Epistemic Trust in Generative AI and the Reshaping of Professional Judgment:

Epistemic Trust in Generative AI and the Reshaping of Professional Judgment: | 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 Research Article Epistemic Trust in Generative AI and the Reshaping of Professional Judgment: Gai Meijuan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-9071480/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 As generative artificial intelligence (GenAI) becomes deeply embedded in university language teaching contexts, EFL teachers are increasingly confronted with a novel epistemic challenge: whether and how to trust AI-generated knowledge in their professional practice. This study examines the psychological and professional consequences of epistemic trust in GenAI (ETGAI) among 420 university EFL teachers in China, proposing a serial mediation model in which professional judgment reshaping (PJR) and moral stress (MS) sequentially transmit the effects of ETGAI on teacher professional development (TPD). Drawing on epistemic trust theory, Conservation of Resources (COR) theory, and professional agency frameworks, a covariance-based structural equation model (CBSEM) was estimated using Mplus 8.3 with 5,000 bootstrap replications. Results supported all seven hypotheses: ETGAI positively predicted PJR (β = .476) and TPD (β = .346), PJR positively predicted MS (β = .490), and MS negatively predicted TPD (β = −.405). The serial indirect effect (ETGAI → PJR → MS → TPD) was significant (β = −.095, 95% BC CI [− .133, − .061]), revealing a moral stress pathway through which epistemic trust in GenAI ultimately undermines professional development when mediated by judgment reshaping. Theoretical and practical implications for AI-integrated EFL teacher education are discussed. epistemic trust generative AI professional judgment reshaping moral stress teacher professional development EFL China structural equation modeling Figures Figure 1 1. Introduction The rapid proliferation of generative artificial intelligence tools—including ChatGPT, Gemini, and DeepSeek—has fundamentally altered the landscape of English as a Foreign Language (EFL) teaching at the university level. Within a remarkably short span, these tools have moved from peripheral novelties to central instruments in lesson planning, feedback provision, assessment design, and student interaction (Lee et al., 2025 ). Yet the integration of GenAI into professional teaching practice is far from a straightforward technical adoption; it carries profound epistemic, ethical, and psychological dimensions that the existing literature has only begun to explore. Central among these dimensions is the question of epistemic trust. Epistemic trust—originally theorized as the disposition to accept communicated knowledge as relevant and reliable (Fonagy & colleagues; Yirmiya & Fonagy, 2025 )—has emerged as a pivotal construct in understanding how professionals engage with AI-generated information. When teachers extend epistemic trust to GenAI systems, they are, in effect, delegating a portion of their professional judgment to an external agent whose reasoning processes remain partially opaque. Pandey et al. ( 2025 , 2026 ) have demonstrated that epistemic trust in GenAI is a distinct and measurable construct in higher education contexts, significantly predicting engagement patterns and information acceptance behaviors. Yet the downstream consequences of this trust—particularly for teachers whose professional identities are bound up in the exercise of expert judgment—remain poorly understood. One important consequence that appears underexplored is the reshaping of professional judgment. As teachers come to rely on GenAI for content generation, assessment feedback, and pedagogical decisions, the locus of professional judgment may shift in ways that threaten professional agency and autonomy (Chen et al., 2026 ; Küçükuncular & Ertugan, 2025 ). This process of professional judgment reshaping (PJR) is not inherently negative; it may enable teachers to perform more efficiently and creatively. However, it also creates the conditions for moral stress: a form of psychological tension arising when the external reconfiguration of one's professional role conflicts with deeply held values about pedagogical authenticity, academic integrity, and professional responsibility (Vassallo, 2026 ; Yang, 2026 ). Despite the conceptual plausibility of this pathway, no empirical study has yet traced the full serial mediation chain from epistemic trust in GenAI through professional judgment reshaping and moral stress to teacher professional development. The existing literature on GenAI in EFL education has largely focused on attitudes and adoption (Fatalaki et al., 2025 ; Mutlu, 2025 ), perceptions of student writing (Chen & Pi, 2026 ), or identity negotiation (Hu et al., 2025 ; Li et al., 2025 ), without attending to the psychological stress mechanisms that mediate these relationships. Meanwhile, research on moral stress has developed primarily in healthcare settings (Ansari et al., 2025 ; Awad et al., 2026 ; Barr, 2025 ), with limited translation to educational contexts. The present study addresses this gap by investigating the following research questions: (1) Does epistemic trust in GenAI positively predict professional judgment reshaping among university EFL teachers in China? (2) Does professional judgment reshaping in turn predict moral stress? (3) Does moral stress mediate the relationship between professional judgment reshaping and teacher professional development? (4) Do professional judgment reshaping and moral stress jointly and serially mediate the relationship between epistemic trust in GenAI and teacher professional development? Using data from 420 university EFL teachers in China, and employing covariance-based SEM with bootstrapped confidence intervals, this study tests a theoretically grounded serial mediation model (ETGAI → PJR → MS → TPD). The study makes three contributions. Theoretically, it extends epistemic trust theory and Conservation of Resources (COR) theory to the AI-integrated teaching context, proposing and empirically validating a novel serial mediation model. Methodologically, it provides a rigorously validated instrument battery adapted for Chinese university EFL contexts. Practically, it offers evidence-based implications for institutional policy, teacher education programs, and individual reflective practice in the rapidly evolving GenAI era. The remainder of the paper is organized as follows: Section 2 reviews the literature and develops the theoretical framework; Section 3 describes the methodology; Section 4 presents the results; Section 5 discusses the findings; and Section 6 offers conclusions, limitations, and future directions. 2. Literature Review and Theoretical Framework 2.1 Epistemic Trust: From Developmental Psychology to Artificial Intelligence Epistemic trust was originally conceptualized in developmental psychology as the fundamental human disposition to regard communicated information as genuine, relevant, and applicable to the self (Yirmiya & Fonagy, 2025 ). In its original formulation, epistemic trust serves as an adaptive mechanism that enables efficient cultural learning: rather than independently verifying every piece of received knowledge, individuals selectively extend trust to sources they perceive as credible, benevolent, and contextually appropriate. Critically, this trust is not blind credulity; it coexists with epistemic vigilance—a parallel evaluative disposition that screens for unreliability or deception (Sedlakova et al., 2025 ). The extension of epistemic trust theory to technological agents is a relatively recent development, accelerated by the widespread deployment of large language models. McCarthy ( 2026 ) traces the discursive construction of AI as an epistemic authority in public imaginaries, arguing that institutional trust in AI involves a form of delegated epistemic authority analogous to trust in expert human informants. Sedlakova et al. ( 2025 ) develop a normative analysis of "human-like epistemic trust" in conversational AI, concluding that while AI systems can elicit trust responses phenomenologically similar to those directed at humans, important differences in reliability, accountability, and relational reciprocity warrant caution. Békés and Doorn ( 2026 ) further document that individuals with higher attachment anxiety and mental health vulnerabilities exhibit significantly elevated epistemic trust in AI agents, suggesting that epistemic trust in AI is a heterogeneous construct shaped by individual difference variables. In the specific domain of higher education, Pandey et al. ( 2026 ) developed and validated the Epistemic Trust in Generative AI for Higher Education Scale (ETGAI-HE), demonstrating that epistemic trust in GenAI is empirically distinguishable from general AI acceptance and predicts qualitatively distinct behavioral patterns. A companion bibliometric study (Pandey et al., 2025 ) mapped the evolution of this construct across disciplines, identifying education and health as the two sectors in which epistemic trust in AI has attracted the greatest research attention. Sacco et al. ( 2025 ) found that epistemic trust—operationalized as openness to accepting AI output—moderated the relationship between general AI attitudes and behavioral intentions, underscoring its role as a psychological mechanism rather than merely an attitudinal disposition. Boyd and Markowitz ( 2026 ) further argue that AI systems that simulate epistemic responsiveness—appearing to attend to, understand, and personalize information—are particularly effective at eliciting trust, with potentially significant implications for professional reliance on AI outputs. For EFL teachers, epistemic trust in GenAI carries a specific professional valence. Language teachers exercise professional judgment over matters of linguistic accuracy, pedagogical appropriateness, and cultural sensitivity—domains in which GenAI systems demonstrate uneven competence. The decision to extend epistemic trust to GenAI in these domains, therefore, represents a consequential professional act with downstream effects on instructional quality, professional identity, and ethical practice. 2.2 Generative AI in University EFL Teaching Contexts The uptake of GenAI tools among EFL teachers has been rapid and uneven. Lee et al. ( 2025 ) provide a comprehensive overview from a global Englishes perspective, documenting that while GenAI offers significant affordances for language teaching—including on-demand text generation, adaptive feedback, and multilingual capability—it also raises fundamental questions about native-speakerism, linguistic authority, and the role of the teacher as language model. Chen et al. ( 2026 ) conducted an ethnographic investigation of how EFL teachers navigate pedagogical decision-making in GenAI-integrated classrooms, coining the notion of "teacherness" to describe the constellation of professional attitudes and practices that teachers mobilize to maintain pedagogical agency in the face of AI encroachment. In Chinese university contexts specifically, the pressures of digital transformation and national AI development agendas have accelerated GenAI adoption in ways that often outpace teacher preparation and institutional support (Zhao et al., 2025 ; Mutlu, 2025 ). Hoang ( 2025 ) documents how cultural-linguistic factors moderate teacher emotional responses to AI implementation in Vietnamese EFL classrooms—findings with potential relevance to Chinese contexts given shared Confucian educational values. Hu et al. ( 2025 ) explore how non-native English researchers negotiate professional identity when using ChatGPT for academic writing, documenting tensions between efficiency gains and concerns about authentic scholarly voice—a dynamic that parallels the identity challenges facing EFL teachers who rely on GenAI for professional tasks. Pham and Le ( 2026 ) offer a narrative inquiry of Vietnamese EFL teachers' experiences of co-reflection with AI, revealing that while AI co-reflection affords new forms of reflective depth, it simultaneously raises ethical concerns about the nature of professional judgment when externally mediated. Similarly, Fatalaki et al. ( 2025 ) developed and validated an instrument measuring EFL teachers' attitudes toward AI in education, finding that perceived professional utility was the strongest predictor of positive attitudes, though concerns about accuracy and academic integrity remained significant barriers. Zhang and Wang ( 2025 ) employed epistemic network analysis to examine how GenAI influences in-service teachers' knowledge-building processes, finding evidence of both cognitive enrichment and epistemic dependency—a pattern consistent with the dual-edged nature of epistemic trust proposed in this study. 2.3 Professional Judgment Reshaping in AI-Integrated Practice Professional judgment in teaching refers to the expert decision-making capacity that enables teachers to read complex classroom situations and respond appropriately, drawing on a synthesis of content knowledge, pedagogical knowledge, relational knowledge, and professional values (Erbay-çetinkaya, 2025 ). This capacity is not static; it develops through experience, reflection, and engagement with professional communities. Critically, professional judgment is also the site at which teachers exercise professional agency—the capacity to act intentionally on the basis of professional goals and values, rather than merely responding to external directives or automated cues (Li et al., 2025 ). The introduction of GenAI into professional practice creates conditions for what we term professional judgment reshaping: the process by which teachers' habitual patterns of professional decision-making are reconfigured through sustained interaction with AI systems. This reshaping is not synonymous with de-skilling or replacement; it may involve enhanced efficiency, expanded creativity, or novel forms of human-AI collaboration (Alreiahi & Alrwaished, 2025 ; Xia et al., 2025 ). However, when epistemic trust leads teachers to defer systematically to AI recommendations in domains that were previously the province of professional expertise, the result may be a gradual attenuation of autonomous professional judgment—a phenomenon related to automation bias documented in other high-stakes professional domains (Roe et al., 2025 ). Sánchez-Trujillo et al. ( 2025 ) found that pre-service teachers with higher general AI acceptance showed greater willingness to adopt AI-generated lesson plans with minimal modification, a pattern suggestive of incipient judgment reshaping even at the beginning stages of professional formation. Küçükuncular and Ertugan ( 2025 ) offer a more critical perspective, arguing from a Marxian standpoint that AI-mediated professional judgment involves a form of intellectual labor extraction that may ultimately undermine teacher professional identity. Together, these studies suggest that professional judgment reshaping is a real and consequential process, though its psychological and developmental effects have yet to be systematically examined. Based on the conceptual logic that higher epistemic trust in GenAI promotes greater reliance on AI in professional decision-making, we propose: H1: Epistemic trust in generative AI positively predicts professional judgment reshaping among university EFL teachers. 2.4 Moral Stress in AI-Integrated Teaching The concept of moral stress draws on a substantial literature originating in healthcare settings, where moral distress was originally defined as the psychological tension arising when professionals know the ethically appropriate course of action but are constrained by external forces from acting accordingly (cf. Ansari et al., 2025 ; Frush & Gaffney, 2025 ). Subsequent elaborations have emphasized the role of role conflict, value misalignment, and institutional constraints in generating sustained moral stress that depletes psychological resources and contributes to professional burnout (Barr, 2025 ; Trueblood et al., 2025 ). In educational contexts, moral stress from AI adoption takes distinctive forms. Vassallo ( 2026 ) conceptualizes what he terms the "AI guilt complex"—a cluster of moral emotions including guilt, complicity, and ethical uncertainty experienced by academics who adopt GenAI tools while remaining uncertain about their professional and ethical implications. Yang ( 2026 ) documents a parallel phenomenon among Chinese patients who experience moral anxiety alongside epistemic empowerment when engaging with AI-assisted medical diagnosis—a dynamic remarkably consonant with what EFL teachers may experience when GenAI reshapes their professional judgment. Drawing on cross-disciplinary evidence, Awad et al. ( 2026 ) demonstrated in a nursing context that AI integration mediates the relationship between moral distress and professional integrity—a finding with important structural parallels to the mechanism proposed here. When professional judgment reshaping creates discrepancies between teachers' performed professional role (relying on AI outputs) and their internalized professional values (exercising independent expert judgment), the resultant cognitive dissonance is expected to manifest as moral stress. This stress is further amplified by the opacity of GenAI reasoning processes, which makes it difficult for teachers to evaluate the epistemological provenance of AI-generated content (Sedlakova et al., 2025 ). We therefore propose: H2: Professional judgment reshaping positively predicts moral stress among university EFL teachers. 