustration emerges from the historical evolution of bicycles: young male cyclists viewed bicycles as competitive racing apparatuses, while women/elderly users prioritized safety, driving technical iterations from high-wheelers to safety bicycles (Pinch & Bijker, 1984). Mobile phone controversies similarly demonstrate the dynamic: users emphasize communication convenience while bystanders denounce “cell yell” as a public nuisance. These conflicting interpretations of technology led to design adjustments like vibration modes and text messaging, demonstrating how conflicting user needs shape practical solutions (Humphreys, 2005a). Extending this analysis, MacKay and Gillespie (1992) theorize technological “openness”—the degree to which systems permit user reinterpretation—as a critical mediator of interpretive flexibility. Closed designs (e.g., locked software) maintain power imbalances by restricting customization, whereas open designs (e.g., customizable hardware) let users adapt technologies against standard practices. Closure and stabilization constitute interconnected yet distinct processes in technological evolution. Closure occurs when social groups agree on a technology’s purpose, often through discussions, persuasive strategies, or problem redefinition. Following moments of closure, technologies enter processes of stabilization, whereby provisional agreements solidify into taken-for-granted norms. It unfolds when designs become standardized and widely accepted as “natural” parts of technological systems (Humphreys, 2005b). While closure reflects momentary social compromises, stabilization develops through repeated agreements over time. SCOT’s conceptualization of closure undergoes crucial paradigm shifts in the internet age. Early research framed closure as a final resolution through stable consensus (Pinch & Bijker, 1984)—a premise rooted in the relative stability of industrial-era technologies. In the context of digital technologies, Bijker’s (1993) notion of “degrees of stabilization” reconceptualizes closure as a series of temporary compromises under specific socio-temporal conditions, dynamically adjusting with power relations. Humphreys (2005b) advances “temporary closure” to explicitly manifest SCOT’s dynamism, arguing that technological closure in networked societies inherently features locality and recursivity. Each instance of closure simultaneously functions as the resolution of preceding negotiations and the genesis of subsequent controversies. This theoretical shift deconstructs linear evolutionary myths, refocusing analytical attention from static solutions to dynamic power negotiations. In doing so, it marks SCOT’s critical transition from a tool of technological historiography tool to a research paradigm for analyzing digital societies. 2.2. Generative AI in Higher Education The integration of generative AI technologies like ChatGPT into higher education has accelerated amidst global digital transformation and post-pandemic reforms in teaching and learning. As universities increasingly adopt hybrid learning models, these tools demonstrate multifaceted applications in assessment systems, pedagogical interactions, and technical skill development. Specifically, universities have begun employing generative AI to enhance various institutional processes—creating dynamic quizzes (Adiguzel et al., 2023), simulating academic writing scenarios (Kasneci et al., 2023), and providing real-time feedback on student work (Zhao et al., 2023). In programming education, AI-driven code explanation systems (MacNeil et al., 2022) and data science training modules (Moore et al., 2022) demonstrate practical implementations. These applications extend beyond pedagogical functions to contribute to accessibility improvements, for example, by assisting students with disabilities through adaptive interfaces and multilingual support (Kasneci et al., 2023). The adoption of generative AI presents transformative opportunities for personalized learning. ChatGPT’s capacity to generate tailored educational content aligns with evolving pedagogical trends toward individualized instruction, enabling educators to optimize content delivery based on learner analytics (Zhu et al., 2018). Its ability to overcome language barriers and democratize access to quality education for non-native speakers and underserved populations marks a significant advancement (Kasneci et al., 2023). Technologically, AI-assisted problem decomposition in data science (Moore et al., 2022) and automated grading systems reduce administrative burdens, allowing instructors to focus on higher-order teaching objectives. Social media sentiment analyses further reveal public optimism about AI’s potential to enhance educational equity and innovation (Li et al., 2023). Despite its potential, generative AI introduces systemic risks to academic integrity and pedagogical quality. The technology’s ability to produce graduate-level exam responses (Dwivedi et al., 2023) and even pass standardized law exams (Koetsier, 2023) raises fundamental concerns about the validity of traditional assessment frameworks, plagiarism and credential devaluation (AlAfnan et al., 2023; Sok & Heng, 2023). Pedagogically, overreliance on AI-generated content correlates with diminished critical thinking skills (Huang et al., 2025) and oversimplification of complex disciplinary dialogues (Tack & Piech, 2022). These concerns are further exacerbated by technical limitations, including factual inaccuracies (Choi et al., 2023) and algorithmic biases in culturally sensitive contexts (Qadir, 2022). Institutional barriers include high infrastructure costs and ethical dilemmas surrounding data privacy and workforce displacement (Thurzo et al., 2023). Policy responses remain polarized, ranging from outright bans in schools (Johnson, 2023) to proactive ethical guidelines, highlighting the urgent need for balanced regulatory frameworks (Mhlanga, 2023; Rudolph et al., 2023). 2.3. Research Gaps The Social Construction of Technology (SCOT) theory emphasizes that technological innovation is a process co-shaped by diverse social groups through interactive negotiation (Pinch & Bijker, 1984). While a growing body of scholarship has highlighted ethical risks and pedagogical disruptions associated with generative AI in higher education, much of this literature remains grounded in expert commentary or top-down theoretical reflection (AlAfnan et al., 2023; Kasneci et al., 2023). Emerging empirical studies based on Twitter data have revealed more nuanced, bottom-up perspectives on AI adoption that public discourse on social media encompasses richer dimensions of discussion, such as academic integrity, skill development, and technological limitations (Tlili et al., 2023; Leiter et al., 2023). These discursive practices, which reflect lived experiences of various stakeholders, not only supplements the multifaceted realities of technology adoption (Haque et al., 2022) but also unveils the dynamic interplay of “relevant social groups” during educational technology development within the SCOT framework (Taecharungroj, 2023). However, current studies predominantly focus on Western social media platforms (Feng et al., 2023), with limited exploration of how role negotiation and consensus-building mechanisms unfold among key stakeholders in Chinese higher education context—including policymakers, university administrators, educators, students, and technology developers—in AI implementation. Differences in cultural context and institutional environments may significantly influence the sociotechnical construction pathways. For instance, the unique ecosystem of Chinese higher education system, characterized by the coexistence of administrative dominance and rapid market-driven digitalization, may shape AI adoption models distinct from Western paradigms. To address the above-mentioned gap, this study proposes the following research questions: RQ1: What core themes emerge from discussions about AI in higher education on Chinese social media? RQ2: How do these discussions reflect the interpretive flexibility of technology among different social groups within the SCOT framework? RQ3: How are intergroup controversies resolved through closure mechanisms within the SCOT framework, and what implications does such consensus hold for the future trajectory of AI adoption in Chinese higher education? 3. Methodology 3.1 Research Framework This study adopts a mixed-methods research design to systematically investigate the social construction process of AI usage in higher education as manifested on Chinese social media. The research procedure consists of three stages. First, Latent Dirichlet Allocation (LDA) model was employed to conduct topic modelling on the collected social media data, aiming to uncover the core dimensions of public discourse (RQ1). The selection of LDA stems from its unique strength at capturing latent semantic structures within large-scale online discussions (Blei et al., 2003). Unlike traditional manual coding, its unsupervised learning mechanism effectively circumvents researcher bias by identifying naturally emergent thematic patterns through probabilistic distribution models (Taecharungroj, 2023; Li et al., 2024). This approach aligns with the SCOT framework’s emphasis on “interpretive flexibility” research needs, specifically, how technological meanings self-organize through interactions among different social groups. Second, building upon the SCOT framework, directed content analysis was applied to code social groups within the collected discourse data, examining technological interpretation strategies across distinct actor groups (RQ2). Finally, to answer RQ3, the study integrated the findings from the previous two stages with policy documents and related materials. This synthesis allowed for an analysis of closure mechanisms, that is, how competing interpretations of technology are negotiated, stabilized, or contested, thus providing insight into the ongoing consensus-building pathways for technological controversies and their implications for the future trajectory of AI adoption in Chinese higher education. Table 1 Research Framework: Social Construction of AI in Higher Education 3.2. Data collection This study collected publicly available data related to AI-assisted writing from Sina Weibo Open API v2 over the period from December 20, 2024, to January 30, 2025. Adopting Bruns and Stieglitz’s (2013) hashtag-based data collection framework, which prioritizes precision over generic keyword crawling by leveraging platform-native topic indexing, the dataset is identified as original posts and their corresponding comments under four hashtags (see Table 1). This approach leverages the structural similarities between Weibo and Twitter (X) in hashtag functionality (i.e., centralized topic aggregation and user-driven engagement patterns) to minimize semantic noise and exclude irrelevant or contextually peripheral content, supporting subsequent topic modelling and SCOT-based group classification. Table 2 Selected Weibo Hashtags for Data Collection Hashtag Launch Date View Count #大学生作业里充满了AI味# #College Students’ Homework is Full of AI-generated Content# 2024/12/20 26.42 million #大学教师称学生快失去了原创写作能力# #University Professors Warn Students Are Losing Original Writing Abilities# 2024//12/20 1.59 million #AI辅助写作一刀切禁止不现实# #A Blanket Ban on AI-assisted Writing is Unrealistic# 2024/12/26 6.97 million #近8成学霸备考离不开AI# #Nearly 80% of Top Students Rely on AI for Exam Preparation# 2024/12/25 15.98 million To ensure research validity in keyword selection, as shown in the Table 2, we first retained only hashtags explicitly addressing AI’s application in educational writing assistance (e.g., #大学生作业里充满了AI味#, i.e. #College Students’ Homework is Full of AI-generated Content#) while excluding topics with insufficient semantic relevance (e.g., #AI绘画争议#, i.e. #AI Generated Art Controversy#). This filtering protocol aligns with Bruns and Stieglitz’s (2013) emphasis on hashtag-driven semantic anchoring, which isolates context-specific discourse and reduces the ambiguity inherent in broad keyword queries. Second, following Tufekci’s (2014) observation that a decline in hashtag activity does not equate to discussion cessation of a discussion, we implemented a 30–40 day delayed crawling protocol to capture stabilized user-generated content, thereby improving temporal validity of the data. To mitigate platform disruption risks highlighted by Driscoll and Walker’s (2014) analysis of Twitter API limitations, API requests were capped at 120 calls per minute. Finally, all procedures adhered to the ‘ethical pluralism’ principle for public data reuse (Fiesler & Proferes, 2018), including data anonymization and exclusion of non-consenting private accounts. This protocol yielded 4,750 valid raw data entries (N = 4,750). 3.3. LDA-Based Topic Modeling Latent Dirichlet Allocation (LDA) (Blei et al., 2003) is a widely-used probabilistic generative model for uncovering latent thematic structures within large document collections. To improve LDA analysis,we first performed essential preprocessing on the collected data before building the model. 3.3.1. Data Preprocessing The dataset refinement process employed a multi-stage cleansing protocol tailored to the linguistic characteristics of the Sina platform. An initial round of manual curation removed three categories of non-discursive content: (1) numeric repetition artifacts (“111”, “2222”) indicative of bot-generated interactions, (2) platform engagement prompts (e.g., “@xxx”), and (3) formulaic phatic expressions (e.g., “Good afternoon”) that lack analytical relevance. Subsequently, automated processing via Python’s re module systematically eliminated emojis, URLs, and special symbols (#, @), addressing the noise prevalence characteristic of Chinese Sina corpora. This hybrid cleaning pipeline validated 4,122 texts (86.8% of 4,750 initial entries), achieving an optimal balance between noise reduction and semantic preservation. The final preprocessing step prior to formal text segmentation involved noise reduction through application of a customized stop-word list. To build this resource, three authoritative Chinese lexical resources were integrated: the Baidu, SCU, and HIT stopword lists. Following methodological precedents in social media text processing (Xie et al., 2019), the composite lexicon combines 1) standard linguistic stopwords, 2) platform-specific noise patterns (e.g., “点击展开全文”, i.e., Read more ), and 3) domain-specific formulaic expressions, achieving comprehensive filtering for contemporary web-based Mandarin corpora. This comprehensive approach ensured that filtering was suited to the contemporary web-based Mandarin discourse environment. The Chinese text segmentation process employed the jieba tokenizer in precision mode, a widely-adopted tool for handling Mandarin’s character-bound morphology (Peng et el., 2015; Liu et al., 2020; Zhang et al., 2020; Guo et al., 2024). Unlike English-oriented tools such as NLTK , which rely on whitespace delimiters for word seperation(Loper & Bird, 2002), jieba ’s architecture is specifically designed to resolve the agglutinative nature of Chinese script—a critical advantage given the absence of inherent word separators in written Mandarin (Huang et al., 2017). Domain adaptation was achieved through proactive lexicon expansion, where AI-specific terminology (e.g., “ChatGPT”, “文心一言/ERNIE Bot”) was manually integrated into jieba’s segmentation rules. This adaptation enabled the preservation of semantically coherent compound terms, aligning with established practices in Chinese NLP for topic modeling (Zhou et al., 2005). The inclusion of a controlled vocabulary ensured LDA’s ability to capture domain-specific semantic patterns, as validated in probabilistic topic models (Blei et al., 2003; Griffiths & Steyvers, 2004). 3.3.2. LDA Modeling We conducted LDA analysis using Gensim ’s specialized LdaMulticore implementation (Řehůřek & Sojka, 2010), a parallel computing variant designed for efficient topic modeling on multi-core systems. This implementation preserves the theoretical foundations of standard LDA while substantially accelerating computation via distributed workload processing, utilizing optimization strategies such as data-aware scheduling and pipeline parallelism as demonstrated in distributed LDA frameworks (Liu et al., 2011). Computational efficiency was prioritized by configuring three parallel processes (workers = 3) and 20 training iterations (passes = 20), while a fixed random seed (random_state = 7) ensured full reproducibility of the stochastic initialization. Text preprocessing involved generating a dictionary through tokenization and constructing a document-term matrix using bag-of-words (BoW) vectorization (Zhang et al., 2010). Rather than imposing a predefined topic count, the model progressively refined latent thematic structures through repeated passes over the corpus. 3.3.3. Optimal Topic Number Selection To determine the optimal number of topics (k) in LDA models, a hybrid approach combining perplexity minimization, quantitative coherence metrics (e.g., C_v, U_mass), and human evaluation of topic interpretability was employed. Topic coherence quantifies the semantic consistency of a topic’s top terms by assessing their co-occurrence patterns in reference texts (Lau et al., 2014). Statistical measures metrics calculate how frequently terms appear together relative to their individual occurrence rates, with higher coherence scores indicating stronger contextual associations between terms. Robust co-occurrence patterns typically reflects human-interpretable themes (Röder et al., 2015). Thus, topic coherence offers distinct advantages for evaluating topic models, particularly in specialized domains. By quantifying the semantic relatedness of high-probability terms within topics, coherence metrics provide interpretability guarantees that perplexity inherently lacks (Chang et al., 2009). To evaluate topic coherence in our LDA model, we employed the C_v coherence metric (Röder et al., 2015). First, we trained LDA models with varying topic counts (1–30) and calculated their C_v coherence scores. Next, we identified the optimal topic count by locating the C_v score peak (indicating maximal semantic coherence) and manually inspecting topic keywords for interpretability. As shown in Fig. 1, the coherence curve reaches a peak coherence score of 0.532 at 14 topics, beyond which scores plateau with diminishing returns. Manual inspection of topic keywords confirmed that 14 topics preserved thematic granularity (reflecting nuanced opinion patterns) while avoiding semantic redundancy. Therefore, k = 14 was selected as the optimal topic count, balancing statistical validity and human interpretability. 3.4. Content Analysis To address RQ2, we first conducted a content analysis in formed by the SCOT framework. A codebook was developed through the integration of the definitions of social groups proposed by in prior SCOT literature (Humphreys, 2005b) and the specific contextual characteristics of AI applications in higher education. Two researchers independently coded a subset of 2895 comments based on the preliminary codebook to identify the different group tendencies implied in the contents. Before formal coding, three rounds of rounds were conducted to refine coding rules and resolve discrepancies, achieving high inter-coder reliability with a Cohen’s κ of 0.906 (indicating substantial agreement). To ensure analytical clarity, each comment was assigned exclusively to one social group category and no overlaps were allowed. Therefore, the final coding results revealed the distribution of social groups within the dataset. Following the completion of coding, textual analysis was conducted separately for comments classified within each of the four social group categories. Table 3 presents the final five-category coding scheme, specifying the operational definitions of four types of social groups in this study and a residual category, which includes comments that do not clearly fall into any of the defined social-group categories. Table 3 Classification Schema for Social Groups with Residual Category in Content Analysis Group Operational Definition Example Producers Groups involved in the development, funding, or marketing of AI technologies (e.g., engineers, investors). Through technical specifications and commercialization strategies, they determine the core capabilities of AI tools, including both dedicated educational AI systems and general-purpose platforms later adapted for pedagogy (e.g., ChatGPT repurposed for academic writing). “The AI features on [Brand]’s tablet are exceptionally user-friendly. Let’s go try it out together!” “Effortlessly take notes and organize materials with AI, [Brand]’s tablet model helps you stand out from information overload!” Advocates Groups actively shaping the legitimacy and institutional boundaries of AI applications in higher education through academic discourse, media engagement, or policymaking (e.g., policymakers, researchers, professors). Their stances (pro or con) redefine the sociotechnical significance of AI by framing its permissible uses within institutional teaching, learning, and administrative practices. “A teacher remarked, ‘AI has made students’ work carry a mechanical feel, lacking independent thinking and originality.’ ” “Many university faculty members report that ... though widespread adoption of AI tools has boosted academic productivity, students’ works are increasingly devoid of vitality, with originality facing severe challenges.” Users Groups directly interacting with AI tools (whether purpose-built for education or not) within higher education contexts. Their adaptive practices drive technical refinements through educational repurposing, while redefining the pedagogical utility and context-specific ethical norms of AI tools. “How can you know that I’m AI-generating my term paper that’s had me stuck for hours?” “ChatGPT has become a trusted partner for university students and postgraduates worldwide.” Bystanders Groups comprising 1) non-users (e.g., parents unaware of AI, pedagogy scholars) whose perceptions are mediated through social ties, and 2) AI users outside academia (e.g., general consumers).Their perspectives (e.g., media critiques of AI existential risks) or cross-domain engagements (e.g., corporate AI skill requirements) indirectly mediate the adoption patterns and social legitimacy of AI in higher education. “Honestly, I’ve started outsourcing repetitive and formulaic tasks to AI text generators these days... Can’t blame college students either...” “Artificial Intelligence is the future itself, we must seize the moment. This revolution stands as nothing less than the Industrial Revolution or the Internet Revolution!” Residual Category Comments that cannot be clearly assigned to any of the above categories due to vagueness, neutrality, or irrelevance to educational contexts. “I’m not quite getting it.” “Is AI making us lazy or leveling up our skills? Let’s hear firsthand from university students across campuses!” Note. Two processing measures applied: a) All quoted comments remain anonymous; b) Commercial brand names quoted in producers’ comments was replaced with the neutral identifier [Brand]. 4. Results 4.1. LDA-Based Topic Modeling Results Table 4 presents the results of the LDA analysis. Based on the C_v coherence metric, the optimal number of topics was identified as 14. Subsequently, two researchers independently reviewed the top 30 high-frequency keywords for each topic (translated into English terminology), and through discussion, assigned thematic labels to each topic. To enhance interpretability, thematically related topics were further grouped into higher-order conceptual categories. This table also displays the most relevant high-frequency words associated with the theme of each topic. The first category focuses on the instrumental application of AI technology in learning scenarios, primarily involving usage scenarios and underlying motivations. The feature words “writing/exam preparation” in Topic 12 indicate that AI has permeated the entire learning process, from information retrieval and content generation to exam preparation. Topic 2 further reveals the core motivations behind college students’ reliance on AI for academic tasks. High-frequency terms such as “thesis/assignments/low-effort courses” indicate that AI is predominantly applied to writing course papers and handling routine coursework for “low-effort courses”. The term “watered-down courses (水课) ” originates from the Chinese context, reflecting students’ subjective evaluation of courses perceived to offer low knowledge acquisition or skill application. Data indicate that students tend to employ AI tool to complete assignments for such courses to increase efficiency, especially when facing repetitive or low-stake assignments. Furthermore, users generally cite that excessive workloads and tight deadlines as key drivers of AI use, as reflected by terms such as “days/weeks”. These temporal markers suggest that during finals periods, they often need to complete large volumes of thesis assignments within two weeks or even just a few days. Notably, Topic 9 reveals that AI usage is not limited to struggling students. In fact, many top-performing students adopt AI to optimize their learning strategies, with users reporting that academically successful peers around them routinely integrate AI into their study routines. This finding suggests a normalization of AI-assisted learning across academic performance levels. The second category highlights controversies arising from the application of AI technology in academia. Topics 5 and 6 reflect both advantages and disadvantages. The juxtaposition of “generation/templates/dependence” with “convenience/time-saving” demonstrates AI’s efficiency improvements in exam preparation while raising concerns about learning capability degradation due to overreliance on technological. A particularly notable issue is the homogenization in student submissions, as assignments generated using AI often follow identifiable structural templates. Some instructors report being able to even identify specific AI tools used for the assignments, reflecting a growing awareness of stylistic convergence in student work. Topic 14, which features keywords “innovation ability/weakening/coping” directly underscores the potential structural deficiencies in academic capabilities caused by heavy dependency. Students who use AI to complete assignments may fail to develop targeted competencies, ultimately weakening their innovation capacity over time. In parellel, defined by terms such as “Integrity/Risks”, Topic 11 exposes how AI assistance may undermine fundamental values of academic ethics, particularly integrity. This constitutes a primary rationale for advocating prohibitions on AI in academic contexts. The third category predominantly reflects users’ positive evaluations of AI technology value. Terms “efficiency/intelligence/convenience” in Topic 1 and “productivity/cost-effectiveness” in Topic 3 collectively construct a technical efficacy assessment framework, indicating widespread recognition of AI’s breakthrough effects in optimizing workflows and enhancing resource utilization. Topic 4’s high-frequency terms “development/excellent” suggest societal optimism regarding AI’s technological evolution trajectory. Meanwhile, “omnipresent” particularly demonstrates AI’s high penetration rate in China, reinforcing its strategic position as a new productivity tool in the educational system. The fourth category reveals universities’ governance challenges in addressing AI technological impacts. Represented by terms “universities/prohibition/originality”, Topic 8 points to typical regulatory measures, as exemplified by Fudan University’s written policy banning AI use in thesis writing to preserve academic originality. In contrast, Topic 10 features terms including “one-size-fits-all/improvement/efficiency”, which exposes public discontent with overly rigid regulatory practices. “One-size-fits-all” ban are often deemed impractical or misaligned with actual educational needs in the public discussion. Consequently, current AI policy-making in higher education faces dual pressures of keeping pace with technological evolution and upholding academic ethics. The last category examines the forward-looking impacts of AI technological development. As characterized by “era/intelligence/satisfaction”, Topic 7 indicates a framework connecting technological and social evolution, emphasizing AI’s role in steering contemporary society toward intelligent societal transformation. Within this framing, AI is portrayed not merely as a tool, but as a defining force of the current era, with some users even asserting that the we are already living in the AI era. Focusing specifically in the education domain, Topic 13 includes high-frequency terms such as “education/future/technology”, which indicate a consensus that AI urgently requires and will profoundly engage in reconstructing pedagogical systems, thereby driving reforms to address existing issues. Table 4 LDA Topic Clusters on AI in Higher Education Category Theme Topic No. Topic Feature Words AI Academic Assistance Practice College Students Using AI to Complete Assignments 2 College Students, AI, Thesis, Assignments, Low-Effort Courses, Days, Weeks AI-Assisted Full-Cycle Learning 12 AI, College Students /Materials, Assistance, Writing, Efficiency, Exam Preparation Top Students Optimizing Study Strategies with AI 9 AI, Top Students, Exam Prep, High Efficiency, Relaxation, Thesis AI Academic Assistance Controversies​​ Pros and Cons of AI Academic Assistance 5 AI, Thesis, Generation, ChatGPT, Polishing, Tools, Time-Saving, Templates 6 AI, Problem-Solving, Dependence, Tutorial Assistance, Convenience, Exams, Learning 14 AI, Practical Use, Dependence, Innovation Ability, Weakening, Savior, Coping AI Dependence and Academic Ethics Risks 11 AI, Learning, Efficiency, Thinking, Universities, Academic, Integrity Positive Evaluation of AI Technical Efficiency​​ AI Enhancing Learning & Work Efficiency 1 AI, Daily Use, Development, Technology, Top Students, Work, Intelligence, Convenience User Recognition of AI Technology 3 Overpowered, Functions, Sharing, Productivity, Cost-Effectiveness, Efficiency 4 AI, Powerful, Development, Technology, Intelligence, Useful, Excellent, Omnipresent Policies and Controversies of AI-Generated Content in Universities​​ University Regulations on AI-Generated Content 8 Universities, Originality, Independent Thinking, Fudan University, Prohibition Efficiency vs. “One-Size-Fits-All” AI Policies 10 Writing, One-Size-Fits-All, Prohibition, Originality, Policies, Efficiency, Improvement AI Era and Future Trends​​ Value of AI in the Intelligent Era 7 AI, User-Friendly, Era, Intelligence, Satisfaction Future Trends of AI in Education 13 AI, College Students, Education, Significance, Future, ​​ “Watered-down” courses, Technology, Teachers 4.2. Content Analysis Results The second research question examines how discussions on social platforms reflect the interpretive flexibility of technology (specifically AI usage in higher education) among different social groups. Coding results indicate the proportional representation of each group’s voices within the sampled discussion data. Specifically, bystanders accounted for the highest proportion of comments (73.75%), while advocates were the least represented group (0.83%). Although there was a significant gap between groups, users ranked second with 14.20%, followed by producers in third place at 8.46%. The residual category (unclassifiable comments) comprised 2.76% of the total. 4.3. Textual Analysis Results The textual analysis reveals distinct interpretive patterns across four social groups within the SCOT framework regarding AI in education. Results indicate a clear contrast: while producers predominantly framed AI through promotional discourse, emphasizing its benefits and efficiency, advocates raise regulatory concerns via mainstream media. Users (students) expressed strong practical endorsement rooted in academic pressures and efficiency considerations and bystanders presented a spectrum of attitudes spanning from positive acceptance to cautious neutrality. Crucially, these differentiated narratives directly exemplify the interpretive flexibility within the SCOT framework, illustrating how each social group assigns meanings to educational AI based on their positions: producers (commercial interests), advocates (institutional authority), users (experiential needs), and bystanders (observational perspectives). The following sections detail these variation patterns. 4.3.1. Producers Marketing content for a brand’s tablet with AI-assisted learning systems constitutes the vast majority (98.8%) of producers’ comments. Specifically, these promotional messages took two main approaches. The first approach is using brief, impactful slogans. These simple, straightforward phrases highlight the product’s key benefit: tablet with AI enhanced system (Example: “The new favorite of top students! AI boosts exam prep, doubles efficiency!”). Second, beyond explicit promotional slogans, producers also adopt a more simulated grassroots approach by mimicking the language style of student users. High-frequency words in this type of messaging are “top student” (学霸, 53.1%, n = 130), “study” (学习, 44.5%, n = 109), and “exam preparation” (备考, 38.8%, n = 95). Collectively, these messages promote the idea that “top students are all using AI for learning.” (Example: “AI is so useful, no wonder top students can’t do without it! I also want to use AI’s power to become one of them!”) The overall tone is enthusiastic and aspirational, often marked with frequent use of exclamation marks (43.7%, n = 107), seeking to generate emotional resonance among student readers. When describing the AI’s educational role, these messages often employ expressions like “a top student’s secret weapon,” “smart tool,” or “helpful assistant.” In summary, within the producers’ messaging, the use of AI in education is portrayed in an overwhelmingly positive light. 4.3.2. Advocates Although advocates represent the smallest portion in the dataset, their commentary is mainly drawn from mainstream media outlets, which grants them disproportionate reach and influence. Notably, one major verified account evens served as the host (the originator) of the Weibo topic: #College Students’ Homework is Full of AI-generated Content#. Comments in this category consist primarily of news reports, supplemented by a smaller number of personal statements from university educators. Typically, the news reports cite the opinions of university teachers or educational institutions to express their stance on AI use in education. For example, they quote teachers interviewed saying: “The homework submitted by students is full of AI-generated content, lacking any traces of independent thinking or originality.” Moreover, these pieces primarily focus on the negative impacts AI use has on students. Additionally, new policies implemented by Chinese universities regarding AI are frequently mentioned in this data. For instance, Fudan University issued regulations titled “Guidelines for the Use of Artificial Intelligence Tools in Undergraduate Thesis Writing”, which was framed as a leading institutional response to emerging ethical dilemmas. The coverage further noted that “many universities in China have started exploring the boundaries of AI technology applications.” Overall, advocates’ viewpoint towards AI use in higher education is predominantly negative, focusing mainly on restrictions and limitations. 4.3.3. Users User-generated comments originate almost entirely from students of in higher education (undergraduates or postgraduates), and the vast majority (92.0%) express support for the use of AI in education. Only 0.07% (n = 28) of users comment with a nuanced view of AI usage. An even smaller proportion, 0.01% (n = 5), hold a negative stance after personal experience with AI tools These critical voices primarily cited technical shortcomings, especially in academic writing tasks, where AI-generated content was described as unreliable due to fabricated references or inaccurate outputs. Among supportive users, 24.1% provided reasons for their favorable stance. These comments primarily focused on using AI-generated text for completing assignments or writing papers. Three main reasons emerged. First, users emphasized the overwhelming workload, especially during exam periods. They argued that using AI is necessary for coping with overwhelming assignments so that they can have time to review for finals. A typical comment stated: “Twelve papers in one month, for different courses. I don’t believe anyone could realistically write them all by themselves...” Second, dissatisfaction was expressed regarding low-value, outdated courses and tedious tasks. Users prefer using AI under such contexts to save time. One student’s comment captured this: “Do teachers for these ‘watered-down’ courses(水课, institutional compulsory courses or courses that pedagogically void yet credit-bearing) even read the papers we write meticulously? Since they don’t, why waste our time? How many unnecessary courses are there in university? We see the same outdated slides, older than the professors, year after year, how can they then criticize students for lacking innovative ideas in their papers?” Third, users reported feeling incapable of writing a paper due to lack of instruction or expertise. One user shared: “In my freshman ‘Innovation’ class, the teacher assigned a 3,000-word paper on infectious diseases, requiring proper citations, all due in a week. I’ve only taken the class for one semester, not years, what am I supposed to use besides AI?” Other reasons included using AI to improve learning efficiency, such as for data organization and generating mind maps. Additionally, some users expressing approval used rhetorical questions or exclamations for emphasis, e.g., “So what?” or “Stop policing college students!” In conclusion, users strongly affirm AI’s role in education. Their expressions of support are grounded in their personal academic struggles and reflect adaptive rationales for AI use. 4.3.4. Bystanders Since bystanders include both people who have used AI (but not in education) and those who have never used it, the attitudes captured in these comments are not limited to AI in education; rather, they cover general opinions about AI itself. These attitudes were categorized into four groups: positive, negative, nuanced (balanced view), and no clear stance. Quantitative coding shows that over half (51.4%, n = 1098) held a positive attitude towards AI. When describing AI’s benefits, one type of comment offered general praise (38.9%, n = 427). High-frequency words here included “good” (不错, n = 119), “smart” (智能, n = 76), “impressive” (厉害, n = 62), “easy to use” (好用, n = 55), “powerful” (强大, n = 51), “practical” (实用, n = 51), and “useful” (有用, n = 13). Another type focused specifically on acknowledging AI’s ability to improve efficiency (23.1%, n = 254), using words like “convenient” (方便, n = 142), “efficiency” (效率, n = 87), and “handy” (便利, n = 25). By contrast, negative attitudes constituted a relatively small share (4.5%, n = 98). These critical comments expressed several key concerns. Some pointed out that abusing AI could weaken human capabilities (e.g., “Choosing to be a happy fool simply because your thinking ability can’t match AI is the beginning of self-destruction.”). Others highlighted flaws in AI-generated content (e.g., “Mechanical AI writing fails to move hearts.”). Additionally, some employed a human vs. machine narrative, expressing fears about AI’s potential threat (e.g., “If AI runs out of control, who will fix it?”). A moderate number of comments expressed nuanced or balanced views (16.1%, n = 345). These fell into two subgroups. The first acknowledged both AI’s advantages and disadvantages. The second, comprising 27.5% (n = 95), emphasized the need for responsible or appropriate AI use, noting that AI is powerful but must be applied thoughtfully and ethically by human users. Finally, 27.8% of comments (n = 594) exhibited no explicit stance. These comments either marveled at AI’s omnipresence (e.g., “AI really is in every part of our work and life.”) or acknowledged AI as an irresistible trend (e.g., “It’s the era of AI for everyone now.”). Further, some shared experiences suggesting AI’s importance in future workplaces (e.g., “It’s okay, working people do this too. It’s just about learning and using it a few years ahead.”). In summary, bystanders demonstrated a complex and multilayered orientation toward AI. While a clear majority viewed it positively, critical concerns and calls for reasonable and ethical usage were also present. 