2.5 Moral Stress and Teacher Professional Development Teacher professional development (TPD) encompasses the full range of processes through which teachers expand their professional knowledge, refine their skills, develop their identities, and enhance their adaptive capacity in response to changing educational demands (Zainuddin et al., 2026 ; Sharma & Rai, 2026 ). In the AI era, TPD increasingly involves the development of critical AI literacy, ethical reasoning about technology use, and the ability to coordinate human and machine competencies in service of learning (Watts, 2025 ; Sawyer et al., 2025 ). Conservation of Resources (COR) theory (Hobfoll, 1989 ) provides a theoretical framework for predicting how moral stress affects TPD. COR theory proposes that individuals strive to acquire, maintain, and protect valued resources—including psychological resources such as professional identity, self-efficacy, and motivational energy. Stress arises when resources are threatened, lost, or fail to be gained following significant investment. Moral stress, which depletes the psychological resource of ethical coherence and professional self-regard, is expected to impair the cognitive and motivational resources necessary for professional learning and development. Consistent with this reasoning, Lipschuetz and Topaz ( 2026 ) found in a nursing digital transformation context that moral distress significantly constrained nurses' capacity for professional adaptation and learning—a pattern analogous to what we predict for EFL teachers facing GenAI-induced moral stress. Furthermore, digital resilience research in nursing education (Su et al., 2026 ) has documented that psychological distress from technology-mediated role conflict undermines professional learning engagement—a cross-disciplinary parallel supporting our prediction. Woo et al. ( 2026 ) similarly noted in a scoping review of LLM use in clinical documentation that clinician distrust and moral concerns about AI use were associated with reduced professional adoption and development outcomes. We therefore propose: H3: Moral stress negatively predicts teacher professional development. 2.6 Direct Effect of Epistemic Trust on Teacher Professional Development While the mediated pathway through PJR and MS is the central theoretical contribution of this study, epistemic trust in GenAI may also exert a direct positive effect on TPD. When teachers extend epistemic trust to GenAI, they are more likely to engage productively with AI tools, experiment with novel pedagogical approaches, and leverage AI-generated resources in ways that expand their professional repertoire (Zhang & Wang, 2025 ; Zhao et al., 2025 ). This direct effect represents the beneficial dimension of epistemic trust—the productive engagement with GenAI as a professional resource—and is expected to operate independently of the moral stress pathway. Consistent with this reasoning: H4: Epistemic trust in generative AI positively predicts teacher professional development. 2.7 Serial Mediation Model and Hypotheses Summary Integrating the theoretical arguments developed above, we propose a serial mediation model in which the relationship between ETGAI and TPD is transmitted through two sequentially ordered psychological mechanisms: professional judgment reshaping (M1) and moral stress (M2). This model draws on three theoretical pillars: (1) epistemic trust theory, which accounts for the conditions under which professionals accept AI-generated knowledge as authoritative; (2) Conservation of Resources theory (Hobfoll, 1989 ), which predicts how moral stress depletes the psychological resources necessary for professional development; and (3) professional agency frameworks (Erbay-çetinkaya, 2025 ; Li et al., 2025 ), which explain why the reconfiguration of professional judgment by AI constitutes a psychologically significant event for professional identity and wellbeing. The mediation hypotheses are as follows: H5: Professional judgment reshaping mediates the relationship between epistemic trust in generative AI and moral stress. H6: Moral stress mediates the relationship between professional judgment reshaping and teacher professional development. H7: Professional judgment reshaping and moral stress serially mediate the relationship between epistemic trust in generative AI and teacher professional development (ETGAI → PJR → MS → TPD). 3. Methodology 3.1 Research Design This study employed a quantitative, cross-sectional survey design using covariance-based structural equation modeling (CBSEM) to test the proposed serial mediation model. A two-stage analytic approach was adopted: a confirmatory factor analysis (CFA) to evaluate the measurement model, followed by structural model estimation with bootstrapped confidence intervals for mediation testing (Kline, 2016 ). Cross-sectional SEM has been widely used in EFL teacher cognition research and AI adoption studies (Fatalaki et al., 2025 ; Zhang & Wang, 2025 ), and is appropriate for simultaneously testing complex systems of relationships among latent variables. 3.2 Participants and Sampling The target population comprised in-service EFL teachers employed at universities and higher education institutions in mainland China who reported current awareness of, or experience with, GenAI tools in their professional practice. Participants were recruited via a combination of convenience and snowball sampling through professional WeChat groups, EFL teacher networks affiliated with provincial English teaching associations, and personal referrals. The sampling strategy, while not probabilistic, reflects common practice in educational SEM research and is consistent with methodological precedents in the field (Lee et al., 2025 ; Mutlu, 2025 ). Following SEM sample size recommendations (Kline, 2016 ), a target of at least 300 responses was established. A total of 456 questionnaires were collected, of which 420 were retained after excluding responses with more than 10% missing data, systematic patterns indicating random responding, or failure on attention check items. The final analytic sample comprised 420 participants (n male = 126, 30.0%; n female = 286, 68.1%; n other/undisclosed = 8, 1.9%). Teaching experience ranged from 1 to over 16 years. Full demographic details are presented in Table 1 (Section 4 ). The sample size satisfies the 10:1 participant-to-parameter ratio recommended for stable CFA solutions (Hair et al., 2019 ). 3.3 Measures Epistemic Trust in Generative AI (ETGAI). ETGAI was measured using 12 items adapted from the Epistemic Trust in Generative AI for Higher Education Scale (ETGAI-HE; Pandey et al., 2026 ), supplemented by items drawing on the bibliometric-informed framework of Pandey et al. ( 2025 ) and the conceptual analysis of Sedlakova et al. ( 2025 ). Items assess the degree to which respondents regard GenAI systems as accurate, credible, and reliable epistemic sources in professional teaching contexts (e.g., "I trust the information generated by GenAI tools to be professionally accurate and reliable"). Response options ranged from 1 (strongly disagree) to 7 (strongly agree). In the present sample, Cronbach's α = .895 and composite reliability (CR) = .921. Professional Judgment Reshaping (PJR). PJR was measured using 10 items developed de novo for this study, drawing on conceptual frameworks from Chen et al. ( 2026 ), Erbay-çetinkaya ( 2025 ), and Xia et al. ( 2025 ). Items assess the degree to which respondents' habitual patterns of professional decision-making—including curriculum design, assessment, and pedagogical choice—have been reconfigured through reliance on GenAI outputs (e.g., "I find myself deferring to GenAI recommendations in areas where I previously exercised independent professional judgment"). Response options ranged from 1 to 7. Cronbach's α = .870, CR = .901. Moral Stress (MS). MS was assessed using 8 items adapted from cross-disciplinary moral distress instruments (cf. Vassallo, 2026 ; Awad et al., 2026 ; Ansari et al., 2025 ) and modified to reflect EFL teacher contexts. Items capture the experience of ethical tension, professional role conflict, and value misalignment arising from AI-mediated teaching practice (e.g., "Using GenAI in my teaching creates an uncomfortable tension between efficiency and my professional responsibility to exercise independent judgment"). Response options ranged from 1 to 7. Cronbach's α = .868, CR = .902. Teacher Professional Development (TPD). TPD was measured using 10 items adapted from validated teacher professional development scales (cf. Zainuddin et al., 2026 ; Sharma & Rai, 2026 ; Watts, 2025 ), modified to reflect AI-era professional learning. Items assess engagement in professional learning activities, development of AI-related competencies, and reflective professional growth (e.g., "My engagement with GenAI has contributed meaningfully to my professional knowledge and skills as an EFL teacher"). Response options ranged from 1 to 7. Cronbach's α = .831, CR = .863. 3.4 Translation and Pilot Testing All items were originally drafted in English and translated into Mandarin Chinese following a forward-backward translation procedure with two bilingual translators. Discrepancies were resolved by consensus. A pilot study involving 35 university EFL teachers (excluded from the main analysis) was conducted to assess item clarity, response distribution, and internal consistency. Minor wording revisions were made to three items based on cognitive interview feedback prior to full-scale data collection. 3.5 Common Method Bias To mitigate common method bias arising from single-source self-report data, procedural remedies recommended by Podsakoff et al. ( 2003 ) were employed: the survey was administered anonymously, predictor and criterion items were separated by unrelated filler items, and scale anchors were varied across constructs. Post hoc, Harman's single-factor test was conducted; the single extracted factor accounted for 28.3% of total variance—well below the 50% threshold—suggesting that common method bias was not a substantial concern in the present data. 3.6 Data Analysis Data analysis proceeded in two stages using Mplus 8.3 (Muthén & Muthén, 1998–2017). In Stage 1, a CFA was conducted to evaluate the measurement model. Fit was assessed using multiple indices: the comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR), with accepted thresholds of CFI > .95, TLI > .95, RMSEA < .06, and SRMR .50), average variance extracted (AVE > .50), and composite reliability (CR > .70; Fornell & Larcker, 1981 ). Discriminant validity was assessed using the Fornell-Larcker criterion. In Stage 2, the structural model was estimated using maximum likelihood estimation with 5,000 bootstrap replications to obtain bias-corrected (BC) 95% confidence intervals for direct and indirect effects. 4. Results 4.1 Sample Characteristics A total of 420 university EFL teachers in China completed the survey. As shown in Table 1 , the sample comprised 126 males (30.0%), 286 females (68.1%), and 8 participants who identified as other or preferred not to disclose (1.9%). The majority of respondents fell in the 26–35 age bracket (38.1%), followed by 36–45 (35.0%). In terms of teaching experience, 30.0% reported 4–8 years, 29.0% reported 9–15 years, 22.9% had 16 or more years, and 18.1% had 1–3 years. Regarding institution type, the largest group taught at regular higher education institutions (45.0%), followed by 985/211 key universities (21.9%), vocational colleges (20.0%), and other institutions (13.1%). In terms of generative AI usage frequency, 31.0% reported using GenAI tools almost daily, 38.1% one to two times per week, 21.9% one to two times per month, and 9.0% rarely or never. Table 1 Sample Characteristics (N = 420) Variable Category n % Gender Male 126 30.0% Female 286 68.1% Other/No disclose 8 1.9% Age < 26 34 8.1% 26–35 160 38.1% 36–45 147 35.0% ≥ 46 79 18.8% Teaching experience 1–3 years 76 18.1% 4–8 years 126 30.0% 9–15 years 122 29.0% ≥ 16 years 96 22.9% Institution type 985/211 92 21.9% Regular HEI 189 45.0% Vocational 84 20.0% Other 55 13.1% GenAI usage frequency Rarely 38 9.0% 1–2×/month 92 21.9% 1–2×/week 160 38.1% Daily 130 31.0% Note. HEI = higher education institution. 4.2 Measurement Model Prior to structural model testing, a confirmatory factor analysis (CFA) was conducted. The model demonstrated excellent fit: χ²(590) = 601.30, p = .365, CFI = .998, TLI = .998, RMSEA = .007 [90% CI: .000, .017], SRMR = .035 (see Table 3 ). Table 2 presents descriptive statistics, reliability estimates, and inter-construct correlations. All standardized factor loadings were statistically significant (p < .001) and ranged from .594 to .753, exceeding the recommended minimum of .50 (Hair et al., 2019 ). Convergent validity was supported: average variance extracted (AVE) values ranged from .494 to .512, and composite reliability (CR) values ranged from .863 to .921, both meeting established thresholds (Fornell & Larcker, 1981 ). Cronbach's α ranged from .831 to .895. Discriminant validity was assessed using the Fornell–Larcker criterion. The square roots of AVE for ETGAI, PJR, MS, and TPD were .703, .709, .711, and .716, respectively, all exceeding the largest inter-construct correlations, supporting discriminant validity. Harman's single-factor test indicated that the single factor accounted for 28.3% of total variance, below the 50% threshold, suggesting common method bias was not a serious concern. Table 2 Descriptive Statistics, Reliability Estimates, and Inter-Construct Correlations Variable M SD α CR AVE 1 2 3 1. ETGAI 3.49 0.56 .895 .921 .494 — 2. PJR 3.50 0.58 .870 .901 .503 .423** — 3. MS 3.55 0.58 .868 .902 .505 .218** .422** — 4. TPD 3.52 0.62 .831 .863 .512 .210** −.085 −.273** Note. M = mean; SD = standard deviation; α = Cronbach's alpha; CR = composite reliability; AVE = average variance extracted. ETGAI = Epistemic Trust in Generative AI; PJR = Professional Judgment Reshaping; MS = Moral Stress; TPD = Teacher Professional Development. Diagonal entries are square roots of AVE. **p < .01. 4.3 Structural Model and Hypothesis Testing The structural model was estimated using maximum likelihood (ML) estimation with 5,000 bootstrap replications in Mplus 8.3. As shown in Table 3 , the model demonstrated excellent fit. R² values for the endogenous variables were: PJR = .227, MS = .240, and TPD = .219. Table 3 Structural Model Fit Indices Model χ² df p CFI TLI RMSEA SRMR Structural model 601.30 590 .365 .998 .998 .007 [.000, .017] .035 Threshold — — > .05 > .95 > .95 < .06 < .08 Note. RMSEA = root mean square error of approximation; SRMR = standardized root mean square residual; CFI = comparative fit index; TLI = Tucker–Lewis index. Thresholds from Hu and Bentler ( 1999 ) and Kline ( 2016 ). Table 4 presents standardized path coefficients and bootstrap confidence intervals for all hypothesized effects. H1 predicted that epistemic trust in generative AI (ETGAI) would positively predict professional judgment reshaping (PJR). Results supported this hypothesis (β = .476, SE = .042, p < .001, 95% CI [.388, .556]), indicating that higher levels of epistemic trust in GenAI were associated with greater AI-driven reconfiguration of teachers' professional judgment. H2 predicted that professional judgment reshaping (PJR) would positively predict moral stress (MS). This hypothesis was supported (β = .490, SE = .045, p < .001, 95% CI [.397, .598]), suggesting that AI-driven reshaping of professional judgment generates substantial moral tensions among university EFL teachers. H3 predicted that moral stress (MS) would negatively predict teacher professional development (TPD). Results confirmed this hypothesis (β = −.405, SE = .049, p < .001, 95% CI [− .499, − .272]), consistent with Conservation of Resources Theory (Hobfoll, 1989 ), whereby moral stress depletes psychological resources necessary for professional growth. H4 predicted a positive direct effect of epistemic trust in GenAI (ETGAI) on teacher professional development (TPD), which was supported (β = .346, SE = .054, p < .001, 95% CI [.238, .484]), demonstrating that epistemic trust simultaneously functions as a motivational resource facilitating professional development. Table 4 Standardized Path Coefficients and Mediation Effects (N = 420) Path / Effect β SE p 95% CI LL 95% CI UL Hypothesis Panel A: Direct Effects ETGAI → PJR .476 .042 < .001 .388 .556 H1 Supported PJR → MS .490 .045 < .001 .397 .598 H2 Supported MS → TPD −.405 .049 < .001 −.499 −.272 H3 Supported ETGAI → TPD (direct) .346 .054 < .001 .238 .484 H4 Supported Panel B: Indirect Effects — Bootstrap BC 95% CI (5,000 replications) ETGAI → PJR → MS .233 .031 < .001 .172 .295 H5 Supported PJR → MS → TPD −.199 .033 < .001 −.266 −.119 H6 Supported ETGAI → PJR → MS → TPD −.095 .019 < .001 −.133 −.061 H7 Supported Note. β = standardized coefficient; SE = standard error; CI = confidence interval; LL = lower limit; UL = upper limit. Bootstrap bias-corrected (BC) 95% CIs based on 5,000 replications. ETGAI = Epistemic Trust in Generative AI; PJR = Professional Judgment Reshaping; MS = Moral Stress; TPD = Teacher Professional Development. 