5. Discussion Within the data collected, producers almost exclusively frame AI technology through the singular narrative of “efficiency enhancement” to promote AI-integrated educational tools. Their marketing discourse centers on a tablet with AI learning features, constantly emphasizing in their advertisements that it can transform learners into “top students.” For example, slogans like “The secret weapon for top students, doubling efficiency” are directly used. Or, disguised as students, they post comments such as “All the top students are using it, I want to try it too” to stimulate peer endorsements. This strategy precisely taps into the group psychology in China’s competitive educational environment: high achievers are portrayed as role models, while average students’ fear of “falling behind” drives herd-like consumption. It’s clear that when interpreting AI technology, producers deliberately choose the single narrative of an “efficiency tool,” avoiding potential controversies about AI in education. They package their product as a “must-have learning device” with no downsides. This strategy excessively amplifies the product’s advantages, essentially lowering the cognitive barrier for consumers by simplifying the technology’s semantic meaning. Ironically, this marketing logic also reflects a real contradiction in China’s education technology market: despite producers claiming that “AI use represents an education revolution”, they continue to anchor technological innovation within the utilitarian framework of measurable improvements in test performances. In contrast to producers’ enthusiastic promotion of AI across all learning scenarios, advocates, including universities, faculties and educational departments adopt a markedly more cautious stance to restrict AI use in higher education. Their chosen narrative for AI is the exact opposite of the producers’, emphasizing academic ethics risks and the danger of AI replacing human capabilities. This attitude is not rooted in anti-technological sentiment but rather stems from problems observed in actual teaching practice. University instructors have reported that student submissions increasingly exhibit a “distinctive AI tone, lacking traces of independent thought,” indicating that concerns about AI’s impact on learning originality have materialized into genuine issues. The measures taken by advocates mainly involve setting rules. For instance, institutions such as Fudan University (2024) have issued AI Use Guidelines for Graduation Theses, whose core idea is to define clear boundaries for technology use, prioritizing restrictions on key activities like graduation theses. This approach reflects a practical consideration in China’s educational management: since a complete AI ban is impossible, policies should prioritize preserving the academic integrity of academic evaluation. However, the problem is that these local restrictions clash with the “all-scenario use” pushed by producers. Through product design and advertising, producers implicitly encourage students to rely on AI for routine homework, class notes, exam preparation, and other learning tasks. If students consistently use AI to take over basic learning tasks, their independent thinking skills may decline. This not only conflicts with the advocates’ rules forbidding AI use in theses but also risks students gradually losing the ability to complete academic work independently. In effect, the advocates’ fear of “innovation ability weakening” is being precisely validated through this daily penetration of AI. The vast majority of users (students) support the use of AI in education. This attitude is not blind enthusiasm for technology, but a survival strategy within the structural contradictions of Chinese higher education. Specifically, students cite three main reasons for using AI: overwhelming academic pressure, being assigned tasks beyond their current ability, and silent resistance against inefficient or outdated courses. These reasons reveal problems in the current education system’s instructional design, task distribution, and evaluation mechanisms. When faced with excessive tasks such as “12 different course papers in one month”, AI is seen as a tool to boost efficiency. When first-year students are required to write 3000-word specialized papers far exceeding their knowledge base in a week, AI serves as a bridge to fill that capability gap. For inefficient courses where “professors reuse decade-old slides”, using AI to write assignments becomes a protest against poor teaching. Students are aware of the ethical risks of AI but prioritize its utilitarian value to cope with their situation. The core motivation is the disconnection between goals and methods in teaching design. Assignments meant to serve skill development are instead used as quantitative metrics in the student evaluation system. To avoid failing or to get better grades, many students turn to AI. The differing interpretations of AI use among bystanders reflect how the public, drawing on diverse experiences, forms distinct cognitive frameworks about AI’s value and risks during the the integration of technology into society. Most affirm the necessity of AI, recognizing its substantial boost to life and work efficiency. This favorable view stems from AI’s significant advantages in areas like information processing and process optimization. Especially in the workplace, AI has become a core element in reshaping productivity, instilling a crisis mindset of “lagging behind if not using AI” in the public. However, many hold a dialectical position, always carefully weighing the boundaries of AI use. Concerns arise when AI evolves from an assistant tool into a potential decision-maker. The public worries that AI might erode human critical thinking. Individuals acknowledge AI’s value while also being vigilant against its potential to alienate human agency. Therefore, some advocate for “scientific use” mechanisms to balance tool dependency and skill retention. What emerges as particularly noteworthy within in the public discourse is the collective narrative of the “AI era”. Phrases like “the era of AI for all” and “AI for work too” reflect a developing social consensus positioning AI as basic infrastructure, akin to water or electricity. AI’s importance fuels the notion that individuals unable to master AI tools risk dual marginalization, both in professional competition and knowledge acquisition. In this context, the adoption by universities of blanket bans on AI are arguably going against the tide. On one hand, such prohibition fails to address the root cause: students are forced to depend on AI due to workload overload and ability mismatches. On the other, an AI-free educational environment deprives students of opportunities to develop critical AI literacy, thereby widening the skill gaps between them and societal demands. Thus, it is urgently required that the education system move beyond simplistic control-oriented governance to adaptative, context-sensitive regulatory frameworks aligned with the realities of technological integration. Distinctively, the closure mechanism for AI use in Chinese higher education will be significantly policy-driven, established through collaboration between the government’s education authorities and universities. Within this framework, a tiered governance model based on different scenarios is recommended. Drawing inspiration from the “four-category scenario” approach in the Shanghai Jiao Tong University (2025) policy “SJTU Guidance on Development and Governance of AI in Education and Teaching (Trial Version)”, AI use scenarios could be classified into: Prohibited (e.g., graduation theses and innovative research requiring strict prevention of AI substituting independent thought), Restricted, Encouraged (e.g., literature search, data processing), and Open Levels. This tiered AI management can help achieve both ethical constraints and technological empowerment simultaneously. It’s important to note that current university oversight often focuses on technical measures like AI-generated text detection rates. However, the key to truly addressing AI dependency lies in reconstructing the course design and evaluation systems. To prevent students from being forced into AI use by excessive tasks, educational institutions must promote course reform. First, reduce inefficient repetitive assignments and replace them with tasks requiring critical thinking to lessen reliance on substitution tools. Then, consider mandating AI literacy training as part of general education, focusing on developing two key abilities: critical usage skills (e.g., verifying information authenticity, identifying algorithmic bias) and human-AI collaborative creativity. Furthermore, developing teachers’ AI competency, positioning them not only as knowledge transmitters because, as AI becomes powerful at knowledge delivery, the irreplaceable role of teachers must increasingly lie in sparking innovative potential and ethical reflection, as well as nurturing morality. In an AI era, higher education should not merely chase tool efficiency. Instead, it should center on capability cultivation, aiming to foster whole persons capable of harnessing tools without becoming alienated by them and maintaining clear thought in the AI age. 6. Conclusion This study applied the Social Construction of Technology theoretical framework and a mixed-methods approach to systematically investigate the social construction process of artificial intelligence adoption in higher education on Chinese social media (Weibo). The key finding indicates that different relevant social groups demonstrated significant “interpretive flexibility” toward AI technology: producers framed AI as an “essential learning tool” for efficiency enhancement due to marketing strategies; advocates focused on academic ethical risks and capability replacement concerns; users (students) strongly supported its instrumental value primarily due to academic pressure; while bystanders generally recognized its epochal significance but called for cautious use. These diverse interpretations reflected distinct group positions. The study revealed a policy-driven characteristic in the “closure” mechanism within China’s higher education context. Its future trajectory may rely on coordinated development between stratified governance models and teaching system reforms. One key contribution of this study lies in empirically examining how relevant social groups negotiate the meaning and application boundaries of AI technology in China’s higher education environment, characterized by administrative dominance coexisting with market mechanisms, using Chinese social media data. Additionally, the study explores a technology dispute coordination mechanism driven by administrative power within the Chinese context. This work provides important empirical evidence for applying the SCOT theory in non-Western higher education settings. It demonstrates how different institutional environments substantially influence technology acceptance and implementation paths in society and further confirms the diverse characteristics of technology adoption patterns across cultures. Nonetheless, thestudy has several limitations. The data primarily derives from public discussions from a single social media platform (Weibo). While effective for capturing public opinion, it may not sufficiently cover internal perspectives or private concerns within specific groups, such as university administrators and some education policymakers. Future research should expand data sources by incorporating in-depth interviews or more diverse social media data to better depict complex stakeholder positions and interaction dynamics. Declarations Data availability The datasets supporting the findings of this study have been deposited in the FigShare repository and are publicly available under the DOI: 10.6084/m9.figshare.29605916. Competing interests The authors declare no competing interests. Ethical statements This article does not contain any studies with human participants performed by any of the authors. All data were collected from Sina Weibo Open API v2 legally. Funding Declaration This study was funded by the Science and Technology Department of Guangxi Zhuang Autonomous Region, Talent Project of Guangxi Science and Technology Department (Funding No. 2022AC21201). The funders had no role in study design, data analysis, or manuscript preparation. References Adiguzel, T., Kaya, M. H., & Cansu, F. K. (2023). Revolutionizing education with AI: Exploring the transformative potential of ChatGPT. Contemporary Educational Technology , 15 (3), ep429. https://doi.org/10.30935/cedtech/13152 AlAfnan, M. A., Dishari, S., Jovic, M., & Lomidze, K. (2023). ChatGPT as an educational tool: Opportunities, challenges, and recommendations for communication, business writing, and composition courses. Journal of Artificial Intelligence and Technology , 3 (2), Article 1. Blei, D. M., Ng, A. Y., & Jordan, M. I. (2003). Latent Dirichlet allocation. Journal of Machine Learning Research , 3 , 993–1022. Bijker, W. E. (1997). Of bicycles, bakelites, and bulbs: Toward a theory of sociotechnical change . MIT Press. Bloor, D. (1976). Knowledge and social imagery . University of Chicago Press. Bruns, A., & Stieglitz, S. (2013). Towards more systematic Twitter analysis: Metrics for tweeting activities. International Journal of Social Research Methodology , 16 (2), 91–108. https://doi.org/10.1080/13645579.2012.756095 Chang, J., Gerrish, S., Wang, C., Boyd-Graber, J., & Blei, D. (2009). Reading tea leaves: How humans interpret topic models. Advances in Neural Information Processing Systems , 22 , 288–296. https://proceedings.neurips.cc/paper/2009/hash/f92586a25bb3145facd64ab20fd554ff-Abstract.html Choi, J. H., Hickman, K. E., Monahan, A., & Schwarcz, D. B. (2023). ChatGPT goes to law school. SSRN Electronic Journal . https://doi.org/10.2139/ssrn.4335905 Clayton, N. (2002). SCOT: Does it answer? Technology and Culture , 43 (2), 351–360. Driscoll, K., & Walker, S. (2014). Big data, big questions | Working within a black box: Transparency in the collection and production of big Twitter data. International Journal of Communication , 8 , Article 20. https://ijoc.org/index.php/ijoc/article/view/2174 Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., ... & Wright, R. (2023). "So what if ChatGPT wrote it?" Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. International Journal of Information Management , 71 , Article 102642. https://doi.org/10.1016/j.ijinfomgt.2023.102642 European Union. (2024). Regulation (EU) 2024/1689 on artificial intelligence. Official Journal of the European Union , L 200 , 1–349. Retrieved July 20, 2025, from https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689 Feng, Y., Poralla, P., Dash, S., Li, K., Desai, V., & Qiu, M. (2023). The impact of ChatGPT on streaming media: A crowdsourced and data-driven analysis using Twitter and Reddit . arXiv. Fiesler, C., & Proferes, N. (2018). "Participant" perceptions of Twitter research ethics. Social Media + Society , 4 (1), 1–14. https://doi.org/10.1177/2056305118763366 Fudan University. (2024). Regulations on the use of AI tools in undergraduate theses (trial implementation) . Retrieved July 20, 2025, from https://www.fudan.design/ai.html Garvey, E. G. (1996). The adman in the parlor: Magazines and the gendering of consumer culture, 1880s to 1910s . Oxford University Press. Griffiths, T. L., & Steyvers, M. (2004). Finding scientific topics. Proceedings of the National Academy of Sciences , 101 , 5228–5235. https://doi.org/10.1073/pnas.0307752101 Guo, F., Liu, Z., Lu, Q., Ji, S., & Zhang, C. (2024). Public opinion about COVID-19 on a microblog platform in China: Topic modeling and multidimensional sentiment analysis of social media. Journal of Medical Internet Research , 26 , Article e47508. https://doi.org/10.2196/47508 Haque, M. U., Dharmadasa, I., Sworna, Z. T., Rajapakse, R. N., & Ahmad, H. (2022). "I think this is the most disruptive technology": Exploring sentiments of ChatGPT early adopters using Twitter data . arXiv. Huang, Q., Lv, C., Lu, L., & Tu, S. (2025). Evaluating the quality of AI-generated digital educational resources for university teaching and learning. Systems , 13 (3), Article 174. https://doi.org/10.3390/systems13030174 Huang, S., Sun, X., & Wang, H. (2017). Addressing domain adaptation for Chinese word segmentation with global recurrent structure. In Proceedings of the Eighth International Joint Conference on Natural Language Processing (pp. 184–193). Taipei, Taiwan: AFNLP. Humphreys, L. (2005a). Cellphones in public: Social interactions in a wireless era. New Media & Society , 7 (6), 810–833. https://doi.org/10.1177/1461444805058164 Humphreys, L. (2005b). Reframing social groups, closure, and stabilization in the social construction of technology. Social Epistemology , 19 (2-3), 231–253. Johnson, A. (2023, January 18). ChatGPT in schools: Here's where it's banned—and how it could potentially help students. Forbes . Retrieved July 20, 2025, from https://www.forbes.com/sites/ariannajohnson/2023/01/18/chatgpt-in-schools-heres-where-its-banned-and-how-it-could-potentially-help-students/ Kasneci, E., Seßler, K., Küchemann, S., Bannert, M., Dementieva, D., Fischer, F., & Kasneci, G. (2023). ChatGPT for good? On opportunities and challenges of large language models for education. Educational Technology & Society , 26 (1), 1–23. https://doi.org/10.30191/ETS.202301_26(1).0003 Koetsier, J. (2023, March 14). GPT-4 scores in the top 10% on the Uniform Bar Exam. Forbes . Retrieved July 20, 2025, from https://www.forbes.com/sites/johnkoetsier/2023/03/14/gpt-4-scores-in-the-top-10-on-the-uniform-bar-exam/ Lau, J. H., Newman, D., & Baldwin, T. (2014). Machine reading tea leaves: Automatically evaluating topic coherence and topic model quality. In Proceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics (pp. 530–539). Gothenburg, Sweden: Association for Computational Linguistics. https://doi.org/10.3115/v1/E14-1056 Leiter, C., Zhang, R., Chen, Y., & Riedl, M. O. (2023). AI in the wild: A longitudinal analysis of public perceptions of ChatGPT on Twitter. Proceedings of the International AAAI Conference on Web and Social Media , 17 , 505–517. Li, L., Ma, Z., Fan, L., Lee, S., Yu, H., & Hemphill, L. (2024). ChatGPT in education: A discourse analysis of worries and concerns on social media. Education and Information Technologies , 29 (9), 10729–10762. Liu, Q., Zheng, Z., Zheng, J., Chen, Q., Liu, G., Chen, S., ... & Ming, W. K. (2020). Health communication through news media during the early stage of the COVID-19 outbreak in China: Digital topic modeling approach. Journal of Medical Internet Research , 22 (4), Article e19118. https://doi.org/10.2196/19118 Liu, Z., Zhang, Y., Chang, E. Y., & Sun, M. (2011). PLDA+: Parallel latent Dirichlet allocation with data placement and pipeline processing. ACM Transactions on Intelligent Systems and Technology , 2 (3), Article 26. https://doi.org/10.1145/1961189.1961198 Loper, E., & Bird, S. (2002). NLTK: The natural language toolkit [Preprint]. arXiv. https://arxiv.org/abs/cs/0205028 Mackay, H., & Gillespie, G. (1992). Extending the social shaping of technology approach: Ideology and appropriation. Social Studies of Science , 22 (4), 685–716. MacNeil, S., Tran, A., Mogil, D., Bernstein, S., Ross, E., & Huang, Z. (2022). Generating diverse code explanations using the GPT-3 language model. In Proceedings of the 55th ACM Technical Symposium on Computer Science Education (pp. 1–7). Portland, OR: ACM. McLuhan, M. (1964). Understanding media: The extensions of man . MIT Press. Mhlanga, D. (2023). Open AI in education, the responsible and ethical use of ChatGPT towards lifelong learning [Preprint]. Social Science Research Network. Moore, R. L., Miller, C. N., & Leskovec, J. (2022). Leveraging AI for scalable data science education. Journal of Machine Learning Research , 23 (1), 1–30. Peng, K. H., Liou, L. H., Chang, C. S., & Lee, D. S. (2015). Predicting personality traits of Chinese users based on Facebook wall posts. In Proceedings of the 24th Wireless and Optical Communication Conference (pp. 9–14). Piscataway, NJ: IEEE. Pinch, T. J., & Bijker, W. E. (1984). The social construction of facts and artefacts: Or how the sociology of science and the sociology of technology might benefit each other. Social Studies of Science , 14 (3), 399–441. https://doi.org/10.1177/030631284014003004 Qadir, J. (2022). Engineering education in the era of ChatGPT: Promise and pitfalls. IEEE Access , 10 , 125813–125826. Řehůřek, R., & Sojka, P. (2010). Software framework for topic modelling with large corpora. In Proceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks (pp. 45–50). Valletta, Malta: ELRA. Röder, M., Both, A., & Hinneburg, A. (2015). Exploring the space of topic coherence measures. In Proceedings of the Eighth ACM International Conference on Web Search and Data Mining (pp. 399–408). Shanghai, China: ACM. https://doi.org/10.1145/2684822.2685324 Rudolph, J., Tan, S., & Tan, S. (2023). ChatGPT: Bullshit spewer or the end of authentic assessments? Journal of Applied Learning and Teaching , 6 (1), Article 1. Shanghai Jiao Tong University. (2025). Guidance on development and governance of AI in education and teaching (trial version) . Retrieved July 20, 2025, from https://www.sjtu.edu.cn/tg/20250304/207682.html Sok, S., & Heng, K. (2023). ChatGPT for educational chatbots: The balance between hype and reality. Journal of Educational Computing Research , 61 (2), 345–367. Stanford Teaching Commons. (2025). Understanding AI literacy . Retrieved July 20, 2025, from https://teachingcommons.stanford.edu/teaching-guides/artificial-intelligence-teaching-guide/understanding-ai-literacy Tack, C., & Piech, C. (2022). The pedagogical limitations of large language models in disciplinary dialogues. In Proceedings of the Ninth ACM Conference on Learning @ Scale (pp. 1–4). Copenhagen, Denmark: ACM. Taecharungroj, V. (2023). What can ChatGPT do? Analyzing early reactions to the innovative AI chatbot on Twitter. Big Data and Cognitive Computing , 7 (1), Article 35. https://doi.org/10.3390/bdcc7010035 Thurzo, A., Urbanová, W., & Waczulíková, I. (2023). The impact of AI on dental education: A scoping review. European Journal of Dental Education , 27 (3), 1–12. Tlili, A., Shehata, B., Adarkwah, M. A., Bozkurt, A., & Hickey, D. T. (2023). What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education. Smart Learning Environments , 10 (1), Article 15. https://doi.org/10.1186/s40561-023-00237-x Tufekci, Z. (2014). Big questions for social media big data: Representativeness, validity and other methodological pitfalls. In Proceedings of the International AAAI Conference on Web and Social Media (pp. 505–514). Ann Arbor, MI: AAAI. Winner, L. (1986). Do artifacts have politics? In L. Winner, The whale and the reactor (pp. 19–39). University of Chicago Press. (Original work published 1980) Xie, T., Qin, P., & Zhu, L. (2019). Study on the topic mining and dynamic visualization in view of LDA model. Modern Applied Science , 13 (1), 204. https://doi.org/10.5539/mas.v13n1p204 Zhang, Y., Jin, R., & Zhou, Z. H. (2010). Understanding bag-of-words model: A statistical framework. International Journal of Machine Learning and Cybernetics , 1 , 43–52. https://doi.org/10.1007/s13042-010-0001-0 Zhang, Y., Sun, M., Ren, Y., & Shen, J. (2020). Sentiment analysis of Sina Weibo users under the impact of super typhoon Lekima using natural language processing tools: A multi-tags case study. Procedia Computer Science , 174 , 478–490. Zhao, R., Yunus, M. M., & Rafiq, K. R. M. (2023). The impact of the use of ChatGPT in enhancing students' engagement and learning outcomes in higher education: A review. International Journal of Academic Research in Business and Social Sciences , 13 (12), Article e1–e15. Zhou, J. S., Dai, X., Ni, R. Y., & Chen, J. (2005). A