4.4 Mediation Analysis H5 proposed that professional judgment reshaping (PJR) would mediate the relationship between epistemic trust in GenAI (ETGAI) and moral stress (MS). The indirect effect was significant (β = .233, SE = .031, p < .001, 95% BC CI [.172, .295]), with the confidence interval excluding zero, thus supporting H5. H6 proposed that moral stress (MS) would mediate the relationship between professional judgment reshaping (PJR) and teacher professional development (TPD). The indirect effect was significant (β = −.199, SE = .033, p < .001, 95% BC CI [− .266, − .119]), supporting H6. H7 proposed the full serial mediation pathway: ETGAI → PJR → MS → TPD. This hypothesis represents the core theoretical contribution of the present study. Results confirmed a significant serial indirect effect (β = −.095, SE = .019, p < .001, 95% BC CI [− .133, − .061]). The confidence interval excluded zero, providing empirical support for the moral stress pathway: epistemic trust in GenAI triggers a sequential process in which professional judgment reshaping amplifies moral stress, which in turn undermines teacher professional development. H7 was supported. Taken together, all seven hypotheses were supported. The standardized path diagram is presented in Fig. 1 . 5. Discussion This study proposed and empirically tested a serial mediation model linking epistemic trust in generative AI (ETGAI) to teacher professional development (TPD) through professional judgment reshaping (PJR) and moral stress (MS) among 420 university EFL teachers in China. All seven hypotheses were supported, yielding several theoretically significant and practically consequential findings. The following sections discuss the implications of each pathway in turn, situate the findings within the broader literature, and consider their cross-disciplinary significance. 5.1 The Positive Effect of Epistemic Trust on Professional Judgment Reshaping (H1) The significant positive effect of ETGAI on PJR (β = .476) indicates that teachers who regard GenAI as a credible and reliable epistemic source are substantially more likely to reconfigure their habitual patterns of professional decision-making around AI inputs. This finding is consistent with theoretical predictions derived from epistemic trust theory (Pandey et al., 2025 , 2026 ; Sedlakova et al., 2025 ): when a novel agent is perceived as epistemically trustworthy, individuals revise their information-seeking and decision-making behaviors accordingly. In professional teaching contexts, this means that epistemic trust functions as the psychological gateway through which AI adoption moves from superficial tool use to deeper integration into professional cognition. The magnitude of this effect is notable. It suggests that epistemic trust is a stronger predictor of professional judgment reshaping than general technology acceptance or behavioral intention—a distinction with important implications for professional development interventions. Chen et al. ( 2026 ) documented ethnographically how EFL teachers' "teacherness" was renegotiated in response to AI integration, and the present findings provide quantitative confirmation that this renegotiation is primarily driven by epistemic, rather than merely instrumental, trust. Erbay-çetinkaya ( 2025 ) similarly found that reflective teacher identity was significantly reshaped through AI-mediated collaborative learning—processes that appear to be mediated by epistemic trust dynamics. The present results suggest that supporting teachers in developing calibrated, critical epistemic trust—neither wholesale acceptance nor blanket rejection—may be a more effective professional development target than generic AI skills training. 5.2 Professional Judgment Reshaping and Moral Stress (H2) The significant positive effect of PJR on MS (β = .490) is perhaps the most theoretically consequential finding of this study. It establishes empirically that the process of deferring professional judgment to AI systems generates genuine moral stress in university EFL teachers—not merely discomfort or uncertainty, but a sustained form of ethical tension that, as subsequent results demonstrate, carries measurable consequences for professional development. Vassallo ( 2026 ) theorized the "AI guilt complex" as a phenomenological response to AI adoption in academic settings, and the present findings provide structural confirmation that this emotional cluster is systematically predicted by the degree of professional judgment reshaping that has occurred. This finding resonates with cross-disciplinary evidence from healthcare settings. Drawing on parallel evidence, Awad et al. ( 2026 ) demonstrated that AI integration mediates the link between moral distress and professional integrity in critical care nursing—a mechanism structurally analogous to our finding that PJR amplifies MS, which in turn undermines TPD. Similarly, Barr ( 2025 ) documented in neonatal intensive care nurses that the relationship between moral distress and professional departure intention was mediated by burnout dimensions, underscoring the cascade of consequences that follows from sustained moral stress in professional settings. The educational parallel is clear: when teachers experience persistent moral stress from AI-driven judgment reshaping, professional resources are depleted in ways that constrain growth. Yang ( 2026 ) further observed that Chinese patients experiencing moral anxiety about AI-assisted diagnosis exhibited similar ambivalence—simultaneously experiencing epistemic empowerment and psychological discomfort—suggesting that this tension may be a cross-domain characteristic of high-stakes AI adoption in expert contexts. 5.3 Moral Stress and Teacher Professional Development (H3) The negative effect of MS on TPD (β = −.405) is consistent with predictions derived from Conservation of Resources (COR) theory (Hobfoll, 1989 ). Moral stress depletes the psychological resources—including professional self-efficacy, motivational energy, and ethical coherence—that are necessary for sustained engagement in professional learning. When these resources are depleted, teachers may withdraw from professional development activities, reduce their willingness to experiment with new pedagogical approaches, or develop defensive avoidance strategies in relation to GenAI—all of which constrain professional growth. Lipschuetz and Topaz ( 2026 ) argued in the nursing context that the moral dimensions of digital transformation are frequently neglected in institutional technology adoption agendas, with the result that efficiency gains from AI come at the cost of professional wellbeing and growth. The present findings suggest that a parallel dynamic may operate in university EFL education. Su et al. ( 2026 ) found that digital resilience in nursing students was significantly undermined by psychological distress from technology-mediated role conflict—a mechanism that appears functionally equivalent to the moral-stress-to-TPD pathway identified here. These cross-professional parallels strengthen our confidence that the COR-based mechanism is not specific to teaching, but reflects a general feature of professional adaptation to AI in expert work contexts. 5.4 The Direct Positive Effect of Epistemic Trust on Teacher Professional Development (H4) The significant positive direct effect of ETGAI on TPD (β = .346)—controlling for the mediated pathways—indicates that epistemic trust in GenAI also functions as a motivational resource that directly facilitates professional growth. This finding reflects the productive dimension of epistemic trust: teachers who regard GenAI as a reliable professional resource are more likely to engage with AI tools in exploratory and developmental ways, experiment with novel pedagogical approaches, and leverage AI-generated content to expand their professional repertoire (Zhang & Wang, 2025 ; Zhao et al., 2025 ). Roe et al. ( 2025 ), in an international study examining higher education stakeholder perceptions of synthetic AI avatars, found that perceived reliability of AI—which overlaps substantially with epistemic trust—was associated with greater openness to AI-mediated professional learning. Sharma and Rai ( 2026 ) similarly documented that Indian higher education educators who expressed trust in ChatGPT for teaching support reported greater professional learning gains from AI integration. The coexistence of positive direct and negative mediated effects of ETGAI on TPD—through the PJR → MS pathway—suggests that epistemic trust has a dual character in professional development contexts. At low to moderate levels of PJR, the direct motivational benefits of epistemic trust dominate, supporting TPD. As PJR intensifies, however, the moral stress mechanism becomes increasingly salient, attenuating or reversing the direct benefit. This pattern is consistent with a resource-based interpretation: moderate reliance on GenAI enables efficient professional learning, while extensive judgment deferral creates moral strain that depletes the very resources that epistemic trust would otherwise mobilize. 5.5 The Serial Mediation Pathway (H5, H6, H7): The Moral Stress Pathway The confirmation of the full serial mediation pathway (ETGAI → PJR → MS → TPD; β = −.095, 95% BC CI [− .133, − .061]) constitutes the core theoretical contribution of this study. This finding reveals what we term the moral stress pathway: a sequential psychological process through which the epistemic disposition to trust GenAI—when it catalyzes sufficient professional judgment reshaping—ultimately undermines teacher professional development via the depletion of psychological resources associated with moral stress. To our knowledge, this is the first empirical study to identify and validate this complete serial pathway in an EFL teacher population, and one of the first in any educational context. The magnitude of the serial indirect effect (β = −.095), while modest relative to the direct effects, is entirely consistent with expectations for serial mediation models, in which the chaining of multiple mediation steps necessarily attenuates the overall indirect effect (Hayes, 2018 ). The significance and directionality of the effect, combined with the precision of the bootstrap confidence interval, provide strong evidence that this pathway is real, meaningful, and theoretically interpretable. It adds a dimension of explanatory depth to the otherwise positive story of GenAI adoption: epistemic trust in GenAI can simultaneously facilitate direct professional growth and, through the moral stress mechanism, constrain it—an irony that reflects the inherent complexity of integrating powerful autonomous AI systems into expert professional practice. The H5 finding—that PJR mediates ETGAI → MS (β = .233)—establishes that professional judgment reshaping is not merely a benign cognitive adaptation but a psychologically consequential process that generates moral tension. The H6 finding—that MS mediates PJR → TPD (β = −.199)—establishes that moral stress is a genuine mechanism rather than a mere correlate of constrained development. Together, H5 and H6 identify two nodes in the causal chain where intervention might interrupt the moral stress pathway: by supporting teachers in maintaining professional agency and identity amid AI integration (reducing PJR-induced MS), and by providing moral support resources that help teachers process and metabolize moral stress before it depletes developmental motivation. 5.6 Theoretical Implications The present study makes several theoretical contributions. First, it extends epistemic trust theory from its original clinical and developmental psychology contexts (Yirmiya & Fonagy, 2025 kés & Doorn, 2026 ) to the domain of professional education, demonstrating that the epistemic dimension of AI trust—distinct from general technology acceptance—has unique predictive power for professional psychological and developmental outcomes. Second, it advances moral stress theory beyond healthcare settings (Ansari et al., 2025 ; Frush & Gaffney, 2025 ; Trueblood et al., 2025 ) into educational contexts, documenting that the construct is theoretically coherent and empirically tractable in university EFL teacher populations. Third, the study contributes to COR-based models of professional development by identifying moral stress as a resource-depleting mechanism specifically generated by AI-induced professional judgment reshaping—a novel application of COR theory to the digital transformation context. The serial mediation architecture of the model also contributes to the methodological literature on AI adoption research in education. Most existing studies examine either attitudinal predictors of AI adoption or aggregate outcomes of adoption, without attending to the psychological mechanisms that connect them (Lee et al., 2025 ; Fatalaki et al., 2025 ; Mutlu, 2025 ). The present model demonstrates that the full causal story from epistemic trust to professional development requires attention to at least two intermediate psychological processes—and that these processes carry opposite valences, generating both positive (direct) and negative (indirect) effects of ETGAI on TPD. 5.7 Practical Implications The findings carry several practical implications for institutions, teacher educators, and individual teachers. At the institutional level, the results suggest that AI integration policies for university EFL programs should extend beyond infrastructure and skills training to include explicit attention to the ethical dimensions of AI use and the psychological wellbeing of teachers navigating professional judgment reshaping. The finding that moral stress mediates negative effects on TPD implies that institutions that invest heavily in GenAI capability without providing parallel moral and emotional support risk undermining the very professional development gains they seek to promote. Regular structured forums for teachers to discuss ethical concerns about AI use—analogous to the ethics rounds and moral case deliberation sessions documented in healthcare settings (cf. Tabata-Kelly et al., 2026 )—may help to contain moral stress accumulation. For teacher educators and professional development designers, the study suggests that critical AI epistemic literacy—the capacity to evaluate AI-generated content with appropriate skepticism and maintain professional judgment autonomy—should be a core component of AI-era teacher education programs (Watts, 2025 ; Sawyer et al., 2025 ). Roe et al. ( 2025 ) found that higher education stakeholders were generally willing to engage productively with AI tools when provided with frameworks for evaluating AI reliability and maintaining critical distance. Developing such frameworks specifically for EFL teachers, and embedding them in pre-service and in-service professional development, appears warranted by the present findings. For individual teachers, the results may tentatively suggest the value of reflective practice focused on the ethical dimensions of GenAI use—not merely on whether and how to use AI tools, but on how AI use is affecting their professional judgment, values, and sense of professional identity. Pham and Le ( 2026 ) found that structured co-reflection with AI, when approached ethically and critically, could generate meaningful professional learning without the moral costs of unreflective AI deference. Such reflective practices, particularly when conducted in professional learning communities, may provide a protective buffer against moral stress accumulation. 6. Conclusion This study investigated the psychological and developmental consequences of epistemic trust in generative AI among 420 university EFL teachers in China, proposing and validating a serial mediation model in which professional judgment reshaping and moral stress sequentially transmit the effects of epistemic trust on teacher professional development. All seven hypotheses were supported. Epistemic trust in GenAI positively predicted professional judgment reshaping, which in turn generated moral stress; moral stress negatively predicted teacher professional development; and epistemic trust exerted a significant positive direct effect on professional development. The full serial indirect pathway (ETGAI → PJR → MS → TPD) was significant, revealing a moral stress pathway through which epistemic trust in GenAI ultimately constrains professional development when mediated by extensive judgment reshaping. These findings advance understanding of the psychological dynamics of AI integration in professional teaching contexts in three ways. Theoretically, they extend epistemic trust theory, Conservation of Resources theory, and moral stress frameworks to university EFL education, proposing a novel explanatory model that captures both the beneficial and constraining effects of epistemic trust in GenAI. Empirically, they provide the first validated structural evidence for the moral stress pathway in EFL teacher professional development, drawing on a well-powered sample with rigorously validated instrumentation. Practically, they identify actionable leverage points—professional judgment autonomy support, ethical AI literacy training, and moral support resources—for mitigating the negative developmental consequences of moral stress in AI-integrated teaching environments. 6.1 Limitations Several limitations should be noted. First, the cross-sectional design precludes causal inference; longitudinal or experimental designs are needed to establish directionality with greater confidence. Second, data were collected via self-report, which introduces potential social desirability bias; future studies might supplement surveys with behavioral measures or experience sampling. Third, the sample, while sizable, was drawn through non-probabilistic methods and is limited to Chinese university EFL contexts; generalizability to other national, institutional, or disciplinary contexts requires empirical verification. Fourth, the newly adapted PJR and MS scales, while demonstrating adequate psychometric properties in the present study, require further cross-cultural validation before use in other populations. Fifth, the present model does not include potential moderating variables—such as institutional support, AI self-efficacy, or professional identity strength—that may condition the proposed pathways. 6.2 Future Directions Future research might profitably pursue several lines of inquiry. Longitudinal studies tracking teachers across AI adoption phases would strengthen causal claims and illuminate the temporal dynamics of moral stress accumulation and professional development. Cross-cultural comparative designs—examining, for example, whether the moral stress pathway operates similarly in Western higher education contexts—would address generalizability constraints. Qualitative research exploring the lived experience of moral stress in GenAI-integrated EFL teaching would provide narrative depth to the structural patterns identified here, and might yield richer insight into the specific professional value conflicts that drive moral stress. Intervention studies testing the efficacy of critical AI epistemic literacy programs, ethical reflection frameworks, and institutional moral support structures would translate the present findings into evidence-based professional development practice. Finally, future models might investigate moderating variables—including institutional AI governance quality, teacher professional identity strength, and peer community support—that may buffer or amplify the moral stress pathway identified in this study. Declarations Ethics approval and consent to participate. This study was approved by the Yinchuan Municipal Education Bureau Social Science Ethics Approval Committee (YMEBSSEAC), Yinchuan, Ningxia Province, Approval No. 86. Informed consent was obtained from all participants prior to data collection. All procedures complied with the Declaration of Helsinki. YMEBSSEAC-0086. (2025.9.18) Consent for publication. Not applicable. Competing interests. The authors declare no competing interests. Clinical trial number. Not applicable. Funding. No funding was received for this study. Author Contribution The author is solely responsible for the conception, design, data collection, analysis, interpretation of the results, and writing of this manuscript. No other individuals contributed to the research. Data Availability The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request. References Alreiahi NJ, Alrwaished N. Integrating AI tools into preservice mathematics teacher education: A qualitative study of lesson planning practices. Contemp Educational Technol. 2025;17(4):ep617. https://doi.org/10.30935/cedtech/17549 . Ansari N, Wilson R, Warner E, Taylor-Swanson L, Van Epps J, Iacob E, Supiano K. We are not heroes: Examining nurses' experiences of moral distress during COVID-19. Teach Learn Nurs. 2025;20(2):e368–75. https://doi.org/10.1016/j.teln.2024.11.021 . Awad NHA, Ashour HMA, Abouelala RASA, Yaseen MM, Awad WHA. When distress meets technology: The mediating role of AI integration in the link between nurses' moral distress and moral integrity in critical care settings. Appl Nurs Res. 2026;88:152065. https://doi.org/10.1016/j.apnr.2026.152065 . Barr P. Mediation effects of burnout measure dimensions on the relationship between moral distress and having considered leaving: A regression-based correlational study in neonatal intensive care unit nurses. Australian Crit Care. 2025;38(6):101446. https://doi.org/10.1016/j.aucc.2025.101446 . Békés V, Doorn KAV. The most vulnerable are prone to use AI therapists: The role of attachment, epistemic trust, and mental health symptoms in acceptance of digital mental health interventions. Psychother Res. 2026. https://doi.org/10.1080/10503307.2026.2615388 . Boyd RL, Markowitz DM. Artificial intelligence and the psychology of human connection. Perspect Psychol Sci. 2026. https://doi.org/10.1177/17456916251404394 . Chen QH, Tawilapakul U, Lin AMY. Teacherness in the age of GenAI: An ethnographic exploration of pedagogical decision-making and AI integration in english language teaching. System. 2026;136:103877. https://doi.org/10.1016/j.system.2025.103877 . Chen XX, Pi XJ. Assessing EFL students' GenAI-assisted writing: Teachers' pains, perceptions and practices. Assess Writ. 2026;67:100996. https://doi.org/10.1016/j.asw.2025.100996 . Erbay-Çetinkaya S. A triadic support framework for reflective teacher identity: Peer, AI, and faculty collaboration within a community of practice. Teach Teacher Educ. 2025;168:105239. https://doi.org/10.1016/j.tate.2025.105239 . Fatalaki JA, Saeedi Z, Marand ES, Firuz ARN. Artificial intelligence in education and EFL teachers' attitudes: Scale development and validation. Innov Lang Learn Teach. 2025. https://doi.org/10.1080/17501229.2025.2494823 . Fornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981;18(1):39–50. https://doi.org/10.1177/002224378101800104 . Frush BW, Gaffney S. Scarcity, futility, and moral distress: A case analysis of the challenges of sound ethical description. J Pain Symptom Manag. 2025;70(5). https://doi.org/10.1016/j.jpainsymman.2025.07.025 . Hair JF, Risher JJ, Sarstedt M, Ringle CM. When to use and how to report the results of PLS-SEM. Eur Bus Rev. 2019;31(1):2–24. https://doi.org/10.1108/EBR-11-2018-0203 . Hayes AF. Introduction to mediation, moderation, and conditional process analysis. 2nd ed. Guilford Press; 2018. Hobfoll SE. Conservation of resources: A new attempt at conceptualizing stress. Am Psychol. 1989;44(3):513–24. https://doi.org/10.1037/0003-066X.44.3.513 . Hoang NH. Cultural-linguistic congruence as a moderator of teacher emotions: Examining AI implementation in Vietnamese EFL classrooms. RELC J. 2025. https://doi.org/10.1177/00336882251367785 . Hu HZ, Zhou Q, Hashim H. Negotiating identity in the age of ChatGPT: Non-native english researchers' experiences with AI-assisted academic writing. Humanit Social Sci Commun. 2025;12(1):965. https://doi.org/10.1057/s41599-025-05351-4 . Hu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equ Model. 1999;6(1):1–55. https://doi.org/10.1080/10705519909540118 . Kline RB. Principles and practice of structural equation modeling. 4th ed. Guilford Press; 2016. Küçükuncular A, Ertugan A. Teaching in today's world: A Marxian critique of AI in education. J Acad Ethics. 2025;24(1):18. https://doi.org/10.1007/s10805-025-09692-2 . Lee S, Jeon J, McKinley J, Rose H. Generative AI and english language teaching: A global Englishes perspective. Annu Rev Appl Linguist. 2025;45:85–108. https://doi.org/10.1017/S0267190525100184 . Li BX, Bhattarai A, Ding ZQ. In search of humanness: Professional identities of qualitative research educators in the age of generative AI. Learn Media Technol. 2025. https://doi.org/10.1080/17439884.2025.2521547 . Lipschuetz M, Topaz M. (2026). Prioritizing nurse-patient relationships in the digital transformation of nursing: A discussion paper. International Journal of Nursing Studies , 177 , 105369. https://doi.org/10.1016/j.ijnurstu.2026.105369 McCarthy DR. A journey to trust for AI: Civil-military relations and epistemic authority in American socio-technical imaginaries of artificial intelligence. Contemp Secur Policy. 2026. https://doi.org/10.1080/13523260.2025.2607593 . Mutlu AK. Unveiling Turkish EFL teachers' generative artificial intelligence (GenAI) literacy: Voices from low-resource contexts. J Educ Students Placed Risk. 2025. https://doi.org/10.1080/10824669.2025.2579972 . Muthén LK, Muthén BO. (1998–2017). Mplus user's guide (8th ed.). Muthén & Muthén. Pandey CS, Mishra P, Pandey SR, Pandey S. Epistemic trust in generative AI for higher education scale (ETGAI-HE scale). AI Soc. 2026;41(2):1387–400. https://doi.org/10.1007/s00146-025-02566-6 . Pandey SR, Pandey CS, Mishra P, Pandey S, Kumar M. Mapping the evolution of epistemic trust in artificial intelligence: A bibliometric analysis of key themes, influences, and global trends. Aslib J Inform Manage. 2025. https://doi.org/10.1108/AJIM-11-2024-0898 . Pham TT, Le TT. Negotiating reflection and ethics in co-reflection with AI: A narrative inquiry of Vietnamese EFL teachers' experiences. Reflective Pract. 2026. https://doi.org/10.1080/14623943.2026.2631417 . Podsakoff PM, MacKenzie SB, Lee JY, Podsakoff NP. Common method biases in behavioral research: A critical review of the literature and recommended remedies. J Appl Psychol. 2003;88(5):879–903. https://doi.org/10.1037/0021-9010.88.5.879 . Roe J, Perkins M, Somoray K, Miller D, Furze L. Can synthetic avatars replace lecturers? An exploratory international study of higher education stakeholder perceptions. Int J Educational Technol High Educ. 2025;22(1):71. https://doi.org/10.1186/s41239-025-00568-4 . Sacco F, Manzi F, Di Dio C, Schepman A, Rodway P, Massaro D, Marchetti A. Validation of the general attitudes towards artificial intelligence scale in the Italian context and the role of trust. Computers Hum Behav Rep. 2025;19:100751. https://doi.org/10.1016/j.chbr.2025.100751 . Sánchez-Trujillo MD, Bernabé-Sánchez E, Sáenz-Egúsquiza FD. (2025). Artificial intelligence and teacher training: Perspectives of education students. Eduweb – Revista de Tecnología de Información y Comunicación en Educación , 19 (3), 22–34. https://doi.org/10.46502/issn.1856-7576/2025.19.03.2 Sawyer K, Dave VS, Gonyeau MJ, Cain J. Teaching tomorrow's pharmacists in an AI world: Risk, responsibility, and reflection. Am J Pharm Educ. 2025;89(12):101905. https://doi.org/10.1016/j.ajpe.2025.101905 . Sedlakova J, Lucivero F, Pavarini G, Kerasidou A. Human-like epistemic trust? A conceptual and normative analysis of conversational AI in mental healthcare. Am J Bioeth. 2025. https://doi.org/10.1080/15265161.2025.2526734 . Sharma K, Rai P. Integrating ChatGPT into online teaching: An interpretative phenomenological study of Indian higher education educators. Am J Distance Educ. 2026. https://doi.org/10.1080/08923647.2026.2627737 . Su Y, Yu ZY, Chen S, Chen J, Wang L. Exploring digital resilience and its related factors among nursing students in the era of AI. Nurse Educ Pract. 2026;90:104636. https://doi.org/10.1016/j.nepr.2025.104636 . Tabata-Kelly M, Bulger AL, Reich AJ, Gazarian PK, Izumi S, Gray TF, Wichmann L, Rinaldi S, Bernacki RE. Nurses' experiences of serious illness conversations with marginalized patients in COVID-19. J Pain Symptom Manag. 2026;71(1):24–42. https://doi.org/10.1016/j.jpainsymman.2025.09.001 . Trueblood AE, Vogelsmeier A, Scott SD, Rantz M. Psychological distress of nurse managers in acute care: A scoping review. Nurs Outlook. 2025;73(4):102447. https://doi.org/10.1016/j.outlook.2025.102447 . Vassallo D. The AI guilt complex: Moral emotions and ethical dilemmas in academic technology adoption. J Acad Ethics. 2026;24(1):53. https://doi.org/10.1007/s10805-026-09726-3 . Watts KJ. Paying the cognitive debt: An experiential learning framework for integrating AI in social work education. Educ Sci. 2025;15(10):1304. https://doi.org/10.3390/educsci15101304 . Woo BFY, Cato K, Cho H, You B, S., Song J. The use of large language models in clinical documentation: A scoping review. Int J Nurs Stud. 2026;176:105322. https://doi.org/10.1016/j.ijnurstu.2025.105322 . Xia Q, Yang YM, Weng XJ, Cilsalar-Sagnak H, Cheng WK, Chiu TKF. Generative artificial intelligence competency self-efficacy scale for teachers and researchers (GAICS-TR) in higher education. J Computers Educ. 2025. https://doi.org/10.1007/s40692-025-00373-y . Yang RH. Between epistemic empowerment and moral anxiety: Chinese patients' ambivalence toward AI-assisted diagnosis. Soc Sci Med. 2026;391:118929. https://doi.org/10.1016/j.socscimed.2026.118929 . Yirmiya K, Fonagy P. Mentalizing without a mind: Psychotherapeutic potential of generative AI. J Med Internet Res. 2025;27:e79156. https://doi.org/10.2196/79156 . Zainuddin Z, Churiyah M, Rahayu WP, Pratikto H, Halili SH. E-portfolios as authentic assessment in pre-service teacher education: Enhancing reflection, professional identity, and digital preparedness. Educ Inform Technol. 2026. https://doi.org/10.1007/s10639-025-13869-y . Zhang H, Wang Q. How could GenAI work on in-service teachers' knowledge building process? An empirical study based on epistemic network analysis. Int J Educational Technol High Educ. 2025;22(1):47. https://doi.org/10.1186/s41239-025-00544-y . Zhao ZJ, Li JJ, Hong YJ, Yun TY. Lesson study as an approach to facilitate the integration of Gen-AI into EFL curriculum design in higher education. Int J Lesson Learn Stud. 2025. https://doi.org/10.1108/IJLLS-01-2025-0029 . 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-9071480","acceptedTermsAndConditions":true,"allowDirectSubmit":true,"archivedVersions":[],"articleType":"Research Article","associatedPublications":[],"authors":[{"id":623146730,"identity":"0b15521d-3297-4eb2-bb71-9a6a7d0442f2","order_by":0,"name":"Gai Meijuan","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA6klEQVRIie3PsWrDMBCAYSmC03Ik64VCn0FBQ1rIw9gYPOUBMoTGRuA1L2DcV8jU2UGQLHmAjHYM3bpnKoldOmSxPBaqHw0nuG84xny+P9iYMcHYip5B8KQM7sPERaAjp4UeS5NW1elFT5MhhGdxWODRzOpsFe6chKJGI1ieUXh/BfEdE/Xl3EtiHSFaAR35IDFnoPWylyyFRbLwS+A1QXhyE2Xxh+SEqhxAIgxiAtwbFSZEboKfepaXCwUyTavgQGpqHLdMZNTQ1zdt3o2s9tf122YrTd30kbYRPnyFY72NXwcs+Xw+3z/uBo6GRTzl3O/MAAAAAElFTkSuQmCC","orcid":"","institution":"Shandong Xiehe University","correspondingAuthor":true,"prefix":"","firstName":"Gai","middleName":"","lastName":"Meijuan","suffix":""}],"badges":[],"createdAt":"2026-03-09 10:08:46","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-9071480/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-9071480/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":107189003,"identity":"f6248acd-ee3f-4745-9715-c8ebd1e58aca","added_by":"auto","created_at":"2026-04-17 20:13:54","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":435952,"visible":true,"origin":"","legend":"\u003cp\u003eStandardized Path Diagram of the Serial Mediation Model\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eNote. \u003c/strong\u003eOvals = latent variables; rectangles = observed indicator items. Standardized path coefficients (β) shown on structural paths. Factor loadings (λ) shown on measurement paths. R² values displayed above endogenous latent variables. Direct path (H4) shown as curved arrow below. ***p \u0026lt; .001.