hybrid approach to Chinese word segmentation around CRFs. In Proceedings of the Fourth SIGHAN Workshop on Chinese Language Processing (pp. 63–68). Jeju Island, South Korea: Asian Federation of Natural Language Processing. Zhu, H., Yu, B., Halfaker, A., & Terveen, L. (2018). Value-sensitive algorithm design: Method, case study, and lessons. Proceedings of the ACM on Human-Computer Interaction , 2 (CSCW), Article 119. Additional Declarations No competing interests reported. Cite Share Download PDF Status: Posted Version 1 posted You are reading this latest preprint version Research Square lets you share your work early, gain feedback from the community, and start making changes to your manuscript prior to peer review in a journal. As a division of Research Square Company, we’re committed to making research communication faster, fairer, and more useful. We do this by developing innovative software and high quality services for the global research community. Our growing team is made up of researchers and industry professionals working together to solve the most critical problems facing scientific publishing. Also discoverable on Platform About Our Team In Review Editorial Policies Advisory Board Help Center Resources Author Services Accessibility API Access RSS feed Manage Cookie Preferences © Research Square 2026 | ISSN 2693-5015 (online) Privacy Policy Terms of Service Do Not Sell My Personal Information {"props":{"pageProps":{"initialData":{"identity":"rs-7171423","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":527758140,"identity":"e5dc739a-3341-4d11-aa28-82063e64b5ce","order_by":0,"name":"Mingyi Yang","email":"","orcid":"","institution":"Guangxi University","correspondingAuthor":false,"prefix":"","firstName":"Mingyi","middleName":"","lastName":"Yang","suffix":""},{"id":527758141,"identity":"a438beaa-0bf0-4883-830b-3683685a8c68","order_by":1,"name":"Qi Song","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA7ElEQVRIiWNgGAWjYNACAxDB2HjgAYMNDz9/A/FaGg4kMKTJSM44QIJlQC2HbQwaEvCrkm/vPfziTQGDPP+0w0BbKs7zGDAcYPzwMQePk86cS7OcY8BgOON2IlDLmds85swNzJIzt+HRIpFjZgw0OYEBpCWx7TaPZcMBNmZePFrkZ0C1yIO1/DvHY3AgAb8Whhs5xo9BWgzAWhoOENZicOaMGSPILxvBfjmWzCM542AzXr/It/cYf3jzh0Fe7nb6wwcfauzs+fmbD374iM9hDAxsEjwM/5EFGBvwqgcC5g88hJSMglEwCkbByAYAfGlWvtT48owAAAAASUVORK5CYII=","orcid":"","institution":"Guangxi University","correspondingAuthor":true,"prefix":"","firstName":"Qi","middleName":"","lastName":"Song","suffix":""},{"id":527758142,"identity":"e2508c55-43e4-4e3c-a274-d434d5336ed7","order_by":2,"name":"Minghui Zou","email":"","orcid":"","institution":"University of Oxford","correspondingAuthor":false,"prefix":"","firstName":"Minghui","middleName":"","lastName":"Zou","suffix":""}],"badges":[],"createdAt":"2025-07-20 18:53:11","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7171423/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7171423/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":93422693,"identity":"6704c1d3-907e-4879-8052-5877ef8d5bf1","added_by":"auto","created_at":"2025-10-13 16:20:13","extension":"docx","order_by":0,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":173873,"visible":true,"origin":"","legend":"","description":"","filename":"EfficiencyToolorAcademicThreatMappingChinasAIControversyinHigherEducationwithMixedMethods.docx","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/e4e97b73b5ab3facd12e9170.docx"},{"id":93423698,"identity":"b2f2a514-c007-440a-801d-b87359b39437","added_by":"auto","created_at":"2025-10-13 16:28:13","extension":"json","order_by":1,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":5529,"visible":true,"origin":"","legend":"","description":"","filename":"44a1933dfdc742e68e5d9dd5b30fb769.json","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/3bcda7813de7f422ccea9e49.json"},{"id":93422690,"identity":"a4a4ecfb-13e6-45dc-9447-863e4f999c6c","added_by":"auto","created_at":"2025-10-13 16:20:13","extension":"xml","order_by":2,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":157069,"visible":true,"origin":"","legend":"","description":"","filename":"44a1933dfdc742e68e5d9dd5b30fb7691enriched.xml","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/809af8e354a0946d09faecad.xml"},{"id":93421901,"identity":"7a86b57a-1a68-4cba-869e-caed00c60036","added_by":"auto","created_at":"2025-10-13 16:12:13","extension":"png","order_by":3,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":28028,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/cdda0db341c720aedc5d2ed6.png"},{"id":93421899,"identity":"e21e7cc1-68b8-4b14-9f96-ac77f200c79b","added_by":"auto","created_at":"2025-10-13 16:12:12","extension":"png","order_by":4,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":77180,"visible":true,"origin":"","legend":"","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/c00b37261cdcd737cc8ee5d0.png"},{"id":93421904,"identity":"19504dc7-dceb-43da-9b8f-240cc80f6bc3","added_by":"auto","created_at":"2025-10-13 16:12:13","extension":"png","order_by":5,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":10483,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/4a4f7e638c554f32f30119a5.png"},{"id":93423699,"identity":"fa33729a-7301-4c44-91b3-998dd9eb6f84","added_by":"auto","created_at":"2025-10-13 16:28:13","extension":"png","order_by":6,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":14839,"visible":true,"origin":"","legend":"","description":"","filename":"Onlinefloatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/37200caf7d31d2bcb9746f3f.png"},{"id":93421907,"identity":"53b83f88-4ad0-427b-94df-0b37b04dab81","added_by":"auto","created_at":"2025-10-13 16:12:13","extension":"xml","order_by":7,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":152119,"visible":true,"origin":"","legend":"","description":"","filename":"44a1933dfdc742e68e5d9dd5b30fb7691structuring.xml","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/6ec0bb1706be38e97a397b98.xml"},{"id":93421906,"identity":"77292563-c038-40a1-a428-f3133f001237","added_by":"auto","created_at":"2025-10-13 16:12:13","extension":"html","order_by":8,"title":"","display":"","copyAsset":false,"role":"acdc-reference","size":166817,"visible":true,"origin":"","legend":"","description":"","filename":"earlyproof.html","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/5835f54df350f0c313e6e688.html"},{"id":93421898,"identity":"011467eb-5d32-42bd-ab03-d1312c7c853a","added_by":"auto","created_at":"2025-10-13 16:12:12","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":69363,"visible":true,"origin":"","legend":"\u003cp\u003eOptimal Topic Selection via Coherence Scores\u003c/p\u003e","description":"","filename":"floatimage2.png","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/5846b3b5a2f967e3b2ccefb5.png"},{"id":105347801,"identity":"6a655e60-c35a-4d34-ab1e-695a4973fe19","added_by":"auto","created_at":"2026-03-25 04:55:36","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1512242,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7171423/v1/3e457798-e7f4-44c2-9795-14613cf52b44.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Efficiency Tool or Academic Threat? Mapping China’s AI Controversy in Higher Education with Mixed Methods","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eGenerative AI is profoundly reshaping the global higher education landscape. Tools such as ChatGPT drive teaching paradigm shifts through adaptive content generation (Kasneci et al., 2023), while simultaneously triggering concerns over academic integrity and skill degradation worldwide. Notably, distinct national cultural contexts have fostered differentiated governance paths.\u003c/p\u003e\u003cp\u003eIn Western higher education systems, AI governance typically emphasizes ethical oversight and multi-stakeholder participation. For example, the European Union\u0026rsquo;s \u003cem\u003eArtificial Intelligence Act\u003c/em\u003e (Regulation (EU) 2024/1689) classifies AI application in educational assessments as high-risk (Annex III), mandating transparency disclosures, human oversight, and GDPR-compliant data governance (Article 13\u0026ndash;14). In parallel, U.S. universities seek to balance innovation with risk by cultivating comprehensive AI literacy among faculty and students. Stanford University (2025), for instance, introduced its \u003cem\u003eAI Literacy Guide\u003c/em\u003e, which outlines a multidimensional framework encompassing functional, ethical, rhetorical, and pedagogical competencies. This encourages technological exploration (e.g., customized AI tool development) while managing ethical risks (e.g., copyright issues, data bias), teaching instructors and students to use AI autonomously and wisely. Such governance logic reflects the Western tradition of academic autonomy and positions AI as a tamable instrumental entity, emphasizing that technology should serve human agency.\u003c/p\u003e\u003cp\u003eChina\u0026rsquo;s AI integration reveals a distinct sociotechnical dynamic. In the Chinese context, AI application in higher education is shaped by a triad of administrative intervention, market momentum, and meritocratic anxiety. AI producers employ marketing rhetoric like \u0026ldquo;super study tool\u0026rdquo; to promote efficiency worship, while university administrators impose bans due to academic ethics crises, exemplified by Fudan University (2024)\u0026rsquo;s \u003cem\u003eRegulations on the use of AI tools in undergraduate theses (trial implementation).\u003c/em\u003e This contradiction highlights the cognitive tensions distinctive to East Asian education. Technology is simultaneously sanctified as a competitive symbol that fuels public anxiety about \u0026ldquo;not using AI meaning obsolescence\u0026rdquo;, and alienated as a defensive target due to deep vigilance against skill replacement. When the same technology carries opposing meanings as both an \u0026ldquo;efficiency tool\u0026rdquo; and an \u0026ldquo;academic threat,\u0026rdquo; how can we decode its conflicting nature and reach a stable consensus?​​\u003c/p\u003e\u003cp\u003eThe social construction of technology (SCOT) framework offers key insights to this question. It posits that technological meaning is not fixed, but rather emergent, from negotiation practices among diverse groups (Pinch \u0026amp; Bijker, 1984). This framework is especially valuable for unpacking how meanings of AI are contested in China\u0026rsquo;s higher education system, where bureaucratic governance coexists with intense market pressure and competition. However, existing research on relevant social groups\u0026rsquo; views about AI in higher education primarily focuses on Western platforms such as X (Feng et al., 2023). Consequently ,critical gaps remain in understanding how Chinese stakeholders negotiate the meanings, risks, and boundaries of generative AI technologies. It also fails to address how closure mechanisms unique to the Chinese context reshape technological trajectories through provisional consensus. This Western-centric paradigm systematically obscures China\u0026rsquo;s distinctive phenomena of technological construction.​​\u003c/p\u003e\u003cp\u003eThis study seeks to address the abovementioned gap by analyzing negotiation dynamics among relevant social groups on Chinese social media. Drawing upon data collected from a Chinese social media platform, Weibo, we examined how contests over technological meaning led to temporary closure. A three-phase mixed methods approach was adopted. First, Laten Dirichlet Allocation (LDA) topic modeling was used to extract key themes from Weibo discussions. Then, directed content coding grounded in the SCOT framework was conducted to categorize producers, advocates, users, and bystanders, while parallel textual analysis examined semantic patterns within these same discussion threads. Finally, we traced the emergence of provisional closure mechanisms by examining how institutional policy interventions resolve disputes, while concurrently assessing their broader social implications. This bottom-up empirical approach moves beyond Western-centric viewpoints while provides actionable insights for adaptive governance with Chinese characteristics.\u003c/p\u003e"},{"header":"2. Literature Review","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e\u003ch2\u003e2.1. The Social Construction of Technology (SCOT)\u003c/h2\u003e\u003cp\u003eThe Social Construction of Technology (SCOT), proposed by Pinch and Bijker (1984) within the academic context of critiquing Technological Determinism, fundamentally challenges the notion that technological development follows an inherent, autonomous logic that unidirectionally determines social change (MacKenzie \u0026amp; Wajcman, 1992), as exemplified by McLuhan\u0026rsquo;s (1964) famous assertion that \u0026ldquo;the medium is the message\u0026rdquo;. Rooted in constructivist epistemology, the SCOT framework posits that technological forms emerge as products of negotiations among social groups, highlighting the contingent and multi-directional nature of developmental pathways (Bijker, 1997). This theoretical stance inherits the \u0026ldquo;Strong Programme\u0026rdquo; from the Sociology of Scientific Knowledge (Bloor, 1976), asserting that technology is inherently non-neutral, with its design, application, and meanings inextricably shaped by power relations, cultural values, and social conflicts (Winner, 1986). Overall, the SCOT theoretical system comprises four core analytical dimensions: Social Groups, Interpretive Flexibility, Closure, and Stabilization. These concepts systematically reveal the dynamic processes among multiple actors in technological evolution.\u003c/p\u003e\u003cp\u003eWithin the SCOT framework, social groups are defined as relevant actors with direct/indirect technological connections who share meanings attached to a specific artifact and participate in technological construction. Categorized by their technological relationships and organizational forms, four primary groups emerge: Producers (e.g., engineers, designers, investors, marketers), Advocates (e.g., policymakers, researchers, lobbyists), Users, and some scholars later refer to as Bystanders (Pinch \u0026amp; Bijker, 1984).\u003c/p\u003e\u003cp\u003eTo begin with, producers shape technological forms directly through design, production and marketing. For example, Dunlop\u0026rsquo;s pneumatic tire design simultaneously enhanced cycling comfort and speed performance (Clayton, 2002). Beyond direct production dynamics, advocates exert indirect yet systemic influence on technological configurations through regulatory frameworks. A paradigmatic example emerges from New York State\u0026rsquo;s legislative prohibition against handheld phone operation while driving. This policy intervention effectively institutionalized earphones as mandatory technical prerequisites for mobile device usage in vehicles, fundamentally reconfiguring both material design conventions and sociocultural perceptions of responsible communication practices (Humphreys, 2005a).\u003c/p\u003e\u003cp\u003eCompleting the triad of relevant social groups, users represent the experiential pole of technological interaction. Users are individuals or groups who interact directly with technological artifacts through practice, assigning diverse meanings to it based on contextualized use. Unlike producers or advocates, users are not formally organized but engage in emergent meaning-making through appropriation and adaptation. Their interpretations are shaped by personal needs, cultural norms, and situational demands, which collectively determine the artifact\u0026rsquo;s social legitimacy (Pinch \u0026amp; Bijker, 1984; Humphreys, 2005a).\u003c/p\u003e\u003cp\u003eWhile users actively engage with technologies through situated practices, bystanders, though not directly interacting with a technology, nevertheless shape its social acceptance and cultural meaning through moral judgments, public discourse, or normative critiques (Pinch \u0026amp; Bijker, 1984). Unlike users or producers, bystanders lack material or institutional stakes in the technology itself but influence its trajectory by constructing societal perceptions of its appropriateness, risks, or symbolic value, as seen in 19th-century anti-cycling movements that condemned female riders for violating gender norms, forcing advertisers to redefine bicycles\u0026rsquo; gendered symbolism (Garvey, 1996).\u003c/p\u003e\u003cp\u003eIn this sense, these social groups do not operate in isolation, rather, their dynamic relationships manifest through continuous interactions involving language, usage, and technical structures. On the one hand, systematic power struggles between producers and advocates occur. The former manipulate physical structures through design and capital, while the latter establish normative frameworks via policy discourse (Humphreys, 2005b). These interactions drive technological adaptations between institutional constraints and market imperatives, potentially fostering mandatory standards or alternative solutions. On the other hand, user-bystander interactions form dynamic tension of meaning production: embodied practices generate diverse interpretations, while moral judgments construct social regulations, collectively reshaping technologies\u0026rsquo; legitimacy boundaries.\u003c/p\u003e\u003cp\u003eInterpretive flexibility, SCOT\u0026rsquo;s core analytical tool, reveals how opposing meanings are inscribed onto a single technological artifact. A classic illustration emerges from the historical evolution of bicycles: young male cyclists viewed bicycles as competitive racing apparatuses, while women/elderly users prioritized safety, driving technical iterations from high-wheelers to safety bicycles (Pinch \u0026amp; Bijker, 1984). Mobile phone controversies similarly demonstrate the dynamic: users emphasize communication convenience while bystanders denounce \u0026ldquo;cell yell\u0026rdquo; as a public nuisance. These conflicting interpretations of technology led to design adjustments like vibration modes and text messaging, demonstrating how conflicting user needs shape practical solutions (Humphreys, 2005a). Extending this analysis, MacKay and Gillespie (1992) theorize technological \u0026ldquo;openness\u0026rdquo;\u0026mdash;the degree to which systems permit user reinterpretation\u0026mdash;as a critical mediator of interpretive flexibility. Closed designs (e.g., locked software) maintain power imbalances by restricting customization, whereas open designs (e.g., customizable hardware) let users adapt technologies against standard practices.\u003c/p\u003e\u003cp\u003eClosure and stabilization constitute interconnected yet distinct processes in technological evolution. Closure occurs when social groups agree on a technology\u0026rsquo;s purpose, often through discussions, persuasive strategies, or problem redefinition. Following moments of closure, technologies enter processes of stabilization, whereby provisional agreements solidify into taken-for-granted norms. It unfolds when designs become standardized and widely accepted as \u0026ldquo;natural\u0026rdquo; parts of technological systems (Humphreys, 2005b). While closure reflects momentary social compromises, stabilization develops through repeated agreements over time.\u003c/p\u003e\u003cp\u003eSCOT\u0026rsquo;s conceptualization of closure undergoes crucial paradigm shifts in the internet age. Early research framed closure as a final resolution through stable consensus (Pinch \u0026amp; Bijker, 1984)\u0026mdash;a premise rooted in the relative stability of industrial-era technologies. In the context of digital technologies, Bijker\u0026rsquo;s (1993) notion of \u0026ldquo;degrees of stabilization\u0026rdquo; reconceptualizes closure as a series of temporary compromises under specific socio-temporal conditions, dynamically adjusting with power relations. Humphreys (2005b) advances \u0026ldquo;temporary closure\u0026rdquo; to explicitly manifest SCOT\u0026rsquo;s dynamism, arguing that technological closure in networked societies inherently features locality and recursivity. Each instance of closure simultaneously functions as the resolution of preceding negotiations and the genesis of subsequent controversies. This theoretical shift deconstructs linear evolutionary myths, refocusing analytical attention from static solutions to dynamic power negotiations. In doing so, it marks SCOT\u0026rsquo;s critical transition from a tool of technological historiography tool to a research paradigm for analyzing digital societies.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec4\" class=\"Section2\"\u003e\u003ch2\u003e2.2. Generative AI in Higher Education\u003c/h2\u003e\u003cp\u003eThe integration of generative AI technologies like ChatGPT into higher education has accelerated amidst global digital transformation and post-pandemic reforms in teaching and learning. As universities increasingly adopt hybrid learning models, these tools demonstrate multifaceted applications in assessment systems, pedagogical interactions, and technical skill development. Specifically, universities have begun employing generative AI to enhance various institutional processes\u0026mdash;creating dynamic quizzes (Adiguzel et al., 2023), simulating academic writing scenarios (Kasneci et al., 2023), and providing real-time feedback on student work (Zhao et al., 2023). In programming education, AI-driven code explanation systems (MacNeil et al., 2022) and data science training modules (Moore et al., 2022) demonstrate practical implementations. These applications extend beyond pedagogical functions to contribute to accessibility improvements, for example, by assisting students with disabilities through adaptive interfaces and multilingual support (Kasneci et al., 2023).\u003c/p\u003e\u003cp\u003eThe adoption of generative AI presents transformative opportunities for personalized learning. ChatGPT\u0026rsquo;s capacity to generate tailored educational content aligns with evolving pedagogical trends toward individualized instruction, enabling educators to optimize content delivery based on learner analytics (Zhu et al., 2018). Its ability to overcome language barriers and democratize access to quality education for non-native speakers and underserved populations marks a significant advancement (Kasneci et al., 2023). Technologically, AI-assisted problem decomposition in data science (Moore et al., 2022) and automated grading systems reduce administrative burdens, allowing instructors to focus on higher-order teaching objectives. Social media sentiment analyses further reveal public optimism about AI\u0026rsquo;s potential to enhance educational equity and innovation (Li et al., 2023).\u003c/p\u003e\u003cp\u003eDespite its potential, generative AI introduces systemic risks to academic integrity and pedagogical quality. The technology\u0026rsquo;s ability to produce graduate-level exam responses (Dwivedi et al., 2023) and even pass standardized law exams (Koetsier, 2023) raises fundamental concerns about the validity of traditional assessment frameworks, plagiarism and credential devaluation (AlAfnan et al., 2023; Sok \u0026amp; Heng, 2023). Pedagogically, overreliance on AI-generated content correlates with diminished critical thinking skills (Huang et al., 2025) and oversimplification of complex disciplinary dialogues (Tack \u0026amp; Piech, 2022). These concerns are further exacerbated by technical limitations, including factual inaccuracies (Choi et al., 2023) and algorithmic biases in culturally sensitive contexts (Qadir, 2022). Institutional barriers include high infrastructure costs and ethical dilemmas surrounding data privacy and workforce displacement (Thurzo et al., 2023). Policy responses remain polarized, ranging from outright bans in schools (Johnson, 2023) to proactive ethical guidelines, highlighting the urgent need for balanced regulatory frameworks (Mhlanga, 2023; Rudolph et al., 2023).\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec5\" class=\"Section2\"\u003e\u003ch2\u003e2.3. Research Gaps\u003c/h2\u003e\u003cp\u003eThe Social Construction of Technology (SCOT) theory emphasizes that technological innovation is a process co-shaped by diverse social groups through interactive negotiation (Pinch \u0026amp; Bijker, 1984). While a growing body of scholarship has highlighted ethical risks and pedagogical disruptions associated with generative AI in higher education, much of this literature remains grounded in expert commentary or top-down theoretical reflection (AlAfnan et al., 2023; Kasneci et al., 2023). Emerging empirical studies based on Twitter data have revealed more nuanced, bottom-up perspectives on AI adoption that public discourse on social media encompasses richer dimensions of discussion, such as academic integrity, skill development, and technological limitations (Tlili et al., 2023; Leiter et al., 2023). These discursive practices, which reflect lived experiences of various stakeholders, not only supplements the multifaceted realities of technology adoption (Haque et al., 2022) but also unveils the dynamic interplay of \u0026ldquo;relevant social groups\u0026rdquo; during educational technology development within the SCOT framework (Taecharungroj, 2023).\u003c/p\u003e\u003cp\u003eHowever, current studies predominantly focus on Western social media platforms (Feng et al., 2023), with limited exploration of how role negotiation and consensus-building mechanisms unfold among key stakeholders in Chinese higher education context\u0026mdash;including policymakers, university administrators, educators, students, and technology developers\u0026mdash;in AI implementation. Differences in cultural context and institutional environments may significantly influence the sociotechnical construction pathways. For instance, the unique ecosystem of Chinese higher education system, characterized by the coexistence of administrative dominance and rapid market-driven digitalization, may shape AI adoption models distinct from Western paradigms.\u003c/p\u003e\u003cp\u003eTo address the above-mentioned gap, this study proposes the following research questions:\u003c/p\u003e\u003cp\u003eRQ1: What core themes emerge from discussions about AI in higher education on Chinese social media?\u003c/p\u003e\u003cp\u003eRQ2: How do these discussions reflect the interpretive flexibility of technology among different social groups within the SCOT framework?\u003c/p\u003e\u003cp\u003eRQ3: How are intergroup controversies resolved through closure mechanisms within the SCOT framework, and what implications does such consensus hold for the future trajectory of AI adoption in Chinese higher education?\u003c/p\u003e\u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cdiv id=\"Sec7\"\u003e\n \u003ch2\u003e3.1 Research Framework\u003c/h2\u003e\n \u003cp\u003eThis study adopts a mixed-methods research design to systematically investigate the social construction process of AI usage in higher education as manifested on Chinese social media. The research procedure consists of three stages.\u003c/p\u003e\n \u003cp\u003eFirst, Latent Dirichlet Allocation (LDA) model was employed to conduct topic modelling on the collected social media data, aiming to uncover the core dimensions of public discourse (RQ1). The selection of LDA stems from its unique strength at capturing latent semantic structures within large-scale online discussions (Blei et al., 2003). Unlike traditional manual coding, its unsupervised learning mechanism effectively circumvents researcher bias by identifying naturally emergent thematic patterns through probabilistic distribution models (Taecharungroj, 2023; Li et al., 2024). This approach aligns with the SCOT framework’s emphasis on “interpretive flexibility” research needs, specifically, how technological meanings self-organize through interactions among different social groups.\u003c/p\u003e\n \u003cp\u003eSecond, building upon the SCOT framework, directed content analysis was applied to code social groups within the collected discourse data, examining technological interpretation strategies across distinct actor groups (RQ2).\u003c/p\u003e\n \u003cp\u003eFinally, to answer RQ3, the study integrated the findings from the previous two stages with policy documents and related materials. This synthesis allowed for an analysis of closure mechanisms, that is, how competing interpretations of technology are negotiated, stabilized, or contested, thus providing insight into the ongoing consensus-building pathways for technological controversies and their implications for the future trajectory of AI adoption in Chinese higher education.\u003c/p\u003e\n \u003cp\u003e\u003cstrong\u003eTable 1\u003c/strong\u003e Research Framework: Social Construction of AI in Higher Education\u003c/p\u003e\n \u003cdiv\u003e\n \u003ctable id=\"Tab1\" border=\"1\"\u003e\u003c/table\u003e\n \u003c/div\u003e\n \u003cp\u003e\u003cimg src=\"data:image/png;base64,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\"\u003e\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec8\"\u003e\n \u003ch2\u003e3.2. Data collection\u003c/h2\u003e\n \u003cp\u003eThis study collected publicly available data related to \u003cem\u003eAI-assisted writing\u003c/em\u003e from Sina Weibo Open API v2 over the period from December 20, 2024, to January 30, 2025. Adopting Bruns and Stieglitz’s (2013) hashtag-based data collection framework, which prioritizes precision over generic keyword crawling by leveraging platform-native topic indexing, the dataset is identified as original posts and their corresponding comments under four hashtags (see Table 1). This approach leverages the structural similarities between Weibo and Twitter (X) in hashtag functionality (i.e., centralized topic aggregation and user-driven engagement patterns) to minimize semantic noise and exclude irrelevant or contextually peripheral content, supporting subsequent topic modelling and SCOT-based group classification.\u003c/p\u003e\n \u003ctable id=\"Tab2\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 2\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eSelected Weibo Hashtags for Data Collection\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eHashtag\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eLaunch Date\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eView Count\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e#大学生作业里充满了AI味#\u003c/p\u003e\n \u003cp\u003e#College Students’ Homework is Full of AI-generated Content#\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2024/12/20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e26.42 million\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e#大学教师称学生快失去了原创写作能力#\u003c/p\u003e\n \u003cp\u003e#University Professors Warn Students Are Losing Original Writing Abilities#\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2024//12/20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e1.59 million\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e#AI辅助写作一刀切禁止不现实#\u003c/p\u003e\n \u003cp\u003e#A Blanket Ban on AI-assisted Writing is Unrealistic#\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2024/12/26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e6.97 million\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e#近8成学霸备考离不开AI#\u003c/p\u003e\n \u003cp\u003e#Nearly 80% of Top Students Rely on AI for Exam Preparation#\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e2024/12/25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e15.98 million\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003eTo ensure research validity in keyword selection, as shown in the Table 2, we first retained only hashtags explicitly addressing AI’s application in educational writing assistance (e.g., #大学生作业里充满了AI味#, i.e. #College Students’ Homework is Full of AI-generated Content#) while excluding topics with insufficient semantic relevance (e.g., #AI绘画争议#, i.e. #AI Generated Art Controversy#). This filtering protocol aligns with Bruns and Stieglitz’s (2013) emphasis on hashtag-driven semantic anchoring, which isolates context-specific discourse and reduces the ambiguity inherent in broad keyword queries. Second, following Tufekci’s (2014) observation that a decline in hashtag activity does not equate to discussion cessation of a discussion, we implemented a 30–40 day delayed crawling protocol to capture stabilized user-generated content, thereby improving temporal validity of the data. To mitigate platform disruption risks highlighted by Driscoll and Walker’s (2014) analysis of Twitter API limitations, API requests were capped at 120 calls per minute. Finally, all procedures adhered to the ‘ethical pluralism’ principle for public data reuse (Fiesler \u0026amp; Proferes, 2018), including data anonymization and exclusion of non-consenting private accounts. This protocol yielded 4,750 valid raw data entries (N = 4,750).\u003c/p\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec9\"\u003e\n \u003ch2\u003e3.3. LDA-Based Topic Modeling\u003c/h2\u003e\n \u003cp\u003eLatent Dirichlet Allocation (LDA) (Blei et al., 2003) is a widely-used probabilistic generative model for uncovering latent thematic structures within large document collections. To improve LDA analysis,we first performed essential preprocessing on the collected data before building the model.\u003c/p\u003e\n \u003cdiv id=\"Sec10\"\u003e\n \u003ch2\u003e3.3.1. Data Preprocessing\u003c/h2\u003e\n \u003cp\u003eThe dataset refinement process employed a multi-stage cleansing protocol tailored to the linguistic characteristics of the Sina platform. An initial round of manual curation removed three categories of non-discursive content: (1) numeric repetition artifacts (“111”, “2222”) indicative of bot-generated interactions, (2) platform engagement prompts (e.g., “@xxx”), and (3) formulaic phatic expressions (e.g., “Good afternoon”) that lack analytical relevance. Subsequently, automated processing via Python’s \u003cstrong\u003ere\u003c/strong\u003e module systematically eliminated emojis, URLs, and special symbols (#, @), addressing the noise prevalence characteristic of Chinese Sina corpora. This hybrid cleaning pipeline validated 4,122 texts (86.8% of 4,750 initial entries), achieving an optimal balance between noise reduction and semantic preservation.\u003c/p\u003e\n \u003cp\u003eThe final preprocessing step prior to formal text segmentation involved noise reduction through application of a customized stop-word list. To build this resource, three authoritative Chinese lexical resources were integrated: the Baidu, SCU, and HIT stopword lists. Following methodological precedents in social media text processing (Xie et al., 2019), the composite lexicon combines 1) standard linguistic stopwords, 2) platform-specific noise patterns (e.g., “点击展开全文”, i.e., \u003cem\u003eRead more\u003c/em\u003e), and 3) domain-specific formulaic expressions, achieving comprehensive filtering for contemporary web-based Mandarin corpora. This comprehensive approach ensured that filtering was suited to the contemporary web-based Mandarin discourse environment.\u003c/p\u003e\n \u003cp\u003eThe Chinese text segmentation process employed the \u003cstrong\u003ejieba\u003c/strong\u003e tokenizer in precision mode, a widely-adopted tool for handling Mandarin’s character-bound morphology (Peng et el., 2015; Liu et al., 2020; Zhang et al., 2020; Guo et al., 2024). Unlike English-oriented tools such as \u003cstrong\u003eNLTK\u003c/strong\u003e, which rely on whitespace delimiters for word seperation(Loper \u0026amp; Bird, 2002), \u003cstrong\u003ejieba\u003c/strong\u003e’s architecture is specifically designed to resolve the agglutinative nature of Chinese script—a critical advantage given the absence of inherent word separators in written Mandarin (Huang et al., 2017).\u003c/p\u003e\n \u003cp\u003eDomain adaptation was achieved through proactive lexicon expansion, where AI-specific terminology (e.g., “ChatGPT”, “文心一言/ERNIE Bot”) was manually integrated into jieba’s segmentation rules. This adaptation enabled the preservation of semantically coherent compound terms, aligning with established practices in Chinese NLP for topic modeling (Zhou et al., 2005). The inclusion of a controlled vocabulary ensured LDA’s ability to capture domain-specific semantic patterns, as validated in probabilistic topic models (Blei et al., 2003; Griffiths \u0026amp; Steyvers, 2004).\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec11\"\u003e\n \u003ch2\u003e3.3.2. LDA Modeling\u003c/h2\u003e\n \u003cp\u003eWe conducted LDA analysis using \u003cstrong\u003eGensim\u003c/strong\u003e’s specialized \u003cstrong\u003eLdaMulticore\u003c/strong\u003e implementation (Řehůřek \u0026amp; Sojka, 2010), a parallel computing variant designed for efficient topic modeling on multi-core systems. This implementation preserves the theoretical foundations of standard LDA while substantially accelerating computation via distributed workload processing, utilizing optimization strategies such as data-aware scheduling and pipeline parallelism as demonstrated in distributed LDA frameworks (Liu et al., 2011). Computational efficiency was prioritized by configuring three parallel processes (workers = 3) and 20 training iterations (passes = 20), while a fixed random seed (random_state = 7) ensured full reproducibility of the stochastic initialization. Text preprocessing involved generating a dictionary through tokenization and constructing a document-term matrix using bag-of-words (BoW) vectorization (Zhang et al., 2010). Rather than imposing a predefined topic count, the model progressively refined latent thematic structures through repeated passes over the corpus.\u003c/p\u003e\n \u003c/div\u003e\n \u003cdiv id=\"Sec12\"\u003e\n \u003ch2\u003e3.3.3. Optimal Topic Number Selection\u003c/h2\u003e\n \u003cp\u003eTo determine the optimal number of topics (k) in LDA models, a hybrid approach combining perplexity minimization, quantitative coherence metrics (e.g., C_v, U_mass), and human evaluation of topic interpretability was employed. Topic coherence quantifies the semantic consistency of a topic’s top terms by assessing their co-occurrence patterns in reference texts (Lau et al., 2014). Statistical measures metrics calculate how frequently terms appear together relative to their individual occurrence rates, with higher coherence scores indicating stronger contextual associations between terms. Robust co-occurrence patterns typically reflects human-interpretable themes (Röder et al., 2015). Thus, topic coherence offers distinct advantages for evaluating topic models, particularly in specialized domains. By quantifying the semantic relatedness of high-probability terms within topics, coherence metrics provide interpretability guarantees that perplexity inherently lacks (Chang et al., 2009).\u003c/p\u003e\n \u003cp\u003eTo evaluate topic coherence in our LDA model, we employed the \u003cstrong\u003eC_v\u003c/strong\u003e coherence metric (Röder et al., 2015). First, we trained LDA models with varying topic counts (1–30) and calculated their \u003cstrong\u003eC_v\u003c/strong\u003e coherence scores. Next, we identified the optimal topic count by locating the \u003cstrong\u003eC_v\u003c/strong\u003e score peak (indicating maximal semantic coherence) and manually inspecting topic keywords for interpretability. As shown in Fig. 1, the coherence curve reaches a peak coherence score of 0.532 at 14 topics, beyond which scores plateau with diminishing returns. Manual inspection of topic keywords confirmed that 14 topics preserved thematic granularity (reflecting nuanced opinion patterns) while avoiding semantic redundancy. Therefore, k = 14 was selected as the optimal topic count, balancing statistical validity and human interpretability.\u003c/p\u003e\n \u003c/div\u003e\n\u003c/div\u003e\n\u003cdiv id=\"Sec13\"\u003e\n \u003ch2\u003e3.4. Content Analysis\u003c/h2\u003e\n \u003cp\u003eTo address RQ2, we first conducted a content analysis in formed by the SCOT framework. A codebook was developed through the integration of the definitions of social groups proposed by in prior SCOT literature (Humphreys, 2005b) and the specific contextual characteristics of AI applications in higher education. Two researchers independently coded a subset of 2895 comments based on the preliminary codebook to identify the different group tendencies implied in the contents. Before formal coding, three rounds of rounds were conducted to refine coding rules and resolve discrepancies, achieving high inter-coder reliability with a Cohen’s κ of 0.906 (indicating substantial agreement). To ensure analytical clarity, each comment was assigned exclusively to one social group category and no overlaps were allowed. Therefore, the final coding results revealed the distribution of social groups within the dataset. Following the completion of coding, textual analysis was conducted separately for comments classified within each of the four social group categories. Table 3 presents the final five-category coding scheme, specifying the operational definitions of four types of social groups in this study and a residual category, which includes comments that do not clearly fall into any of the defined social-group categories.\u003c/p\u003e\n \u003ctable id=\"Tab3\" border=\"1\"\u003e\n \u003ccaption language=\"En\"\u003e\n \u003cdiv\u003eTable 3\u003c/div\u003e\n \u003cdiv\u003e\n \u003cp\u003eClassification Schema for Social Groups with Residual Category in Content Analysis\u003c/p\u003e\n \u003c/div\u003e\n \u003c/caption\u003e\n \u003cthead\u003e\n \u003ctr\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eGroup\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eOperational Definition\u003c/p\u003e\n \u003c/th\u003e\n \u003cth align=\"left\"\u003e\n \u003cp\u003eExample\u003c/p\u003e\n \u003c/th\u003e\n \u003c/tr\u003e\n \u003c/thead\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eProducers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGroups involved in the development, funding, or marketing of AI technologies (e.g., engineers, investors). Through technical specifications and commercialization strategies, they determine the core capabilities of AI tools, including both dedicated educational AI systems and general-purpose platforms later adapted for pedagogy (e.g., ChatGPT repurposed for academic writing).\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e“The AI features on [Brand]’s tablet are exceptionally user-friendly. Let’s go try it out together!” “Effortlessly take notes and organize materials with AI, [Brand]’s tablet model helps you stand out from information overload!”\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eAdvocates\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGroups actively shaping the legitimacy and institutional boundaries of AI applications in higher education through academic discourse, media engagement, or policymaking (e.g., policymakers, researchers, professors). Their stances (pro or con) redefine the sociotechnical significance of AI by framing its permissible uses within institutional teaching, learning, and administrative practices.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e“A teacher remarked, ‘AI has made students’ work carry a mechanical feel, lacking independent thinking and originality.’ ” “Many university faculty members report that ... though widespread adoption of AI tools has boosted academic productivity, students’ works are increasingly devoid of vitality, with originality facing severe challenges.”\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eUsers\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGroups directly interacting with AI tools (whether purpose-built for education or not) within higher education contexts. Their adaptive practices drive technical refinements through educational repurposing, while redefining the pedagogical utility and context-specific ethical norms of AI tools.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e“How can you know that I’m AI-generating my term paper that’s had me stuck for hours?” “ChatGPT has become a trusted partner for university students and postgraduates worldwide.”\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eBystanders\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eGroups comprising 1) non-users (e.g., parents unaware of AI, pedagogy scholars) whose perceptions are mediated through social ties, and 2) AI users outside academia (e.g., general consumers).Their perspectives (e.g., media critiques of AI existential risks) or cross-domain engagements (e.g., corporate AI skill requirements) indirectly mediate the adoption patterns and social legitimacy of AI in higher education.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e“Honestly, I’ve started outsourcing repetitive and formulaic tasks to AI text generators these days... Can’t blame college students either...”\u003c/p\u003e\n \u003cp\u003e“Artificial Intelligence is the future itself, we must seize the moment. This revolution stands as nothing less than the Industrial Revolution or the Internet Revolution!”\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eResidual Category\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003eComments that cannot be clearly assigned to any of the above categories due to vagueness, neutrality, or irrelevance to educational contexts.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd align=\"left\"\u003e\n \u003cp\u003e“I’m not quite getting it.” “Is AI making us lazy or leveling up our skills? Let’s hear firsthand from university students across campuses!”\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n \u003c/table\u003e\n \u003cp\u003e\u003cem\u003eNote.\u003c/em\u003e Two processing measures applied:\u003c/p\u003e\n \u003cp\u003ea) All quoted comments remain anonymous;\u003c/p\u003eb) Commercial brand names quoted in producers’ comments was replaced with the neutral identifier [Brand].\n\u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec15\" class=\"Section2\"\u003e\u003ch2\u003e4.1. LDA-Based Topic Modeling Results\u003c/h2\u003e\u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the results of the LDA analysis. Based on the C_v coherence metric, the optimal number of topics was identified as 14. Subsequently, two researchers independently reviewed the top 30 high-frequency keywords for each topic (translated into English terminology), and through discussion, assigned thematic labels to each topic. To enhance interpretability, thematically related topics were further grouped into higher-order conceptual categories. This table also displays the most relevant high-frequency words associated with the theme of each topic.\u003c/p\u003e\u003cp\u003eThe first category focuses on the instrumental application of AI technology in learning scenarios, primarily involving usage scenarios and underlying motivations. The feature words \u0026ldquo;writing/exam preparation\u0026rdquo; in Topic 12 indicate that AI has permeated the entire learning process, from information retrieval and content generation to exam preparation. Topic 2 further reveals the core motivations behind college students\u0026rsquo; reliance on AI for academic tasks. High-frequency terms such as \u0026ldquo;thesis/assignments/low-effort courses\u0026rdquo; indicate that AI is predominantly applied to writing course papers and handling routine coursework for \u0026ldquo;low-effort courses\u0026rdquo;. The term \u0026ldquo;watered-down courses (水课) \u0026rdquo; originates from the Chinese context, reflecting students\u0026rsquo; subjective evaluation of courses perceived to offer low knowledge acquisition or skill application. Data indicate that students tend to employ AI tool to complete assignments for such courses to increase efficiency, especially when facing repetitive or low-stake assignments. Furthermore, users generally cite that excessive workloads and tight deadlines as key drivers of AI use, as reflected by terms such as \u0026ldquo;days/weeks\u0026rdquo;. These temporal markers suggest that during finals periods, they often need to complete large volumes of thesis assignments within two weeks or even just a few days. Notably, Topic 9 reveals that AI usage is not limited to struggling students. In fact, many top-performing students adopt AI to optimize their learning strategies, with users reporting that academically successful peers around them routinely integrate AI into their study routines. This finding suggests a normalization of AI-assisted learning across academic performance levels.\u003c/p\u003e\u003cp\u003eThe second category highlights controversies arising from the application of AI technology in academia. Topics 5 and 6 reflect both advantages and disadvantages. The juxtaposition of \u0026ldquo;generation/templates/dependence\u0026rdquo; with \u0026ldquo;convenience/time-saving\u0026rdquo; demonstrates AI\u0026rsquo;s efficiency improvements in exam preparation while raising concerns about learning capability degradation due to overreliance on technological. A particularly notable issue is the homogenization in student submissions, as assignments generated using AI often follow identifiable structural templates. Some instructors report being able to even identify specific AI tools used for the assignments, reflecting a growing awareness of stylistic convergence in student work. Topic 14, which features keywords \u0026ldquo;innovation ability/weakening/coping\u0026rdquo; directly underscores the potential structural deficiencies in academic capabilities caused by heavy dependency. Students who use AI to complete assignments may fail to develop targeted competencies, ultimately weakening their innovation capacity over time. In parellel, defined by terms such as \u0026ldquo;Integrity/Risks\u0026rdquo;, Topic 11 exposes how AI assistance may undermine fundamental values of academic ethics, particularly integrity. This constitutes a primary rationale for advocating prohibitions on AI in academic contexts.\u003c/p\u003e\u003cp\u003eThe third category predominantly reflects users\u0026rsquo; positive evaluations of AI technology value. Terms \u0026ldquo;efficiency/intelligence/convenience\u0026rdquo; in Topic 1 and \u0026ldquo;productivity/cost-effectiveness\u0026rdquo; in Topic 3 collectively construct a technical efficacy assessment framework, indicating widespread recognition of AI\u0026rsquo;s breakthrough effects in optimizing workflows and enhancing resource utilization. Topic 4\u0026rsquo;s high-frequency terms \u0026ldquo;development/excellent\u0026rdquo; suggest societal optimism regarding AI\u0026rsquo;s technological evolution trajectory. Meanwhile, \u0026ldquo;omnipresent\u0026rdquo; particularly demonstrates AI\u0026rsquo;s high penetration rate in China, reinforcing its strategic position as a new productivity tool in the educational system.\u003c/p\u003e\u003cp\u003eThe fourth category reveals universities\u0026rsquo; governance challenges in addressing AI technological impacts. Represented by terms \u0026ldquo;universities/prohibition/originality\u0026rdquo;, Topic 8 points to typical regulatory measures, as exemplified by Fudan University\u0026rsquo;s written policy banning AI use in thesis writing to preserve academic originality. In contrast, Topic 10 features terms including \u0026ldquo;one-size-fits-all/improvement/efficiency\u0026rdquo;, which exposes public discontent with overly rigid regulatory practices. \u0026ldquo;One-size-fits-all\u0026rdquo; ban are often deemed impractical or misaligned with actual educational needs in the public discussion. Consequently, current AI policy-making in higher education faces dual pressures of keeping pace with technological evolution and upholding academic ethics.\u003c/p\u003e\u003cp\u003eThe last category examines the forward-looking impacts of AI technological development. As characterized by \u0026ldquo;era/intelligence/satisfaction\u0026rdquo;, Topic 7 indicates a framework connecting technological and social evolution, emphasizing AI\u0026rsquo;s role in steering contemporary society toward intelligent societal transformation. Within this framing, AI is portrayed not merely as a tool, but as a defining force of the current era, with some users even asserting that the we are already living in the AI era. Focusing specifically in the education domain, Topic 13 includes high-frequency terms such as \u0026ldquo;education/future/technology\u0026rdquo;, which indicate a consensus that AI urgently requires and will profoundly engage in reconstructing pedagogical systems, thereby driving reforms to address existing issues.\u003c/p\u003e\u003cp\u003e\u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab4\" border=\"1\"\u003e\u003ccaption language=\"En\"\u003e\u003cdiv class=\"CaptionNumber\"\u003eTable 4\u003c/div\u003e\u003cdiv class=\"CaptionContent\"\u003e\u003cp\u003eLDA Topic Clusters on AI in Higher Education\u003c/p\u003e\u003c/div\u003e\u003c/caption\u003e\u003ccolgroup cols=\"4\"\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c2\" colnum=\"2\"\u003e\u003c/div\u003e\u003cdiv align=\"char\" char=\".\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e\u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e\u003cthead\u003e\u003ctr\u003e\u003cth align=\"left\" colname=\"c1\"\u003e\u003cp\u003eCategory\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTheme\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c3\"\u003e\u003cp\u003eTopic No.\u003c/p\u003e\u003c/th\u003e\u003cth align=\"left\" colname=\"c4\"\u003e\u003cp\u003eTopic Feature Words\u003c/p\u003e\u003c/th\u003e\u003c/tr\u003e\u003c/thead\u003e\u003ctbody\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003eAI Academic Assistance Practice\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eCollege Students Using AI to Complete Assignments\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e2\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eCollege Students, AI, Thesis, Assignments, Low-Effort Courses, Days, Weeks\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAI-Assisted Full-Cycle Learning\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e12\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, College Students /Materials, Assistance, Writing, Efficiency, Exam Preparation\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eTop Students Optimizing Study Strategies with AI\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e9\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Top Students, Exam Prep, High Efficiency, Relaxation, Thesis\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"3\" rowspan=\"4\"\u003e\u003cp\u003eAI Academic Assistance Controversies​​\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003ePros and Cons of AI Academic Assistance\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e5\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Thesis, Generation, ChatGPT, Polishing, Tools, Time-Saving, Templates\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e6\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Problem-Solving, Dependence, Tutorial Assistance, Convenience, Exams, Learning\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e14\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Practical Use, Dependence, Innovation Ability, Weakening, Savior, Coping\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAI Dependence and Academic Ethics Risks\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e11\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Learning, Efficiency, Thinking, Universities, Academic, Integrity\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"2\" rowspan=\"3\"\u003e\u003cp\u003ePositive Evaluation of AI Technical Efficiency​​\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eAI Enhancing Learning \u0026amp; Work Efficiency\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e1\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Daily Use, Development, Technology, Top Students, Work, Intelligence, Convenience\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eUser Recognition of AI Technology\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e3\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eOverpowered, Functions, Sharing, Productivity, Cost-Effectiveness, Efficiency\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e4\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, Powerful, Development, Technology, Intelligence, Useful, Excellent, Omnipresent\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003ePolicies and Controversies of AI-Generated Content in Universities​​\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eUniversity Regulations on AI-Generated Content\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e8\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eUniversities, Originality, Independent Thinking, Fudan University, Prohibition\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eEfficiency vs. \u0026ldquo;One-Size-Fits-All\u0026rdquo; AI Policies\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e10\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eWriting, One-Size-Fits-All, Prohibition, Originality, Policies, Efficiency, Improvement\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c1\" morerows=\"1\" rowspan=\"2\"\u003e\u003cp\u003eAI Era and Future Trends​​\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eValue of AI in the Intelligent Era\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e7\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, User-Friendly, Era, Intelligence, Satisfaction\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003ctr\u003e\u003ctd align=\"left\" colname=\"c2\"\u003e\u003cp\u003eFuture Trends of AI in Education\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e\u003cp\u003e13\u003c/p\u003e\u003c/td\u003e\u003ctd align=\"left\" colname=\"c4\"\u003e\u003cp\u003eAI, College Students, Education, Significance, Future, ​​\u003c/p\u003e\u003cp\u003e\u0026ldquo;Watered-down\u0026rdquo; courses, Technology, Teachers\u003c/p\u003e\u003c/td\u003e\u003c/tr\u003e\u003c/tbody\u003e\u003c/colgroup\u003e\u003c/table\u003e\u003c/div\u003e\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec16\" class=\"Section2\"\u003e\u003ch2\u003e4.2. Content Analysis Results\u003c/h2\u003e\u003cp\u003eThe second research question examines how discussions on social platforms reflect the interpretive flexibility of technology (specifically AI usage in higher education) among different social groups. Coding results indicate the proportional representation of each group\u0026rsquo;s voices within the sampled discussion data. Specifically, bystanders accounted for the highest proportion of comments (73.75%), while advocates were the least represented group (0.83%). Although there was a significant gap between groups, users ranked second with 14.20%, followed by producers in third place at 8.46%. The residual category (unclassifiable comments) comprised 2.76% of the total.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec17\" class=\"Section2\"\u003e\u003ch2\u003e4.3. Textual Analysis Results\u003c/h2\u003e\u003cp\u003eThe textual analysis reveals distinct interpretive patterns across four social groups within the SCOT framework regarding AI in education. Results indicate a clear contrast: while producers predominantly framed AI through promotional discourse, emphasizing its benefits and efficiency, advocates raise regulatory concerns via mainstream media. Users (students) expressed strong practical endorsement rooted in academic pressures and efficiency considerations and bystanders presented a spectrum of attitudes spanning from positive acceptance to cautious neutrality. Crucially, these differentiated narratives directly exemplify the interpretive flexibility within the SCOT framework, illustrating how each social group assigns meanings to educational AI based on their positions: producers (commercial interests), advocates (institutional authority), users (experiential needs), and bystanders (observational perspectives). The following sections detail these variation patterns.\u003c/p\u003e\u003cdiv id=\"Sec18\" class=\"Section3\"\u003e\u003ch2\u003e4.3.1. Producers\u003c/h2\u003e\u003cp\u003eMarketing content for a brand\u0026rsquo;s tablet with AI-assisted learning systems constitutes the vast majority (98.8%) of producers\u0026rsquo; comments. Specifically, these promotional messages took two main approaches. The first approach is using brief, impactful slogans. These simple, straightforward phrases highlight the product\u0026rsquo;s key benefit: tablet with AI enhanced system (Example: \u0026ldquo;The new favorite of top students! AI boosts exam prep, doubles efficiency!\u0026rdquo;). Second, beyond explicit promotional slogans, producers also adopt a more simulated grassroots approach by mimicking the language style of student users. High-frequency words in this type of messaging are \u0026ldquo;top student\u0026rdquo; (学霸, 53.1%, n\u0026thinsp;=\u0026thinsp;130), \u0026ldquo;study\u0026rdquo; (学习, 44.5%, n\u0026thinsp;=\u0026thinsp;109), and \u0026ldquo;exam preparation\u0026rdquo; (备考, 38.8%, n\u0026thinsp;=\u0026thinsp;95). Collectively, these messages promote the idea that \u0026ldquo;top students are all using AI for learning.\u0026rdquo; (Example: \u0026ldquo;AI is so useful, no wonder top students can\u0026rsquo;t do without it! I also want to use AI\u0026rsquo;s power to become one of them!\u0026rdquo;) The overall tone is enthusiastic and aspirational, often marked with frequent use of exclamation marks (43.7%, n\u0026thinsp;=\u0026thinsp;107), seeking to generate emotional resonance among student readers. When describing the AI\u0026rsquo;s educational role, these messages often employ expressions like \u0026ldquo;a top student\u0026rsquo;s secret weapon,\u0026rdquo; \u0026ldquo;smart tool,\u0026rdquo; or \u0026ldquo;helpful assistant.\u0026rdquo; In summary, within the producers\u0026rsquo; messaging, the use of AI in education is portrayed in an overwhelmingly positive light.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec19\" class=\"Section3\"\u003e\u003ch2\u003e4.3.2. Advocates\u003c/h2\u003e\u003cp\u003eAlthough advocates represent the smallest portion in the dataset, their commentary is mainly drawn from mainstream media outlets, which grants them disproportionate reach and influence. Notably, one major verified account evens served as the host (the originator) of the Weibo topic: #College Students\u0026rsquo; Homework is Full of AI-generated Content#. Comments in this category consist primarily of news reports, supplemented by a smaller number of personal statements from university educators. Typically, the news reports cite the opinions of university teachers or educational institutions to express their stance on AI use in education. For example, they quote teachers interviewed saying: \u0026ldquo;The homework submitted by students is full of AI-generated content, lacking any traces of independent thinking or originality.\u0026rdquo; Moreover, these pieces primarily focus on the negative impacts AI use has on students. Additionally, new policies implemented by Chinese universities regarding AI are frequently mentioned in this data. For instance, Fudan University issued regulations titled \u0026ldquo;Guidelines for the Use of Artificial Intelligence Tools in Undergraduate Thesis Writing\u0026rdquo;, which was framed as a leading institutional response to emerging ethical dilemmas. The coverage further noted that \u0026ldquo;many universities in China have started exploring the boundaries of AI technology applications.\u0026rdquo; Overall, advocates\u0026rsquo; viewpoint towards AI use in higher education is predominantly negative, focusing mainly on restrictions and limitations.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec20\" class=\"Section3\"\u003e\u003ch2\u003e4.3.3. Users\u003c/h2\u003e\u003cp\u003eUser-generated comments originate almost entirely from students of in higher education (undergraduates or postgraduates), and the vast majority (92.0%) express support for the use of AI in education. Only 0.07% (n\u0026thinsp;=\u0026thinsp;28) of users comment with a nuanced view of AI usage. An even smaller proportion, 0.01% (n\u0026thinsp;=\u0026thinsp;5), hold a negative stance after personal experience with AI tools These critical voices primarily cited technical shortcomings, especially in academic writing tasks, where AI-generated content was described as unreliable due to fabricated references or inaccurate outputs.\u003c/p\u003e\u003cp\u003eAmong supportive users, 24.1% provided reasons for their favorable stance. These comments primarily focused on using AI-generated text for completing assignments or writing papers. Three main reasons emerged. First, users emphasized the overwhelming workload, especially during exam periods. They argued that using AI is necessary for coping with overwhelming assignments so that they can have time to review for finals. A typical comment stated: \u0026ldquo;Twelve papers in one month, for different courses. I don\u0026rsquo;t believe anyone could realistically write them all by themselves...