\u003c/p\u003e","description":"","filename":"floatimage1.png","url":"https://assets-eu.researchsquare.com/files/rs-9071480/v1/10971d02ba368a984c6296c2.png"},{"id":107515439,"identity":"1c458276-3ddb-434b-99b8-fc827d0f3463","added_by":"auto","created_at":"2026-04-22 08:28:48","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":902823,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-9071480/v1/80dd7df9-b47a-4006-80e1-e441006866e5.pdf"}],"financialInterests":"No competing interests reported.","formattedTitle":"Epistemic Trust in Generative AI and the Reshaping of Professional Judgment:","fulltext":[{"header":"1. Introduction","content":"\u003cp\u003eThe rapid proliferation of generative artificial intelligence tools\u0026mdash;including ChatGPT, Gemini, and DeepSeek\u0026mdash;has fundamentally altered the landscape of English as a Foreign Language (EFL) teaching at the university level. Within a remarkably short span, these tools have moved from peripheral novelties to central instruments in lesson planning, feedback provision, assessment design, and student interaction (Lee et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Yet the integration of GenAI into professional teaching practice is far from a straightforward technical adoption; it carries profound epistemic, ethical, and psychological dimensions that the existing literature has only begun to explore.\u003c/p\u003e \u003cp\u003eCentral among these dimensions is the question of epistemic trust. Epistemic trust\u0026mdash;originally theorized as the disposition to accept communicated knowledge as relevant and reliable (Fonagy \u0026amp; colleagues; Yirmiya \u0026amp; Fonagy, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2025\u003c/span\u003e)\u0026mdash;has emerged as a pivotal construct in understanding how professionals engage with AI-generated information. When teachers extend epistemic trust to GenAI systems, they are, in effect, delegating a portion of their professional judgment to an external agent whose reasoning processes remain partially opaque. Pandey et al. (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) have demonstrated that epistemic trust in GenAI is a distinct and measurable construct in higher education contexts, significantly predicting engagement patterns and information acceptance behaviors. Yet the downstream consequences of this trust\u0026mdash;particularly for teachers whose professional identities are bound up in the exercise of expert judgment\u0026mdash;remain poorly understood.\u003c/p\u003e \u003cp\u003eOne important consequence that appears underexplored is the reshaping of professional judgment. As teachers come to rely on GenAI for content generation, assessment feedback, and pedagogical decisions, the locus of professional judgment may shift in ways that threaten professional agency and autonomy (Chen et al., \u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; K\u0026uuml;\u0026ccedil;\u0026uuml;kuncular \u0026amp; Ertugan, \u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This process of professional judgment reshaping (PJR) is not inherently negative; it may enable teachers to perform more efficiently and creatively. However, it also creates the conditions for moral stress: a form of psychological tension arising when the external reconfiguration of one's professional role conflicts with deeply held values about pedagogical authenticity, academic integrity, and professional responsibility (Vassallo, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Yang, \u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2026\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eDespite the conceptual plausibility of this pathway, no empirical study has yet traced the full serial mediation chain from epistemic trust in GenAI through professional judgment reshaping and moral stress to teacher professional development. The existing literature on GenAI in EFL education has largely focused on attitudes and adoption (Fatalaki et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mutlu, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), perceptions of student writing (Chen \u0026amp; Pi, \u003cspan citationid=\"CR8\" class=\"CitationRef\"\u003e2026\u003c/span\u003e), or identity negotiation (Hu et al., \u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), without attending to the psychological stress mechanisms that mediate these relationships. Meanwhile, research on moral stress has developed primarily in healthcare settings (Ansari et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Awad et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Barr, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), with limited translation to educational contexts.\u003c/p\u003e \u003cp\u003eThe present study addresses this gap by investigating the following research questions: (1) Does epistemic trust in GenAI positively predict professional judgment reshaping among university EFL teachers in China? (2) Does professional judgment reshaping in turn predict moral stress? (3) Does moral stress mediate the relationship between professional judgment reshaping and teacher professional development? (4) Do professional judgment reshaping and moral stress jointly and serially mediate the relationship between epistemic trust in GenAI and teacher professional development? Using data from 420 university EFL teachers in China, and employing covariance-based SEM with bootstrapped confidence intervals, this study tests a theoretically grounded serial mediation model (ETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD).\u003c/p\u003e \u003cp\u003eThe study makes three contributions. Theoretically, it extends epistemic trust theory and Conservation of Resources (COR) theory to the AI-integrated teaching context, proposing and empirically validating a novel serial mediation model. Methodologically, it provides a rigorously validated instrument battery adapted for Chinese university EFL contexts. Practically, it offers evidence-based implications for institutional policy, teacher education programs, and individual reflective practice in the rapidly evolving GenAI era. The remainder of the paper is organized as follows: Section \u003cspan refid=\"Sec2\" class=\"InternalRef\"\u003e2\u003c/span\u003e reviews the literature and develops the theoretical framework; Section \u003cspan refid=\"Sec10\" class=\"InternalRef\"\u003e3\u003c/span\u003e describes the methodology; Section \u003cspan refid=\"Sec17\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents the results; Section \u003cspan refid=\"Sec22\" class=\"InternalRef\"\u003e5\u003c/span\u003e discusses the findings; and Section \u003cspan refid=\"Sec30\" class=\"InternalRef\"\u003e6\u003c/span\u003e offers conclusions, limitations, and future directions.\u003c/p\u003e"},{"header":"2. Literature Review and Theoretical Framework","content":"\u003cdiv id=\"Sec3\" class=\"Section2\"\u003e \u003ch2\u003e2.1 Epistemic Trust: From Developmental Psychology to Artificial Intelligence\u003c/h2\u003e \u003cp\u003eEpistemic trust was originally conceptualized in developmental psychology as the fundamental human disposition to regard communicated information as genuine, relevant, and applicable to the self (Yirmiya \u0026amp; Fonagy, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). In its original formulation, epistemic trust serves as an adaptive mechanism that enables efficient cultural learning: rather than independently verifying every piece of received knowledge, individuals selectively extend trust to sources they perceive as credible, benevolent, and contextually appropriate. Critically, this trust is not blind credulity; it coexists with epistemic vigilance\u0026mdash;a parallel evaluative disposition that screens for unreliability or deception (Sedlakova et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe extension of epistemic trust theory to technological agents is a relatively recent development, accelerated by the widespread deployment of large language models. McCarthy (\u003cspan citationid=\"CR24\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) traces the discursive construction of AI as an epistemic authority in public imaginaries, arguing that institutional trust in AI involves a form of delegated epistemic authority analogous to trust in expert human informants. Sedlakova et al. (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) develop a normative analysis of \"human-like epistemic trust\" in conversational AI, concluding that while AI systems can elicit trust responses phenomenologically similar to those directed at humans, important differences in reliability, accountability, and relational reciprocity warrant caution. B\u0026eacute;k\u0026eacute;s and Doorn (\u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) further document that individuals with higher attachment anxiety and mental health vulnerabilities exhibit significantly elevated epistemic trust in AI agents, suggesting that epistemic trust in AI is a heterogeneous construct shaped by individual difference variables.\u003c/p\u003e \u003cp\u003eIn the specific domain of higher education, Pandey et al. (\u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) developed and validated the Epistemic Trust in Generative AI for Higher Education Scale (ETGAI-HE), demonstrating that epistemic trust in GenAI is empirically distinguishable from general AI acceptance and predicts qualitatively distinct behavioral patterns. A companion bibliometric study (Pandey et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) mapped the evolution of this construct across disciplines, identifying education and health as the two sectors in which epistemic trust in AI has attracted the greatest research attention. Sacco et al. (\u003cspan citationid=\"CR32\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) found that epistemic trust\u0026mdash;operationalized as openness to accepting AI output\u0026mdash;moderated the relationship between general AI attitudes and behavioral intentions, underscoring its role as a psychological mechanism rather than merely an attitudinal disposition. Boyd and Markowitz (\u003cspan citationid=\"CR6\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) further argue that AI systems that simulate epistemic responsiveness\u0026mdash;appearing to attend to, understand, and personalize information\u0026mdash;are particularly effective at eliciting trust, with potentially significant implications for professional reliance on AI outputs.\u003c/p\u003e \u003cp\u003eFor EFL teachers, epistemic trust in GenAI carries a specific professional valence. Language teachers exercise professional judgment over matters of linguistic accuracy, pedagogical appropriateness, and cultural sensitivity\u0026mdash;domains in which GenAI systems demonstrate uneven competence. The decision to extend epistemic trust to GenAI in these domains, therefore, represents a consequential professional act with downstream effects on instructional quality, professional identity, and ethical practice.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec4\" class=\"Section2\"\u003e \u003ch2\u003e2.2 Generative AI in University EFL Teaching Contexts\u003c/h2\u003e \u003cp\u003eThe uptake of GenAI tools among EFL teachers has been rapid and uneven. Lee et al. (\u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) provide a comprehensive overview from a global Englishes perspective, documenting that while GenAI offers significant affordances for language teaching\u0026mdash;including on-demand text generation, adaptive feedback, and multilingual capability\u0026mdash;it also raises fundamental questions about native-speakerism, linguistic authority, and the role of the teacher as language model. Chen et al. (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) conducted an ethnographic investigation of how EFL teachers navigate pedagogical decision-making in GenAI-integrated classrooms, coining the notion of \"teacherness\" to describe the constellation of professional attitudes and practices that teachers mobilize to maintain pedagogical agency in the face of AI encroachment.\u003c/p\u003e \u003cp\u003eIn Chinese university contexts specifically, the pressures of digital transformation and national AI development agendas have accelerated GenAI adoption in ways that often outpace teacher preparation and institutional support (Zhao et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mutlu, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Hoang (\u003cspan citationid=\"CR16\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) documents how cultural-linguistic factors moderate teacher emotional responses to AI implementation in Vietnamese EFL classrooms\u0026mdash;findings with potential relevance to Chinese contexts given shared Confucian educational values. Hu et al. (\u003cspan citationid=\"CR17\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) explore how non-native English researchers negotiate professional identity when using ChatGPT for academic writing, documenting tensions between efficiency gains and concerns about authentic scholarly voice\u0026mdash;a dynamic that parallels the identity challenges facing EFL teachers who rely on GenAI for professional tasks.\u003c/p\u003e \u003cp\u003ePham and Le (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) offer a narrative inquiry of Vietnamese EFL teachers' experiences of co-reflection with AI, revealing that while AI co-reflection affords new forms of reflective depth, it simultaneously raises ethical concerns about the nature of professional judgment when externally mediated. Similarly, Fatalaki et al. (\u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) developed and validated an instrument measuring EFL teachers' attitudes toward AI in education, finding that perceived professional utility was the strongest predictor of positive attitudes, though concerns about accuracy and academic integrity remained significant barriers. Zhang and Wang (\u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) employed epistemic network analysis to examine how GenAI influences in-service teachers' knowledge-building processes, finding evidence of both cognitive enrichment and epistemic dependency\u0026mdash;a pattern consistent with the dual-edged nature of epistemic trust proposed in this study.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec5\" class=\"Section2\"\u003e \u003ch2\u003e2.3 Professional Judgment Reshaping in AI-Integrated Practice\u003c/h2\u003e \u003cp\u003eProfessional judgment in teaching refers to the expert decision-making capacity that enables teachers to read complex classroom situations and respond appropriately, drawing on a synthesis of content knowledge, pedagogical knowledge, relational knowledge, and professional values (Erbay-\u0026ccedil;etinkaya, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This capacity is not static; it develops through experience, reflection, and engagement with professional communities. Critically, professional judgment is also the site at which teachers exercise professional agency\u0026mdash;the capacity to act intentionally on the basis of professional goals and values, rather than merely responding to external directives or automated cues (Li et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eThe introduction of GenAI into professional practice creates conditions for what we term professional judgment reshaping: the process by which teachers' habitual patterns of professional decision-making are reconfigured through sustained interaction with AI systems. This reshaping is not synonymous with de-skilling or replacement; it may involve enhanced efficiency, expanded creativity, or novel forms of human-AI collaboration (Alreiahi \u0026amp; Alrwaished, \u003cspan citationid=\"CR1\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Xia et al., \u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). However, when epistemic trust leads teachers to defer systematically to AI recommendations in domains that were previously the province of professional expertise, the result may be a gradual attenuation of autonomous professional judgment\u0026mdash;a phenomenon related to automation bias documented in other high-stakes professional domains (Roe et al., \u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eS\u0026aacute;nchez-Trujillo et al. (\u003cspan citationid=\"CR33\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) found that pre-service teachers with higher general AI acceptance showed greater willingness to adopt AI-generated lesson plans with minimal modification, a pattern suggestive of incipient judgment reshaping even at the beginning stages of professional formation. K\u0026uuml;\u0026ccedil;\u0026uuml;kuncular and Ertugan (\u003cspan citationid=\"CR20\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) offer a more critical perspective, arguing from a Marxian standpoint that AI-mediated professional judgment involves a form of intellectual labor extraction that may ultimately undermine teacher professional identity. Together, these studies suggest that professional judgment reshaping is a real and consequential process, though its psychological and developmental effects have yet to be systematically examined. Based on the conceptual logic that higher epistemic trust in GenAI promotes greater reliance on AI in professional decision-making, we propose:\u003c/p\u003e \u003cp\u003eH1: Epistemic trust in generative AI positively predicts professional judgment reshaping among university EFL teachers.