\u0026rdquo; Second, dissatisfaction was expressed regarding low-value, outdated courses and tedious tasks. Users prefer using AI under such contexts to save time. One student\u0026rsquo;s comment captured this: \u0026ldquo;Do teachers for these \u0026lsquo;watered-down\u0026rsquo; courses(水课, institutional compulsory courses or courses that pedagogically void yet credit-bearing) even read the papers we write meticulously? Since they don\u0026rsquo;t, why waste our time? How many unnecessary courses are there in university? We see the same outdated slides, older than the professors, year after year, how can they then criticize students for lacking innovative ideas in their papers?\u0026rdquo; Third, users reported feeling incapable of writing a paper due to lack of instruction or expertise. One user shared: \u0026ldquo;In my freshman \u0026lsquo;Innovation\u0026rsquo; class, the teacher assigned a 3,000-word paper on infectious diseases, requiring proper citations, all due in a week. I\u0026rsquo;ve only taken the class for one semester, not years, what am I supposed to use besides AI?\u0026rdquo; Other reasons included using AI to improve learning efficiency, such as for data organization and generating mind maps. Additionally, some users expressing approval used rhetorical questions or exclamations for emphasis, e.g., \u0026ldquo;So what?\u0026rdquo; or \u0026ldquo;Stop policing college students!\u0026rdquo;\u003c/p\u003e\u003cp\u003eIn conclusion, users strongly affirm AI\u0026rsquo;s role in education. Their expressions of support are grounded in their personal academic struggles and reflect adaptive rationales for AI use.\u003c/p\u003e\u003c/div\u003e\u003cdiv id=\"Sec21\" class=\"Section3\"\u003e\u003ch2\u003e4.3.4. Bystanders\u003c/h2\u003e\u003cp\u003eSince bystanders include both people who have used AI (but not in education) and those who have never used it, the attitudes captured in these comments are not limited to AI in education; rather, they cover general opinions about AI itself. These attitudes were categorized into four groups: positive, negative, nuanced (balanced view), and no clear stance.\u003c/p\u003e\u003cp\u003eQuantitative coding shows that over half (51.4%, n\u0026thinsp;=\u0026thinsp;1098) held a positive attitude towards AI. When describing AI\u0026rsquo;s benefits, one type of comment offered general praise (38.9%, n\u0026thinsp;=\u0026thinsp;427). High-frequency words here included \u0026ldquo;good\u0026rdquo; (不错, n\u0026thinsp;=\u0026thinsp;119), \u0026ldquo;smart\u0026rdquo; (智能, n\u0026thinsp;=\u0026thinsp;76), \u0026ldquo;impressive\u0026rdquo; (厉害, n\u0026thinsp;=\u0026thinsp;62), \u0026ldquo;easy to use\u0026rdquo; (好用, n\u0026thinsp;=\u0026thinsp;55), \u0026ldquo;powerful\u0026rdquo; (强大, n\u0026thinsp;=\u0026thinsp;51), \u0026ldquo;practical\u0026rdquo; (实用, n\u0026thinsp;=\u0026thinsp;51), and \u0026ldquo;useful\u0026rdquo; (有用, n\u0026thinsp;=\u0026thinsp;13). Another type focused specifically on acknowledging AI\u0026rsquo;s ability to improve efficiency (23.1%, n\u0026thinsp;=\u0026thinsp;254), using words like \u0026ldquo;convenient\u0026rdquo; (方便, n\u0026thinsp;=\u0026thinsp;142), \u0026ldquo;efficiency\u0026rdquo; (效率, n\u0026thinsp;=\u0026thinsp;87), and \u0026ldquo;handy\u0026rdquo; (便利, n\u0026thinsp;=\u0026thinsp;25).\u003c/p\u003e\u003cp\u003eBy contrast, negative attitudes constituted a relatively small share (4.5%, n\u0026thinsp;=\u0026thinsp;98). These critical comments expressed several key concerns. Some pointed out that abusing AI could weaken human capabilities (e.g., \u0026ldquo;Choosing to be a happy fool simply because your thinking ability can\u0026rsquo;t match AI is the beginning of self-destruction.\u0026rdquo;). Others highlighted flaws in AI-generated content (e.g., \u0026ldquo;Mechanical AI writing fails to move hearts.\u0026rdquo;). Additionally, some employed a human vs. machine narrative, expressing fears about AI\u0026rsquo;s potential threat (e.g., \u0026ldquo;If AI runs out of control, who will fix it?\u0026rdquo;).\u003c/p\u003e\u003cp\u003eA moderate number of comments expressed nuanced or balanced views (16.1%, n\u0026thinsp;=\u0026thinsp;345). These fell into two subgroups. The first acknowledged both AI\u0026rsquo;s advantages and disadvantages. The second, comprising 27.5% (n\u0026thinsp;=\u0026thinsp;95), emphasized the need for responsible or appropriate AI use, noting that AI is powerful but must be applied thoughtfully and ethically by human users.\u003c/p\u003e\u003cp\u003eFinally, 27.8% of comments (n\u0026thinsp;=\u0026thinsp;594) exhibited no explicit stance. These comments either marveled at AI\u0026rsquo;s omnipresence (e.g., \u0026ldquo;AI really is in every part of our work and life.\u0026rdquo;) or acknowledged AI as an irresistible trend (e.g., \u0026ldquo;It\u0026rsquo;s the era of AI for everyone now.\u0026rdquo;). Further, some shared experiences suggesting AI\u0026rsquo;s importance in future workplaces (e.g., \u0026ldquo;It\u0026rsquo;s okay, working people do this too. It\u0026rsquo;s just about learning and using it a few years ahead.\u0026rdquo;).\u003c/p\u003e\u003cp\u003eIn summary, bystanders demonstrated a complex and multilayered orientation toward AI. While a clear majority viewed it positively, critical concerns and calls for reasonable and ethical usage were also present.\u003c/p\u003e\u003c/div\u003e\u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eWithin the data collected, producers almost exclusively frame AI technology through the singular narrative of \u0026ldquo;efficiency enhancement\u0026rdquo; to promote AI-integrated educational tools. Their marketing discourse centers on a tablet with AI learning features, constantly emphasizing in their advertisements that it can transform learners into \u0026ldquo;top students.\u0026rdquo; For example, slogans like \u0026ldquo;The secret weapon for top students, doubling efficiency\u0026rdquo; are directly used. Or, disguised as students, they post comments such as \u0026ldquo;All the top students are using it, I want to try it too\u0026rdquo; to stimulate peer endorsements. This strategy precisely taps into the group psychology in China\u0026rsquo;s competitive educational environment: high achievers are portrayed as role models, while average students\u0026rsquo; fear of \u0026ldquo;falling behind\u0026rdquo; drives herd-like consumption. It\u0026rsquo;s clear that when interpreting AI technology, producers deliberately choose the single narrative of an \u0026ldquo;efficiency tool,\u0026rdquo; avoiding potential controversies about AI in education. They package their product as a \u0026ldquo;must-have learning device\u0026rdquo; with no downsides. This strategy excessively amplifies the product\u0026rsquo;s advantages, essentially lowering the cognitive barrier for consumers by simplifying the technology\u0026rsquo;s semantic meaning. Ironically, this marketing logic also reflects a real contradiction in China\u0026rsquo;s education technology market: despite producers claiming that \u0026ldquo;AI use represents an education revolution\u0026rdquo;, they continue to anchor technological innovation within the utilitarian framework of measurable improvements in test performances.\u003c/p\u003e\u003cp\u003eIn contrast to producers\u0026rsquo; enthusiastic promotion of AI across all learning scenarios, advocates, including universities, faculties and educational departments adopt a markedly more cautious stance to restrict AI use in higher education. Their chosen narrative for AI is the exact opposite of the producers\u0026rsquo;, emphasizing academic ethics risks and the danger of AI replacing human capabilities. This attitude is not rooted in anti-technological sentiment but rather stems from problems observed in actual teaching practice. University instructors have reported that student submissions increasingly exhibit a \u0026ldquo;distinctive AI tone, lacking traces of independent thought,\u0026rdquo; indicating that concerns about AI\u0026rsquo;s impact on learning originality have materialized into genuine issues. The measures taken by advocates mainly involve setting rules. For instance, institutions such as Fudan University (2024) have issued AI Use Guidelines for Graduation Theses, whose core idea is to define clear boundaries for technology use, prioritizing restrictions on key activities like graduation theses. This approach reflects a practical consideration in China\u0026rsquo;s educational management: since a complete AI ban is impossible, policies should prioritize preserving the academic integrity of academic evaluation. However, the problem is that these local restrictions clash with the \u0026ldquo;all-scenario use\u0026rdquo; pushed by producers. Through product design and advertising, producers implicitly encourage students to rely on AI for routine homework, class notes, exam preparation, and other learning tasks. If students consistently use AI to take over basic learning tasks, their independent thinking skills may decline. This not only conflicts with the advocates\u0026rsquo; rules forbidding AI use in theses but also risks students gradually losing the ability to complete academic work independently. In effect, the advocates\u0026rsquo; fear of \u0026ldquo;innovation ability weakening\u0026rdquo; is being precisely validated through this daily penetration of AI.\u003c/p\u003e\u003cp\u003eThe vast majority of users (students) support the use of AI in education. This attitude is not blind enthusiasm for technology, but a survival strategy within the structural contradictions of Chinese higher education. Specifically, students cite three main reasons for using AI: overwhelming academic pressure, being assigned tasks beyond their current ability, and silent resistance against inefficient or outdated courses. These reasons reveal problems in the current education system\u0026rsquo;s instructional design, task distribution, and evaluation mechanisms. When faced with excessive tasks such as \u0026ldquo;12 different course papers in one month\u0026rdquo;, AI is seen as a tool to boost efficiency. When first-year students are required to write 3000-word specialized papers far exceeding their knowledge base in a week, AI serves as a bridge to fill that capability gap. For inefficient courses where \u0026ldquo;professors reuse decade-old slides\u0026rdquo;, using AI to write assignments becomes a protest against poor teaching. Students are aware of the ethical risks of AI but prioritize its utilitarian value to cope with their situation. The core motivation is the disconnection between goals and methods in teaching design. Assignments meant to serve skill development are instead used as quantitative metrics in the student evaluation system. To avoid failing or to get better grades, many students turn to AI.\u003c/p\u003e\u003cp\u003eThe differing interpretations of AI use among bystanders reflect how the public, drawing on diverse experiences, forms distinct cognitive frameworks about AI\u0026rsquo;s value and risks during the the integration of technology into society. Most affirm the necessity of AI, recognizing its substantial boost to life and work efficiency. This favorable view stems from AI\u0026rsquo;s significant advantages in areas like information processing and process optimization. Especially in the workplace, AI has become a core element in reshaping productivity, instilling a crisis mindset of \u0026ldquo;lagging behind if not using AI\u0026rdquo; in the public. However, many hold a dialectical position, always carefully weighing the boundaries of AI use. Concerns arise when AI evolves from an assistant tool into a potential decision-maker. The public worries that AI might erode human critical thinking. Individuals acknowledge AI\u0026rsquo;s value while also being vigilant against its potential to alienate human agency. Therefore, some advocate for \u0026ldquo;scientific use\u0026rdquo; mechanisms to balance tool dependency and skill retention.\u003c/p\u003e\u003cp\u003eWhat emerges as particularly noteworthy within in the public discourse is the collective narrative of the \u0026ldquo;AI era\u0026rdquo;. Phrases like \u0026ldquo;the era of AI for all\u0026rdquo; and \u0026ldquo;AI for work too\u0026rdquo; reflect a developing social consensus positioning AI as basic infrastructure, akin to water or electricity. AI\u0026rsquo;s importance fuels the notion that individuals unable to master AI tools risk dual marginalization, both in professional competition and knowledge acquisition. In this context, the adoption by universities of blanket bans on AI are arguably going against the tide. On one hand, such prohibition fails to address the root cause: students are forced to depend on AI due to workload overload and ability mismatches. On the other, an AI-free educational environment deprives students of opportunities to develop critical AI literacy, thereby widening the skill gaps between them and societal demands. Thus, it is urgently required that the education system move beyond simplistic control-oriented governance to adaptative, context-sensitive regulatory frameworks aligned with the realities of technological integration.\u003c/p\u003e\u003cp\u003eDistinctively, the closure mechanism for AI use in Chinese higher education will be significantly policy-driven, established through collaboration between the government\u0026rsquo;s education authorities and universities. Within this framework, a tiered governance model based on different scenarios is recommended. Drawing inspiration from the \u0026ldquo;four-category scenario\u0026rdquo; approach in the Shanghai Jiao Tong University (2025) policy \u0026ldquo;SJTU Guidance on Development and Governance of AI in Education and Teaching (Trial Version)\u0026rdquo;, AI use scenarios could be classified into: Prohibited (e.g., graduation theses and innovative research requiring strict prevention of AI substituting independent thought), Restricted, Encouraged (e.g., literature search, data processing), and Open Levels. This tiered AI management can help achieve both ethical constraints and technological empowerment simultaneously.\u003c/p\u003e\u003cp\u003eIt\u0026rsquo;s important to note that current university oversight often focuses on technical measures like AI-generated text detection rates. However, the key to truly addressing AI dependency lies in reconstructing the course design and evaluation systems. To prevent students from being forced into AI use by excessive tasks, educational institutions must promote course reform. First, reduce inefficient repetitive assignments and replace them with tasks requiring critical thinking to lessen reliance on substitution tools. Then, consider mandating AI literacy training as part of general education, focusing on developing two key abilities: critical usage skills (e.g., verifying information authenticity, identifying algorithmic bias) and human-AI collaborative creativity. Furthermore, developing teachers\u0026rsquo; AI competency, positioning them not only as knowledge transmitters because, as AI becomes powerful at knowledge delivery, the irreplaceable role of teachers must increasingly lie in sparking innovative potential and ethical reflection, as well as nurturing morality. In an AI era, higher education should not merely chase tool efficiency. Instead, it should center on capability cultivation, aiming to foster whole persons capable of harnessing tools without becoming alienated by them and maintaining clear thought in the AI age.\u003c/p\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study applied the Social Construction of Technology theoretical framework and a mixed-methods approach to systematically investigate the social construction process of artificial intelligence adoption in higher education on Chinese social media (Weibo). The key finding indicates that different relevant social groups demonstrated significant \u0026ldquo;interpretive flexibility\u0026rdquo; toward AI technology: producers framed AI as an \u0026ldquo;essential learning tool\u0026rdquo; for efficiency enhancement due to marketing strategies; advocates focused on academic ethical risks and capability replacement concerns; users (students) strongly supported its instrumental value primarily due to academic pressure; while bystanders generally recognized its epochal significance but called for cautious use. These diverse interpretations reflected distinct group positions. The study revealed a policy-driven characteristic in the \u0026ldquo;closure\u0026rdquo; mechanism within China\u0026rsquo;s higher education context. Its future trajectory may rely on coordinated development between stratified governance models and teaching system reforms.\u003c/p\u003e\u003cp\u003eOne key contribution of this study lies in empirically examining how relevant social groups negotiate the meaning and application boundaries of AI technology in China\u0026rsquo;s higher education environment, characterized by administrative dominance coexisting with market mechanisms, using Chinese social media data. Additionally, the study explores a technology dispute coordination mechanism driven by administrative power within the Chinese context. This work provides important empirical evidence for applying the SCOT theory in non-Western higher education settings. It demonstrates how different institutional environments substantially influence technology acceptance and implementation paths in society and further confirms the diverse characteristics of technology adoption patterns across cultures.\u003c/p\u003e\u003cp\u003eNonetheless, thestudy has several limitations. The data primarily derives from public discussions from a single social media platform (Weibo). While effective for capturing public opinion, it may not sufficiently cover internal perspectives or private concerns within specific groups, such as university administrators and some education policymakers. Future research should expand data sources by incorporating in-depth interviews or more diverse social media data to better depict complex stakeholder positions and interaction dynamics.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe datasets supporting the findings of this study have been deposited in the FigShare repository and are publicly available under the DOI: 10.6084/m9.figshare.29605916.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eCompeting interests\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eEthical statements\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis article does not contain any studies with human participants performed by any of the authors. All data were collected from Sina Weibo Open API v2 legally.\u003c/p\u003e\n\u003cp\u003e\u0026nbsp;\u003cstrong\u003eFunding Declaration\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was funded by the Science and Technology Department of Guangxi Zhuang Autonomous Region, Talent Project of Guangxi Science and Technology Department (Funding No. 2022AC21201). The funders had no role in study design, data analysis, or manuscript preparation.\u0026nbsp;\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003e\u003cstrong\u003eAdiguzel, T., Kaya, M. H., \u0026amp; Cansu, F. K.\u003c/strong\u003e (2023). Revolutionizing education with AI: Exploring the transformative potential of ChatGPT. \u003cem\u003eContemporary Educational Technology\u003c/em\u003e, \u003cem\u003e15\u003c/em\u003e(3), ep429. https://doi.org/10.30935/cedtech/13152\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eAlAfnan, M. A., Dishari, S., Jovic, M., \u0026amp; Lomidze, K. \u003c/strong\u003e(2023). ChatGPT as an educational tool: Opportunities, challenges, and recommendations for communication, business writing, and composition courses. \u003cem\u003eJournal of Artificial Intelligence and Technology\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e(2), Article 1. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eBlei, D. M., Ng, A. Y., \u0026amp; Jordan, M. I. \u003c/strong\u003e(2003). Latent Dirichlet allocation. \u003cem\u003eJournal of Machine Learning Research\u003c/em\u003e, \u003cem\u003e3\u003c/em\u003e, 993\u0026ndash;1022. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eBijker, W. E. \u003c/strong\u003e(1997). \u003cem\u003eOf bicycles, bakelites, and bulbs: Toward a theory of sociotechnical change\u003c/em\u003e. MIT Press.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eBloor, D.\u003c/strong\u003e (1976). \u003cem\u003eKnowledge and social imagery\u003c/em\u003e. University of Chicago Press.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eBruns, A., \u0026amp; Stieglitz, S. \u003c/strong\u003e(2013). Towards more systematic Twitter analysis: Metrics for tweeting activities. \u003cem\u003eInternational Journal of Social Research Methodology\u003c/em\u003e, \u003cem\u003e16\u003c/em\u003e(2), 91\u0026ndash;108. https://doi.org/10.1080/13645579.2012.756095\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eChang, J., Gerrish, S., Wang, C., Boyd-Graber, J., \u0026amp; Blei, D.\u003c/strong\u003e (2009). Reading tea leaves: How humans interpret topic models. \u003cem\u003eAdvances in Neural Information Processing Systems\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e, 288\u0026ndash;296. https://proceedings.neurips.cc/paper/2009/hash/f92586a25bb3145facd64ab20fd554ff-Abstract.html\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eChoi, J. H., Hickman, K. E., Monahan, A., \u0026amp; Schwarcz, D. B. \u003c/strong\u003e(2023). ChatGPT goes to law school. \u003cem\u003eSSRN Electronic Journal\u003c/em\u003e. https://doi.org/10.2139/ssrn.4335905\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003eClayton, N. \u003c/strong\u003e(2002). SCOT: Does it answer? \u003cem\u003eTechnology and Culture\u003c/em\u003e, \u003cem\u003e43\u003c/em\u003e(2), 351\u0026ndash;360. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Driscoll, K., \u0026amp; Walker, S. \u003c/strong\u003e(2014). Big data, big questions | Working within a black box: Transparency in the collection and production of big Twitter data. \u003cem\u003eInternational Journal of Communication\u003c/em\u003e, \u003cem\u003e8\u003c/em\u003e, Article 20. https://ijoc.org/index.php/ijoc/article/view/2174\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Dwivedi, Y. K., Kshetri, N., Hughes, L., Slade, E. L., Jeyaraj, A., Kar, A. K., ... \u0026amp; Wright, R.\u003c/strong\u003e (2023). \u0026quot;So what if ChatGPT wrote it?\u0026quot; Multidisciplinary perspectives on opportunities, challenges and implications of generative conversational AI for research, practice and policy. \u003cem\u003eInternational Journal of Information Management\u003c/em\u003e, \u003cem\u003e71\u003c/em\u003e, Article 102642. https://doi.org/10.1016/j.ijinfomgt.2023.102642\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e European Union. \u003c/strong\u003e(2024). Regulation (EU) 2024/1689 on artificial intelligence. \u003cem\u003eOfficial Journal of the European Union\u003c/em\u003e, \u003cem\u003eL 200\u003c/em\u003e, 1\u0026ndash;349. Retrieved July 20, 2025, from https://eur-lex.europa.eu/legal-content/EN/TXT/?uri=CELEX:32024R1689\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Feng, Y., Poralla, P., Dash, S., Li, K., Desai, V., \u0026amp; Qiu, M. \u003c/strong\u003e(2023). \u003cem\u003eThe impact of ChatGPT on streaming media: A crowdsourced and data-driven analysis using Twitter and Reddit\u003c/em\u003e. arXiv. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Fiesler, C., \u0026amp; Proferes, N. \u003c/strong\u003e(2018). \u0026quot;Participant\u0026quot; perceptions of Twitter research ethics. \u003cem\u003eSocial Media + Society\u003c/em\u003e, \u003cem\u003e4\u003c/em\u003e(1), 1\u0026ndash;14. https://doi.org/10.1177/2056305118763366\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Fudan University. \u003c/strong\u003e(2024). \u003cem\u003eRegulations on the use of AI tools in undergraduate theses (trial implementation)\u003c/em\u003e. Retrieved July 20, 2025, from https://www.fudan.design/ai.html\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Garvey, E. G. \u003c/strong\u003e(1996). \u003cem\u003eThe adman in the parlor: Magazines and the gendering of consumer culture, 1880s to 1910s\u003c/em\u003e. Oxford University Press.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Griffiths, T. L., \u0026amp; Steyvers, M.\u003c/strong\u003e (2004). Finding scientific topics. \u003cem\u003eProceedings of the National Academy of Sciences\u003c/em\u003e, \u003cem\u003e101\u003c/em\u003e, 5228\u0026ndash;5235. https://doi.org/10.1073/pnas.0307752101\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Guo, F., Liu, Z., Lu, Q., Ji, S., \u0026amp; Zhang, C. \u003c/strong\u003e(2024). Public opinion about COVID-19 on a microblog platform in China: Topic modeling and multidimensional sentiment analysis of social media. \u003cem\u003eJournal of Medical Internet Research\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e, Article e47508. https://doi.org/10.2196/47508\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Haque, M. U., Dharmadasa, I., Sworna, Z. T., Rajapakse, R. N., \u0026amp; Ahmad, H. \u003c/strong\u003e(2022). \u003cem\u003e\u0026quot;I think this is the most disruptive technology\u0026quot;: Exploring sentiments of ChatGPT early adopters using Twitter data\u003c/em\u003e. arXiv. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Huang, Q., Lv, C., Lu, L., \u0026amp; Tu, S. \u003c/strong\u003e(2025). Evaluating the quality of AI-generated digital educational resources for university teaching and learning. \u003cem\u003eSystems\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(3), Article 174. https://doi.org/10.3390/systems13030174\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Huang, S., Sun, X., \u0026amp; Wang, H. \u003c/strong\u003e(2017). Addressing domain adaptation for Chinese word segmentation with global recurrent structure. In \u003cem\u003eProceedings of the Eighth International Joint Conference on Natural Language Processing\u003c/em\u003e (pp. 184\u0026ndash;193). Taipei, Taiwan: AFNLP. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Humphreys, L.\u003c/strong\u003e (2005a). Cellphones in public: Social interactions in a wireless era. \u003cem\u003eNew Media \u0026amp; Society\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(6), 810\u0026ndash;833. https://doi.org/10.1177/1461444805058164\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Humphreys, L. \u003c/strong\u003e(2005b). Reframing social groups, closure, and stabilization in the social construction of technology. \u003cem\u003eSocial Epistemology\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(2-3), 231\u0026ndash;253. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Johnson, A. \u003c/strong\u003e(2023, January 18). ChatGPT in schools: Here\u0026apos;s where it\u0026apos;s banned\u0026mdash;and how it could potentially help students. \u003cem\u003eForbes\u003c/em\u003e. Retrieved July 20, 2025, from https://www.forbes.com/sites/ariannajohnson/2023/01/18/chatgpt-in-schools-heres-where-its-banned-and-how-it-could-potentially-help-students/\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Kasneci, E., Se\u0026szlig;ler, K., K\u0026uuml;chemann, S., Bannert, M., Dementieva, D., Fischer, F., \u0026amp; Kasneci, G. \u003c/strong\u003e(2023). ChatGPT for good? On opportunities and challenges of large language models for education. \u003cem\u003eEducational Technology \u0026amp; Society\u003c/em\u003e, \u003cem\u003e26\u003c/em\u003e(1), 1\u0026ndash;23. https://doi.org/10.30191/ETS.202301_26(1).0003\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Koetsier, J. \u003c/strong\u003e(2023, March 14). GPT-4 scores in the top 10% on the Uniform Bar Exam. \u003cem\u003eForbes\u003c/em\u003e. Retrieved July 20, 2025, from https://www.forbes.com/sites/johnkoetsier/2023/03/14/gpt-4-scores-in-the-top-10-on-the-uniform-bar-exam/\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Lau, J. H., Newman, D., \u0026amp; Baldwin, T. \u003c/strong\u003e(2014). Machine reading tea leaves: Automatically evaluating topic coherence and topic model quality. In \u003cem\u003eProceedings of the 14th Conference of the European Chapter of the Association for Computational Linguistics\u003c/em\u003e (pp. 530\u0026ndash;539). Gothenburg, Sweden: Association for Computational Linguistics. https://doi.org/10.3115/v1/E14-1056\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Leiter, C., Zhang, R., Chen, Y., \u0026amp; Riedl, M. O. \u003c/strong\u003e(2023). AI in the wild: A longitudinal analysis of public perceptions of ChatGPT on Twitter. \u003cem\u003eProceedings of the International AAAI Conference on Web and Social Media\u003c/em\u003e, \u003cem\u003e17\u003c/em\u003e, 505\u0026ndash;517. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Li, L., Ma, Z., Fan, L., Lee, S., Yu, H., \u0026amp; Hemphill, L. \u003c/strong\u003e(2024). ChatGPT in education: A discourse analysis of worries and concerns on social media. \u003cem\u003eEducation and Information Technologies\u003c/em\u003e, \u003cem\u003e29\u003c/em\u003e(9), 10729\u0026ndash;10762. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Liu, Q., Zheng, Z., Zheng, J., Chen, Q., Liu, G., Chen, S., ... \u0026amp; Ming, W. K. \u003c/strong\u003e(2020). Health communication through news media during the early stage of the COVID-19 outbreak in China: Digital topic modeling approach. \u003cem\u003eJournal of Medical Internet Research\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(4), Article e19118. https://doi.org/10.2196/19118\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Liu, Z., Zhang, Y., Chang, E. Y., \u0026amp; Sun, M. \u003c/strong\u003e(2011). PLDA+: Parallel latent Dirichlet allocation with data placement and pipeline processing. \u003cem\u003eACM Transactions on Intelligent Systems and Technology\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(3), Article 26. https://doi.org/10.1145/1961189.1961198\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Loper, E., \u0026amp; Bird, S. \u003c/strong\u003e(2002). \u003cem\u003eNLTK: The natural language toolkit\u003c/em\u003e [Preprint]. arXiv. https://arxiv.org/abs/cs/0205028\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Mackay, H., \u0026amp; Gillespie, G. \u003c/strong\u003e(1992). Extending the social shaping of technology approach: Ideology and appropriation. \u003cem\u003eSocial Studies of Science\u003c/em\u003e, \u003cem\u003e22\u003c/em\u003e(4), 685\u0026ndash;716. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e MacNeil, S., Tran, A., Mogil, D., Bernstein, S., Ross, E., \u0026amp; Huang, Z. \u003c/strong\u003e(2022). Generating diverse code explanations using the GPT-3 language model. In \u003cem\u003eProceedings of the 55th ACM Technical Symposium on Computer Science Education\u003c/em\u003e (pp. 1\u0026ndash;7). Portland, OR: ACM. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e McLuhan, M. \u003c/strong\u003e(1964). \u003cem\u003eUnderstanding media: The extensions of man\u003c/em\u003e. MIT Press.\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Mhlanga, D. \u003c/strong\u003e(2023). \u003cem\u003eOpen AI in education, the responsible and ethical use of ChatGPT towards lifelong learning\u003c/em\u003e [Preprint]. Social Science Research Network. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Moore, R. L., Miller, C. N., \u0026amp; Leskovec, J. \u003c/strong\u003e(2022). Leveraging AI for scalable data science education. \u003cem\u003eJournal of Machine Learning Research\u003c/em\u003e, \u003cem\u003e23\u003c/em\u003e(1), 1\u0026ndash;30. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Peng, K. H., Liou, L. H., Chang, C. S., \u0026amp; Lee, D. S. \u003c/strong\u003e(2015). Predicting personality traits of Chinese users based on Facebook wall posts. In \u003cem\u003eProceedings of the 24th Wireless and Optical Communication Conference\u003c/em\u003e (pp. 9\u0026ndash;14). Piscataway, NJ: IEEE. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Pinch, T. J., \u0026amp; Bijker, W. E. \u003c/strong\u003e(1984). The social construction of facts and artefacts: Or how the sociology of science and the sociology of technology might benefit each other. \u003cem\u003eSocial Studies of Science\u003c/em\u003e, \u003cem\u003e14\u003c/em\u003e(3), 399\u0026ndash;441. https://doi.org/10.1177/030631284014003004\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Qadir, J. \u003c/strong\u003e(2022). Engineering education in the era of ChatGPT: Promise and pitfalls. \u003cem\u003eIEEE Access\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e, 125813\u0026ndash;125826. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Řehůřek, R., \u0026amp; Sojka, P. \u003c/strong\u003e(2010). Software framework for topic modelling with large corpora. In \u003cem\u003eProceedings of the LREC 2010 Workshop on New Challenges for NLP Frameworks\u003c/em\u003e (pp. 45\u0026ndash;50). Valletta, Malta: ELRA. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e R\u0026ouml;der, M., Both, A., \u0026amp; Hinneburg, A. \u003c/strong\u003e(2015). Exploring the space of topic coherence measures. In \u003cem\u003eProceedings of the Eighth ACM International Conference on Web Search and Data Mining\u003c/em\u003e (pp. 399\u0026ndash;408). Shanghai, China: ACM. https://doi.org/10.1145/2684822.2685324\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Rudolph, J., Tan, S., \u0026amp; Tan, S. \u003c/strong\u003e(2023). ChatGPT: Bullshit spewer or the end of authentic assessments? \u003cem\u003eJournal of Applied Learning and Teaching\u003c/em\u003e, \u003cem\u003e6\u003c/em\u003e(1), Article 1. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Shanghai Jiao Tong University. \u003c/strong\u003e(2025). \u003cem\u003eGuidance on development and governance of AI in education and teaching (trial version)\u003c/em\u003e. Retrieved July 20, 2025, from https://www.sjtu.edu.cn/tg/20250304/207682.html \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Sok, S., \u0026amp; Heng, K. \u003c/strong\u003e(2023). ChatGPT for educational chatbots: The balance between hype and reality. \u003cem\u003eJournal of Educational Computing Research\u003c/em\u003e, \u003cem\u003e61\u003c/em\u003e(2), 345\u0026ndash;367. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Stanford Teaching Commons. \u003c/strong\u003e(2025). \u003cem\u003eUnderstanding AI literacy\u003c/em\u003e. Retrieved July 20, 2025, from https://teachingcommons.stanford.edu/teaching-guides/artificial-intelligence-teaching-guide/understanding-ai-literacy\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Tack, C., \u0026amp; Piech, C. \u003c/strong\u003e(2022). The pedagogical limitations of large language models in disciplinary dialogues. In \u003cem\u003eProceedings of the Ninth ACM Conference on Learning @ Scale\u003c/em\u003e (pp. 1\u0026ndash;4). Copenhagen, Denmark: ACM. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Taecharungroj, V.\u003c/strong\u003e (2023). What can ChatGPT do? Analyzing early reactions to the innovative AI chatbot on Twitter. \u003cem\u003eBig Data and Cognitive Computing\u003c/em\u003e, \u003cem\u003e7\u003c/em\u003e(1), Article 35. https://doi.org/10.3390/bdcc7010035\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Thurzo, A., Urbanov\u0026aacute;, W., \u0026amp; Waczul\u0026iacute;kov\u0026aacute;, I.\u003c/strong\u003e (2023). The impact of AI on dental education: A scoping review. \u003cem\u003eEuropean Journal of Dental Education\u003c/em\u003e, \u003cem\u003e27\u003c/em\u003e(3), 1\u0026ndash;12. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Tlili, A., Shehata, B., Adarkwah, M. A., Bozkurt, A., \u0026amp; Hickey, D. T.\u003c/strong\u003e (2023). What if the devil is my guardian angel: ChatGPT as a case study of using chatbots in education. \u003cem\u003eSmart Learning Environments\u003c/em\u003e, \u003cem\u003e10\u003c/em\u003e(1), Article 15. https://doi.org/10.1186/s40561-023-00237-x\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Tufekci, Z.\u003c/strong\u003e (2014). Big questions for social media big data: Representativeness, validity and other methodological pitfalls. In \u003cem\u003eProceedings of the International AAAI Conference on Web and Social Media\u003c/em\u003e (pp. 505\u0026ndash;514). Ann Arbor, MI: AAAI. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Winner, L.\u003c/strong\u003e (1986). Do artifacts have politics? In L. Winner, \u003cem\u003eThe whale and the reactor\u003c/em\u003e (pp. 19\u0026ndash;39). University of Chicago Press. (Original work published 1980)\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Xie, T., Qin, P., \u0026amp; Zhu, L.\u003c/strong\u003e (2019). Study on the topic mining and dynamic visualization in view of LDA model. \u003cem\u003eModern Applied Science\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(1), 204. https://doi.org/10.5539/mas.v13n1p204\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Zhang, Y., Jin, R., \u0026amp; Zhou, Z. H.\u003c/strong\u003e (2010). Understanding bag-of-words model: A statistical framework. \u003cem\u003eInternational Journal of Machine Learning and Cybernetics\u003c/em\u003e, \u003cem\u003e1\u003c/em\u003e, 43\u0026ndash;52. https://doi.org/10.1007/s13042-010-0001-0\u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Zhang, Y., Sun, M., Ren, Y., \u0026amp; Shen, J.\u003c/strong\u003e (2020). Sentiment analysis of Sina Weibo users under the impact of super typhoon Lekima using natural language processing tools: A multi-tags case study. \u003cem\u003eProcedia Computer Science\u003c/em\u003e, \u003cem\u003e174\u003c/em\u003e, 478\u0026ndash;490. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Zhao, R., Yunus, M. M., \u0026amp; Rafiq, K. R. M.\u003c/strong\u003e (2023). The impact of the use of ChatGPT in enhancing students\u0026apos; engagement and learning outcomes in higher education: A review. \u003cem\u003eInternational Journal of Academic Research in Business and Social Sciences\u003c/em\u003e, \u003cem\u003e13\u003c/em\u003e(12), Article e1\u0026ndash;e15. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Zhou, J. S., Dai, X., Ni, R. Y., \u0026amp; Chen, J.\u003c/strong\u003e (2005). A hybrid approach to Chinese word segmentation around CRFs. In \u003cem\u003eProceedings of the Fourth SIGHAN Workshop on Chinese Language Processing\u003c/em\u003e (pp. 63\u0026ndash;68). Jeju Island, South Korea: Asian Federation of Natural Language Processing. \u003c/li\u003e\n\u003cli\u003e\u003cstrong\u003e Zhu, H., Yu, B., Halfaker, A., \u0026amp; Terveen, L.\u003c/strong\u003e (2018). Value-sensitive algorithm design: Method, case study, and lessons. \u003cem\u003eProceedings of the ACM on Human-Computer Interaction\u003c/em\u003e, \u003cem\u003e2\u003c/em\u003e(CSCW), Article 119. \u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":false,"hideJournal":true,"highlight":"","institution":"","isAcceptedByJournal":false,"isAuthorSuppliedPdf":false,"isDeskRejected":"","isHiddenFromSearch":false,"isInQc":false,"isInWorkflow":false,"isPdf":false,"isPdfUpToDate":true,"isWithdrawnOrRetracted":false,"journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true},"keywords":"Generative AI, Chinese higher education, LDA topic modeling, Social Construction of Technology","lastPublishedDoi":"10.21203/rs.3.rs-7171423/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7171423/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"Global higher education is experiencing dual pressures from rapid rise of generative AI tools such as ChatGPT, which drive pedagogical innovation while triggering global concerns over academic integrity and skill degradation. This study addresses a key research gap by investigating how China’s distinct context, marked by government influence, market dynamics, and achievement pressure, influences conflicting understandings of AI in academia. Employing a mixed-methods approach combining Latent Dirichlet Allocation (LDA) topic modeling and Social Construction of Technology (SCOT) framework-guided content analysis, discourse dynamics among key social groups was mapped: producers, advocates, users, and bystanders. Thematic modeling identified five key areas of controversy: AI-enabled academic practices, ethical concerns, efficiency evaluations, policy debates, and the perceived inevitability of AI adoption. Content analysis revealed clear differences in perspectives among key groups: producers highlighted efficiency benefits in their communications; advocates stressed academic integrity risks and endorsed regulatory measures; users explained their reliance on AI as necessary due to academic pressures and curricular dissatisfaction; while bystanders expressed diverse but often practical views on AI’s broader societal impact. Crucially, our analysis indicates that China’s closure mechanism will likely be predominantly policy-driven, advocating tiered governance and pedagogical reforms to mitigate structural pressures fueling student dependency. This research contributes a non-Western perspective to SCOT theory, revealing how administrative-market dynamics shape technology negotiation. Furthermore, it proposes practical implementation pathways for adaptable AI governance frameworks applicable across diverse global higher education contexts.​","manuscriptTitle":"Efficiency Tool or Academic Threat? Mapping China’s AI Controversy in Higher Education with Mixed Methods","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2025-10-13 16:12:08","doi":"10.21203/rs.3.rs-7171423/v1","editorialEvents":[{"type":"communityComments","content":0}],"status":"published","journal":{"display":true,"email":"[email protected]","identity":"researchsquare","isNatureJournal":false,"hasQc":true,"allowDirectSubmit":true,"externalIdentity":"","sideBox":"","snPcode":"","submissionUrl":"/submission","title":"Research Square","twitterHandle":"researchsquare","acdcEnabled":true,"dfaEnabled":false,"editorialSystem":"","reportingPortfolio":"","inReviewEnabled":false,"inReviewRevisionsEnabled":true}}],"origin":"","ownerIdentity":"06dad630-e170-460a-b0c2-bb1bd172e7b3","owner":[],"postedDate":"October 13th, 2025","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":56101604,"name":"Social science/Education"},{"id":56101605,"name":"Business and commerce/Information systems and information technology"},{"id":56101606,"name":"Social science/Politics and international relations"},{"id":56101607,"name":"Social science/Science technology and society"}],"tags":[],"updatedAt":"2026-03-25T04:54:47+00:00","versionOfRecord":[],"versionCreatedAt":"2025-10-13 16:12:08","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7171423","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7171423","identity":"rs-7171423","version":["v1"]},"buildId":"8U1c8b4HqxoKbykW_rLl7","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}