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec6\" class=\"Section2\"\u003e \u003ch2\u003e2.4 Moral Stress in AI-Integrated Teaching\u003c/h2\u003e \u003cp\u003e The concept of moral stress draws on a substantial literature originating in healthcare settings, where moral distress was originally defined as the psychological tension arising when professionals know the ethically appropriate course of action but are constrained by external forces from acting accordingly (cf. Ansari et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Frush \u0026amp; Gaffney, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Subsequent elaborations have emphasized the role of role conflict, value misalignment, and institutional constraints in generating sustained moral stress that depletes psychological resources and contributes to professional burnout (Barr, \u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Trueblood et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eIn educational contexts, moral stress from AI adoption takes distinctive forms. Vassallo (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) conceptualizes what he terms the \"AI guilt complex\"\u0026mdash;a cluster of moral emotions including guilt, complicity, and ethical uncertainty experienced by academics who adopt GenAI tools while remaining uncertain about their professional and ethical implications. Yang (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) documents a parallel phenomenon among Chinese patients who experience moral anxiety alongside epistemic empowerment when engaging with AI-assisted medical diagnosis\u0026mdash;a dynamic remarkably consonant with what EFL teachers may experience when GenAI reshapes their professional judgment.\u003c/p\u003e \u003cp\u003eDrawing on cross-disciplinary evidence, Awad et al. (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) demonstrated in a nursing context that AI integration mediates the relationship between moral distress and professional integrity\u0026mdash;a finding with important structural parallels to the mechanism proposed here. When professional judgment reshaping creates discrepancies between teachers' performed professional role (relying on AI outputs) and their internalized professional values (exercising independent expert judgment), the resultant cognitive dissonance is expected to manifest as moral stress. This stress is further amplified by the opacity of GenAI reasoning processes, which makes it difficult for teachers to evaluate the epistemological provenance of AI-generated content (Sedlakova et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). We therefore propose:\u003c/p\u003e \u003cp\u003eH2: Professional judgment reshaping positively predicts moral stress among university EFL teachers.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec7\" class=\"Section2\"\u003e \u003ch2\u003e2.5 Moral Stress and Teacher Professional Development\u003c/h2\u003e \u003cp\u003eTeacher professional development (TPD) encompasses the full range of processes through which teachers expand their professional knowledge, refine their skills, develop their identities, and enhance their adaptive capacity in response to changing educational demands (Zainuddin et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Sharma \u0026amp; Rai, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2026\u003c/span\u003e). In the AI era, TPD increasingly involves the development of critical AI literacy, ethical reasoning about technology use, and the ability to coordinate human and machine competencies in service of learning (Watts, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Sawyer et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eConservation of Resources (COR) theory (Hobfoll, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1989\u003c/span\u003e) provides a theoretical framework for predicting how moral stress affects TPD. COR theory proposes that individuals strive to acquire, maintain, and protect valued resources\u0026mdash;including psychological resources such as professional identity, self-efficacy, and motivational energy. Stress arises when resources are threatened, lost, or fail to be gained following significant investment. Moral stress, which depletes the psychological resource of ethical coherence and professional self-regard, is expected to impair the cognitive and motivational resources necessary for professional learning and development. Consistent with this reasoning, Lipschuetz and Topaz (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) found in a nursing digital transformation context that moral distress significantly constrained nurses' capacity for professional adaptation and learning\u0026mdash;a pattern analogous to what we predict for EFL teachers facing GenAI-induced moral stress.\u003c/p\u003e \u003cp\u003eFurthermore, digital resilience research in nursing education (Su et al., \u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) has documented that psychological distress from technology-mediated role conflict undermines professional learning engagement\u0026mdash;a cross-disciplinary parallel supporting our prediction. Woo et al. (\u003cspan citationid=\"CR42\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) similarly noted in a scoping review of LLM use in clinical documentation that clinician distrust and moral concerns about AI use were associated with reduced professional adoption and development outcomes. We therefore propose:\u003c/p\u003e \u003cp\u003eH3: Moral stress negatively predicts teacher professional development.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec8\" class=\"Section2\"\u003e \u003ch2\u003e2.6 Direct Effect of Epistemic Trust on Teacher Professional Development\u003c/h2\u003e \u003cp\u003eWhile the mediated pathway through PJR and MS is the central theoretical contribution of this study, epistemic trust in GenAI may also exert a direct positive effect on TPD. When teachers extend epistemic trust to GenAI, they are more likely to engage productively with AI tools, experiment with novel pedagogical approaches, and leverage AI-generated resources in ways that expand their professional repertoire (Zhang \u0026amp; Wang, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). This direct effect represents the beneficial dimension of epistemic trust\u0026mdash;the productive engagement with GenAI as a professional resource\u0026mdash;and is expected to operate independently of the moral stress pathway. Consistent with this reasoning:\u003c/p\u003e \u003cp\u003eH4: Epistemic trust in generative AI positively predicts teacher professional development.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec9\" class=\"Section2\"\u003e \u003ch2\u003e2.7 Serial Mediation Model and Hypotheses Summary\u003c/h2\u003e \u003cp\u003eIntegrating the theoretical arguments developed above, we propose a serial mediation model in which the relationship between ETGAI and TPD is transmitted through two sequentially ordered psychological mechanisms: professional judgment reshaping (M1) and moral stress (M2). This model draws on three theoretical pillars: (1) epistemic trust theory, which accounts for the conditions under which professionals accept AI-generated knowledge as authoritative; (2) Conservation of Resources theory (Hobfoll, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1989\u003c/span\u003e), which predicts how moral stress depletes the psychological resources necessary for professional development; and (3) professional agency frameworks (Erbay-\u0026ccedil;etinkaya, \u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Li et al., \u003cspan citationid=\"CR22\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), which explain why the reconfiguration of professional judgment by AI constitutes a psychologically significant event for professional identity and wellbeing.\u003c/p\u003e \u003cp\u003eThe mediation hypotheses are as follows:\u003c/p\u003e \u003cp\u003eH5: Professional judgment reshaping mediates the relationship between epistemic trust in generative AI and moral stress.\u003c/p\u003e \u003cp\u003eH6: Moral stress mediates the relationship between professional judgment reshaping and teacher professional development.\u003c/p\u003e \u003cp\u003eH7: Professional judgment reshaping and moral stress serially mediate the relationship between epistemic trust in generative AI and teacher professional development (ETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD).\u003c/p\u003e \u003c/div\u003e"},{"header":"3. Methodology","content":"\u003cdiv id=\"Sec11\" class=\"Section2\"\u003e \u003ch2\u003e3.1 Research Design\u003c/h2\u003e \u003cp\u003eThis study employed a quantitative, cross-sectional survey design using covariance-based structural equation modeling (CBSEM) to test the proposed serial mediation model. A two-stage analytic approach was adopted: a confirmatory factor analysis (CFA) to evaluate the measurement model, followed by structural model estimation with bootstrapped confidence intervals for mediation testing (Kline, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Cross-sectional SEM has been widely used in EFL teacher cognition research and AI adoption studies (Fatalaki et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Zhang \u0026amp; Wang, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), and is appropriate for simultaneously testing complex systems of relationships among latent variables.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec12\" class=\"Section2\"\u003e \u003ch2\u003e3.2 Participants and Sampling\u003c/h2\u003e \u003cp\u003eThe target population comprised in-service EFL teachers employed at universities and higher education institutions in mainland China who reported current awareness of, or experience with, GenAI tools in their professional practice. Participants were recruited via a combination of convenience and snowball sampling through professional WeChat groups, EFL teacher networks affiliated with provincial English teaching associations, and personal referrals. The sampling strategy, while not probabilistic, reflects common practice in educational SEM research and is consistent with methodological precedents in the field (Lee et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mutlu, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e).\u003c/p\u003e \u003cp\u003eFollowing SEM sample size recommendations (Kline, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2016\u003c/span\u003e), a target of at least 300 responses was established. A total of 456 questionnaires were collected, of which 420 were retained after excluding responses with more than 10% missing data, systematic patterns indicating random responding, or failure on attention check items. The final analytic sample comprised 420 participants (n male\u0026thinsp;=\u0026thinsp;126, 30.0%; n female\u0026thinsp;=\u0026thinsp;286, 68.1%; n other/undisclosed\u0026thinsp;=\u0026thinsp;8, 1.9%). Teaching experience ranged from 1 to over 16 years. Full demographic details are presented in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e (Section \u003cspan refid=\"Sec17\" class=\"InternalRef\"\u003e4\u003c/span\u003e). The sample size satisfies the 10:1 participant-to-parameter ratio recommended for stable CFA solutions (Hair et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e).\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec13\" class=\"Section2\"\u003e \u003ch2\u003e3.3 Measures\u003c/h2\u003e \u003cp\u003e \u003cb\u003eEpistemic Trust in Generative AI (ETGAI).\u003c/b\u003e \u003c/p\u003e \u003cp\u003eETGAI was measured using 12 items adapted from the Epistemic Trust in Generative AI for Higher Education Scale (ETGAI-HE; Pandey et al., \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2026\u003c/span\u003e), supplemented by items drawing on the bibliometric-informed framework of Pandey et al. (\u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) and the conceptual analysis of Sedlakova et al. (\u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Items assess the degree to which respondents regard GenAI systems as accurate, credible, and reliable epistemic sources in professional teaching contexts (e.g., \"I trust the information generated by GenAI tools to be professionally accurate and reliable\"). Response options ranged from 1 (strongly disagree) to 7 (strongly agree). In the present sample, Cronbach's α\u0026thinsp;=\u0026thinsp;.895 and composite reliability (CR) = .921.\u003c/p\u003e \u003cp\u003e \u003cb\u003eProfessional Judgment Reshaping (PJR).\u003c/b\u003e \u003c/p\u003e \u003cp\u003ePJR was measured using 10 items developed de novo for this study, drawing on conceptual frameworks from Chen et al. (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2026\u003c/span\u003e), Erbay-\u0026ccedil;etinkaya (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), and Xia et al. (\u003cspan citationid=\"CR43\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Items assess the degree to which respondents' habitual patterns of professional decision-making\u0026mdash;including curriculum design, assessment, and pedagogical choice\u0026mdash;have been reconfigured through reliance on GenAI outputs (e.g., \"I find myself deferring to GenAI recommendations in areas where I previously exercised independent professional judgment\"). Response options ranged from 1 to 7. Cronbach's α\u0026thinsp;=\u0026thinsp;.870, CR = .901.\u003c/p\u003e \u003cp\u003e \u003cb\u003eMoral Stress (MS).\u003c/b\u003e \u003c/p\u003e \u003cp\u003eMS was assessed using 8 items adapted from cross-disciplinary moral distress instruments (cf. Vassallo, \u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Awad et al., \u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Ansari et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) and modified to reflect EFL teacher contexts. Items capture the experience of ethical tension, professional role conflict, and value misalignment arising from AI-mediated teaching practice (e.g., \"Using GenAI in my teaching creates an uncomfortable tension between efficiency and my professional responsibility to exercise independent judgment\"). Response options ranged from 1 to 7. Cronbach's α\u0026thinsp;=\u0026thinsp;.868, CR = .902.\u003c/p\u003e \u003cp\u003e \u003cb\u003eTeacher Professional Development (TPD).\u003c/b\u003e \u003c/p\u003e \u003cp\u003eTPD was measured using 10 items adapted from validated teacher professional development scales (cf. Zainuddin et al., \u003cspan citationid=\"CR46\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Sharma \u0026amp; Rai, \u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Watts, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), modified to reflect AI-era professional learning. Items assess engagement in professional learning activities, development of AI-related competencies, and reflective professional growth (e.g., \"My engagement with GenAI has contributed meaningfully to my professional knowledge and skills as an EFL teacher\"). Response options ranged from 1 to 7. Cronbach's α\u0026thinsp;=\u0026thinsp;.831, CR = .863.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec14\" class=\"Section2\"\u003e \u003ch2\u003e3.4 Translation and Pilot Testing\u003c/h2\u003e \u003cp\u003eAll items were originally drafted in English and translated into Mandarin Chinese following a forward-backward translation procedure with two bilingual translators. Discrepancies were resolved by consensus. A pilot study involving 35 university EFL teachers (excluded from the main analysis) was conducted to assess item clarity, response distribution, and internal consistency. Minor wording revisions were made to three items based on cognitive interview feedback prior to full-scale data collection.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec15\" class=\"Section2\"\u003e \u003ch2\u003e3.5 Common Method Bias\u003c/h2\u003e \u003cp\u003eTo mitigate common method bias arising from single-source self-report data, procedural remedies recommended by Podsakoff et al. (\u003cspan citationid=\"CR30\" class=\"CitationRef\"\u003e2003\u003c/span\u003e) were employed: the survey was administered anonymously, predictor and criterion items were separated by unrelated filler items, and scale anchors were varied across constructs. Post hoc, Harman's single-factor test was conducted; the single extracted factor accounted for 28.3% of total variance\u0026mdash;well below the 50% threshold\u0026mdash;suggesting that common method bias was not a substantial concern in the present data.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec16\" class=\"Section2\"\u003e \u003ch2\u003e3.6 Data Analysis\u003c/h2\u003e \u003cp\u003eData analysis proceeded in two stages using Mplus 8.3 (Muth\u0026eacute;n \u0026amp; Muth\u0026eacute;n, 1998\u0026ndash;2017). In Stage 1, a CFA was conducted to evaluate the measurement model. Fit was assessed using multiple indices: the comparative fit index (CFI), Tucker-Lewis index (TLI), root mean square error of approximation (RMSEA), and standardized root mean square residual (SRMR), with accepted thresholds of CFI \u0026gt; .95, TLI \u0026gt; .95, RMSEA \u0026lt; .06, and SRMR \u0026lt; .08 (Hu \u0026amp; Bentler, \u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e1999\u003c/span\u003e; Kline, \u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2016\u003c/span\u003e). Convergent validity was evaluated through standardized factor loadings (\u0026gt;\u0026thinsp;.50), average variance extracted (AVE \u0026gt; .50), and composite reliability (CR \u0026gt; .70; Fornell \u0026amp; Larcker, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1981\u003c/span\u003e). Discriminant validity was assessed using the Fornell-Larcker criterion. In Stage 2, the structural model was estimated using maximum likelihood estimation with 5,000 bootstrap replications to obtain bias-corrected (BC) 95% confidence intervals for direct and indirect effects.\u003c/p\u003e \u003c/div\u003e"},{"header":"4. Results","content":"\u003cdiv id=\"Sec18\" class=\"Section2\"\u003e \u003ch2\u003e4.1 Sample Characteristics\u003c/h2\u003e \u003cp\u003eA total of 420 university EFL teachers in China completed the survey. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab1\" class=\"InternalRef\"\u003e1\u003c/span\u003e, the sample comprised 126 males (30.0%), 286 females (68.1%), and 8 participants who identified as other or preferred not to disclose (1.9%). The majority of respondents fell in the 26\u0026ndash;35 age bracket (38.1%), followed by 36\u0026ndash;45 (35.0%). In terms of teaching experience, 30.0% reported 4\u0026ndash;8 years, 29.0% reported 9\u0026ndash;15 years, 22.9% had 16 or more years, and 18.1% had 1\u0026ndash;3 years. Regarding institution type, the largest group taught at regular higher education institutions (45.0%), followed by 985/211 key universities (21.9%), vocational colleges (20.0%), and other institutions (13.1%). In terms of generative AI usage frequency, 31.0% reported using GenAI tools almost daily, 38.1% one to two times per week, 21.9% one to two times per month, and 9.0% rarely or never.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab1\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 1\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eSample Characteristics (N\u0026thinsp;=\u0026thinsp;420)\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=\"char\" char=\".\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eCategory\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003en\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003e%\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGender\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eMale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e126\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eFemale\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e286\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e68.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOther/No disclose\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e8\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e1.9%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eAge\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;26\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e34\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e8.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e26\u0026ndash;35\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e160\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e38.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e36\u0026ndash;45\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e147\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e35.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;46\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e79\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e18.8%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eTeaching experience\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;3 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e76\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e18.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e4\u0026ndash;8 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e126\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e30.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e9\u0026ndash;15 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e122\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e29.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026ge;\u0026thinsp;16 years\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e96\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e22.9%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eInstitution type\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e985/211\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e21.9%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRegular HEI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e189\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e45.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eVocational\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e84\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e20.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eOther\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e13.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eGenAI usage frequency\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eRarely\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e38\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e9.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;2\u0026times;/month\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e92\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e21.9%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e1\u0026ndash;2\u0026times;/week\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e160\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e38.1%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003eDaily\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e130\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c4\"\u003e \u003cp\u003e31.0%\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"4\"\u003e\u003cb\u003eNote.\u003c/b\u003e HEI\u0026thinsp;=\u0026thinsp;higher education institution.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec19\" class=\"Section2\"\u003e \u003ch2\u003e4.2 Measurement Model\u003c/h2\u003e \u003cp\u003ePrior to structural model testing, a confirmatory factor analysis (CFA) was conducted. The model demonstrated excellent fit: χ\u0026sup2;(590)\u0026thinsp;=\u0026thinsp;601.30, p = .365, CFI = .998, TLI = .998, RMSEA = .007 [90% CI: .000, .017], SRMR = .035 (see Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e). Table\u0026nbsp;\u003cspan refid=\"Tab2\" class=\"InternalRef\"\u003e2\u003c/span\u003e presents descriptive statistics, reliability estimates, and inter-construct correlations.\u003c/p\u003e \u003cp\u003eAll standardized factor loadings were statistically significant (p \u0026lt; .001) and ranged from .594 to .753, exceeding the recommended minimum of .50 (Hair et al., \u003cspan citationid=\"CR13\" class=\"CitationRef\"\u003e2019\u003c/span\u003e). Convergent validity was supported: average variance extracted (AVE) values ranged from .494 to .512, and composite reliability (CR) values ranged from .863 to .921, both meeting established thresholds (Fornell \u0026amp; Larcker, \u003cspan citationid=\"CR11\" class=\"CitationRef\"\u003e1981\u003c/span\u003e). Cronbach's α ranged from .831 to .895.\u003c/p\u003e \u003cp\u003eDiscriminant validity was assessed using the Fornell\u0026ndash;Larcker criterion. The square roots of AVE for ETGAI, PJR, MS, and TPD were .703, .709, .711, and .716, respectively, all exceeding the largest inter-construct correlations, supporting discriminant validity. Harman's single-factor test indicated that the single factor accounted for 28.3% of total variance, below the 50% threshold, suggesting common method bias was not a serious concern.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab2\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 2\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eDescriptive Statistics, Reliability Estimates, and Inter-Construct Correlations\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"9\"\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c1\" colnum=\"1\"\u003e\u003c/div\u003e \u003cdiv align=\"char\" char=\".\" 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 \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c9\" colnum=\"9\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eVariable\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eM\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSD\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003eα\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCR\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eAVE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003e1\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003e2\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c9\"\u003e \u003cp\u003e3\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e1. ETGAI\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.49\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.56\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.895\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.921\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.494\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e\u0026nbsp;\u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e2. PJR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.50\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.870\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.901\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.503\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e.423**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e\u0026nbsp;\u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e3. MS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.55\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.58\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.868\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.902\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.505\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e.218**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e.422**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003e4. TPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c2\"\u003e \u003cp\u003e3.52\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"char\" char=\".\" colname=\"c3\"\u003e \u003cp\u003e0.62\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.831\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.863\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.512\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e.210**\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026minus;.085\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c9\"\u003e \u003cp\u003e\u0026minus;.273**\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"9\"\u003e\u003cb\u003eNote.\u003c/b\u003e M\u0026thinsp;=\u0026thinsp;mean; SD\u0026thinsp;=\u0026thinsp;standard deviation; α\u0026thinsp;=\u0026thinsp;Cronbach's alpha; CR\u0026thinsp;=\u0026thinsp;composite reliability; AVE\u0026thinsp;=\u0026thinsp;average variance extracted. ETGAI\u0026thinsp;=\u0026thinsp;Epistemic Trust in Generative AI; PJR\u0026thinsp;=\u0026thinsp;Professional Judgment Reshaping; MS\u0026thinsp;=\u0026thinsp;Moral Stress; TPD\u0026thinsp;=\u0026thinsp;Teacher Professional Development. Diagonal entries are square roots of AVE. **p \u0026lt; .01.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec20\" class=\"Section2\"\u003e \u003ch2\u003e4.3 Structural Model and Hypothesis Testing\u003c/h2\u003e \u003cp\u003eThe structural model was estimated using maximum likelihood (ML) estimation with 5,000 bootstrap replications in Mplus 8.3. As shown in Table\u0026nbsp;\u003cspan refid=\"Tab3\" class=\"InternalRef\"\u003e3\u003c/span\u003e, the model demonstrated excellent fit. R\u0026sup2; values for the endogenous variables were: PJR = .227, MS = .240, and TPD = .219.\u003c/p\u003e \u003cp\u003e \u003cdiv class=\"gridtable\"\u003e\u003ctable float=\"Yes\" id=\"Tab3\" border=\"1\"\u003e \u003ccaption language=\"En\"\u003e \u003cdiv class=\"CaptionNumber\"\u003eTable 3\u003c/div\u003e \u003cdiv class=\"CaptionContent\"\u003e \u003cp\u003eStructural Model Fit Indices\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"8\"\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=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c8\" colnum=\"8\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003eModel\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eχ\u0026sup2;\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003edf\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003eCFI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003eTLI\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eRMSEA\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c8\"\u003e \u003cp\u003eSRMR\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eStructural model\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e601.30\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e590\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e.365\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.998\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.998\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e.007 [.000, .017]\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e.035\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eThreshold\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e\u0026mdash;\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026gt;\u0026thinsp;.05\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026gt;\u0026thinsp;.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026gt;\u0026thinsp;.95\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.06\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c8\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.08\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"8\"\u003e\u003cb\u003eNote.\u003c/b\u003e RMSEA\u0026thinsp;=\u0026thinsp;root mean square error of approximation; SRMR\u0026thinsp;=\u0026thinsp;standardized root mean square residual; CFI\u0026thinsp;=\u0026thinsp;comparative fit index; TLI\u0026thinsp;=\u0026thinsp;Tucker\u0026ndash;Lewis index. Thresholds from Hu and Bentler (\u003cspan citationid=\"CR18\" class=\"CitationRef\"\u003e1999\u003c/span\u003e) and Kline (\u003cspan citationid=\"CR19\" class=\"CitationRef\"\u003e2016\u003c/span\u003e).\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003cp\u003eTable\u0026nbsp;\u003cspan refid=\"Tab4\" class=\"InternalRef\"\u003e4\u003c/span\u003e presents standardized path coefficients and bootstrap confidence intervals for all hypothesized effects.\u003c/p\u003e \u003cp\u003eH1 predicted that epistemic trust in generative AI (ETGAI) would positively predict professional judgment reshaping (PJR). Results supported this hypothesis (β\u0026thinsp;=\u0026thinsp;.476, SE = .042, p \u0026lt; .001, 95% CI [.388, .556]), indicating that higher levels of epistemic trust in GenAI were associated with greater AI-driven reconfiguration of teachers' professional judgment.\u003c/p\u003e \u003cp\u003eH2 predicted that professional judgment reshaping (PJR) would positively predict moral stress (MS). This hypothesis was supported (β\u0026thinsp;=\u0026thinsp;.490, SE = .045, p \u0026lt; .001, 95% CI [.397, .598]), suggesting that AI-driven reshaping of professional judgment generates substantial moral tensions among university EFL teachers.\u003c/p\u003e \u003cp\u003eH3 predicted that moral stress (MS) would negatively predict teacher professional development (TPD). Results confirmed this hypothesis (β = \u0026minus;.405, SE = .049, p \u0026lt; .001, 95% CI [\u0026minus;\u0026thinsp;.499, \u0026minus;\u0026thinsp;.272]), consistent with Conservation of Resources Theory (Hobfoll, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1989\u003c/span\u003e), whereby moral stress depletes psychological resources necessary for professional growth.\u003c/p\u003e \u003cp\u003eH4 predicted a positive direct effect of epistemic trust in GenAI (ETGAI) on teacher professional development (TPD), which was supported (β\u0026thinsp;=\u0026thinsp;.346, SE = .054, p \u0026lt; .001, 95% CI [.238, .484]), demonstrating that epistemic trust simultaneously functions as a motivational resource facilitating professional development.\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\u003eStandardized Path Coefficients and Mediation Effects (N\u0026thinsp;=\u0026thinsp;420)\u003c/p\u003e \u003c/div\u003e \u003c/caption\u003e \u003ccolgroup cols=\"7\"\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=\"left\" class=\"colspec\" colname=\"c3\" colnum=\"3\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c4\" colnum=\"4\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c5\" colnum=\"5\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c6\" colnum=\"6\"\u003e\u003c/div\u003e \u003cdiv align=\"left\" class=\"colspec\" colname=\"c7\" colnum=\"7\"\u003e\u003c/div\u003e \u003cthead\u003e \u003ctr\u003e \u003cth align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePath / Effect\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c2\"\u003e \u003cp\u003eβ\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c3\"\u003e \u003cp\u003eSE\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c4\"\u003e \u003cp\u003ep\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c5\"\u003e \u003cp\u003e95% CI LL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c6\"\u003e \u003cp\u003e95% CI UL\u003c/p\u003e \u003c/th\u003e \u003cth align=\"left\" colname=\"c7\"\u003e \u003cp\u003eHypothesis\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003ctr\u003e \u003cth align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003ePanel A: Direct Effects\u003c/p\u003e \u003c/th\u003e \u003c/tr\u003e \u003c/thead\u003e \u003ctbody\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eETGAI \u0026rarr; PJR\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.476\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.042\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.388\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.556\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH1 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePJR \u0026rarr; MS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.490\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.045\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.397\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.598\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH2 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eMS \u0026rarr; TPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;.405\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.049\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026minus;.499\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026minus;.272\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH3 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eETGAI \u0026rarr; TPD (direct)\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.346\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.054\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.238\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.484\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH4 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colspan=\"7\" nameend=\"c7\" namest=\"c1\"\u003e \u003cp\u003e\u003cb\u003ePanel B: Indirect Effects \u0026mdash; Bootstrap BC 95% CI (5,000 replications)\u003c/b\u003e\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eETGAI \u0026rarr; PJR \u0026rarr; MS\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e.233\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.031\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e.172\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e.295\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH5 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003ePJR \u0026rarr; MS \u0026rarr; TPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;.199\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.033\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026minus;.266\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026minus;.119\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH6 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003ctr\u003e \u003ctd align=\"left\" colname=\"c1\"\u003e \u003cp\u003eETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c2\"\u003e \u003cp\u003e\u0026minus;.095\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c3\"\u003e \u003cp\u003e.019\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c4\"\u003e \u003cp\u003e\u0026lt;\u0026thinsp;.001\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c5\"\u003e \u003cp\u003e\u0026minus;.133\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c6\"\u003e \u003cp\u003e\u0026minus;.061\u003c/p\u003e \u003c/td\u003e \u003ctd align=\"left\" colname=\"c7\"\u003e \u003cp\u003eH7 Supported\u003c/p\u003e \u003c/td\u003e \u003c/tr\u003e \u003c/tbody\u003e \u003c/colgroup\u003e \u003ctfoot\u003e \u003ctr\u003e\u003ctd colspan=\"7\"\u003e\u003cb\u003eNote.\u003c/b\u003e β\u0026thinsp;=\u0026thinsp;standardized coefficient; SE\u0026thinsp;=\u0026thinsp;standard error; CI\u0026thinsp;=\u0026thinsp;confidence interval; LL\u0026thinsp;=\u0026thinsp;lower limit; UL\u0026thinsp;=\u0026thinsp;upper limit. Bootstrap bias-corrected (BC) 95% CIs based on 5,000 replications. ETGAI\u0026thinsp;=\u0026thinsp;Epistemic Trust in Generative AI; PJR\u0026thinsp;=\u0026thinsp;Professional Judgment Reshaping; MS\u0026thinsp;=\u0026thinsp;Moral Stress; TPD\u0026thinsp;=\u0026thinsp;Teacher Professional Development.\u003c/td\u003e\u003c/tr\u003e \u003c/tfoot\u003e \u003c/table\u003e\u003c/div\u003e \u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec21\" class=\"Section2\"\u003e \u003ch2\u003e4.4 Mediation Analysis\u003c/h2\u003e \u003cp\u003eH5 proposed that professional judgment reshaping (PJR) would mediate the relationship between epistemic trust in GenAI (ETGAI) and moral stress (MS). The indirect effect was significant (β\u0026thinsp;=\u0026thinsp;.233, SE = .031, p \u0026lt; .001, 95% BC CI [.172, .295]), with the confidence interval excluding zero, thus supporting H5.\u003c/p\u003e \u003cp\u003eH6 proposed that moral stress (MS) would mediate the relationship between professional judgment reshaping (PJR) and teacher professional development (TPD). The indirect effect was significant (β = \u0026minus;.199, SE = .033, p \u0026lt; .001, 95% BC CI [\u0026minus;\u0026thinsp;.266, \u0026minus;\u0026thinsp;.119]), supporting H6.\u003c/p\u003e \u003cp\u003eH7 proposed the full serial mediation pathway: ETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD. This hypothesis represents the core theoretical contribution of the present study. Results confirmed a significant serial indirect effect (β = \u0026minus;.095, SE = .019, p \u0026lt; .001, 95% BC CI [\u0026minus;\u0026thinsp;.133, \u0026minus;\u0026thinsp;.061]). The confidence interval excluded zero, providing empirical support for the moral stress pathway: epistemic trust in GenAI triggers a sequential process in which professional judgment reshaping amplifies moral stress, which in turn undermines teacher professional development. H7 was supported.\u003c/p\u003e \u003cp\u003eTaken together, all seven hypotheses were supported. The standardized path diagram is presented in Fig.\u0026nbsp;\u003cspan refid=\"Fig1\" class=\"InternalRef\"\u003e1\u003c/span\u003e.\u003c/p\u003e \u003c/div\u003e"},{"header":"5. Discussion","content":"\u003cp\u003eThis study proposed and empirically tested a serial mediation model linking epistemic trust in generative AI (ETGAI) to teacher professional development (TPD) through professional judgment reshaping (PJR) and moral stress (MS) among 420 university EFL teachers in China. All seven hypotheses were supported, yielding several theoretically significant and practically consequential findings. The following sections discuss the implications of each pathway in turn, situate the findings within the broader literature, and consider their cross-disciplinary significance.\u003c/p\u003e \u003cdiv id=\"Sec23\" class=\"Section2\"\u003e \u003ch2\u003e5.1 The Positive Effect of Epistemic Trust on Professional Judgment Reshaping (H1)\u003c/h2\u003e \u003cp\u003eThe significant positive effect of ETGAI on PJR (β\u0026thinsp;=\u0026thinsp;.476) indicates that teachers who regard GenAI as a credible and reliable epistemic source are substantially more likely to reconfigure their habitual patterns of professional decision-making around AI inputs. This finding is consistent with theoretical predictions derived from epistemic trust theory (Pandey et al., \u003cspan citationid=\"CR28\" class=\"CitationRef\"\u003e2025\u003c/span\u003e, \u003cspan citationid=\"CR27\" class=\"CitationRef\"\u003e2026\u003c/span\u003e; Sedlakova et al., \u003cspan citationid=\"CR35\" class=\"CitationRef\"\u003e2025\u003c/span\u003e): when a novel agent is perceived as epistemically trustworthy, individuals revise their information-seeking and decision-making behaviors accordingly. In professional teaching contexts, this means that epistemic trust functions as the psychological gateway through which AI adoption moves from superficial tool use to deeper integration into professional cognition.\u003c/p\u003e \u003cp\u003eThe magnitude of this effect is notable. It suggests that epistemic trust is a stronger predictor of professional judgment reshaping than general technology acceptance or behavioral intention\u0026mdash;a distinction with important implications for professional development interventions. Chen et al. (\u003cspan citationid=\"CR7\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) documented ethnographically how EFL teachers' \"teacherness\" was renegotiated in response to AI integration, and the present findings provide quantitative confirmation that this renegotiation is primarily driven by epistemic, rather than merely instrumental, trust. Erbay-\u0026ccedil;etinkaya (\u003cspan citationid=\"CR9\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) similarly found that reflective teacher identity was significantly reshaped through AI-mediated collaborative learning\u0026mdash;processes that appear to be mediated by epistemic trust dynamics. The present results suggest that supporting teachers in developing calibrated, critical epistemic trust\u0026mdash;neither wholesale acceptance nor blanket rejection\u0026mdash;may be a more effective professional development target than generic AI skills training.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec24\" class=\"Section2\"\u003e \u003ch2\u003e5.2 Professional Judgment Reshaping and Moral Stress (H2)\u003c/h2\u003e \u003cp\u003eThe significant positive effect of PJR on MS (β\u0026thinsp;=\u0026thinsp;.490) is perhaps the most theoretically consequential finding of this study. It establishes empirically that the process of deferring professional judgment to AI systems generates genuine moral stress in university EFL teachers\u0026mdash;not merely discomfort or uncertainty, but a sustained form of ethical tension that, as subsequent results demonstrate, carries measurable consequences for professional development. Vassallo (\u003cspan citationid=\"CR40\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) theorized the \"AI guilt complex\" as a phenomenological response to AI adoption in academic settings, and the present findings provide structural confirmation that this emotional cluster is systematically predicted by the degree of professional judgment reshaping that has occurred.\u003c/p\u003e \u003cp\u003eThis finding resonates with cross-disciplinary evidence from healthcare settings. Drawing on parallel evidence, Awad et al. (\u003cspan citationid=\"CR3\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) demonstrated that AI integration mediates the link between moral distress and professional integrity in critical care nursing\u0026mdash;a mechanism structurally analogous to our finding that PJR amplifies MS, which in turn undermines TPD. Similarly, Barr (\u003cspan citationid=\"CR4\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) documented in neonatal intensive care nurses that the relationship between moral distress and professional departure intention was mediated by burnout dimensions, underscoring the cascade of consequences that follows from sustained moral stress in professional settings. The educational parallel is clear: when teachers experience persistent moral stress from AI-driven judgment reshaping, professional resources are depleted in ways that constrain growth. Yang (\u003cspan citationid=\"CR44\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) further observed that Chinese patients experiencing moral anxiety about AI-assisted diagnosis exhibited similar ambivalence\u0026mdash;simultaneously experiencing epistemic empowerment and psychological discomfort\u0026mdash;suggesting that this tension may be a cross-domain characteristic of high-stakes AI adoption in expert contexts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec25\" class=\"Section2\"\u003e \u003ch2\u003e5.3 Moral Stress and Teacher Professional Development (H3)\u003c/h2\u003e \u003cp\u003eThe negative effect of MS on TPD (β = \u0026minus;.405) is consistent with predictions derived from Conservation of Resources (COR) theory (Hobfoll, \u003cspan citationid=\"CR15\" class=\"CitationRef\"\u003e1989\u003c/span\u003e). Moral stress depletes the psychological resources\u0026mdash;including professional self-efficacy, motivational energy, and ethical coherence\u0026mdash;that are necessary for sustained engagement in professional learning. When these resources are depleted, teachers may withdraw from professional development activities, reduce their willingness to experiment with new pedagogical approaches, or develop defensive avoidance strategies in relation to GenAI\u0026mdash;all of which constrain professional growth.\u003c/p\u003e \u003cp\u003eLipschuetz and Topaz (\u003cspan citationid=\"CR23\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) argued in the nursing context that the moral dimensions of digital transformation are frequently neglected in institutional technology adoption agendas, with the result that efficiency gains from AI come at the cost of professional wellbeing and growth. The present findings suggest that a parallel dynamic may operate in university EFL education. Su et al. (\u003cspan citationid=\"CR37\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) found that digital resilience in nursing students was significantly undermined by psychological distress from technology-mediated role conflict\u0026mdash;a mechanism that appears functionally equivalent to the moral-stress-to-TPD pathway identified here. These cross-professional parallels strengthen our confidence that the COR-based mechanism is not specific to teaching, but reflects a general feature of professional adaptation to AI in expert work contexts.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec26\" class=\"Section2\"\u003e \u003ch2\u003e5.4 The Direct Positive Effect of Epistemic Trust on Teacher Professional Development (H4)\u003c/h2\u003e \u003cp\u003eThe significant positive direct effect of ETGAI on TPD (β\u0026thinsp;=\u0026thinsp;.346)\u0026mdash;controlling for the mediated pathways\u0026mdash;indicates that epistemic trust in GenAI also functions as a motivational resource that directly facilitates professional growth. This finding reflects the productive dimension of epistemic trust: teachers who regard GenAI as a reliable professional resource are more likely to engage with AI tools in exploratory and developmental ways, experiment with novel pedagogical approaches, and leverage AI-generated content to expand their professional repertoire (Zhang \u0026amp; Wang, \u003cspan citationid=\"CR47\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Zhao et al., \u003cspan citationid=\"CR48\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Roe et al. (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2025\u003c/span\u003e), in an international study examining higher education stakeholder perceptions of synthetic AI avatars, found that perceived reliability of AI\u0026mdash;which overlaps substantially with epistemic trust\u0026mdash;was associated with greater openness to AI-mediated professional learning. Sharma and Rai (\u003cspan citationid=\"CR36\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) similarly documented that Indian higher education educators who expressed trust in ChatGPT for teaching support reported greater professional learning gains from AI integration.\u003c/p\u003e \u003cp\u003eThe coexistence of positive direct and negative mediated effects of ETGAI on TPD\u0026mdash;through the PJR \u0026rarr; MS pathway\u0026mdash;suggests that epistemic trust has a dual character in professional development contexts. At low to moderate levels of PJR, the direct motivational benefits of epistemic trust dominate, supporting TPD. As PJR intensifies, however, the moral stress mechanism becomes increasingly salient, attenuating or reversing the direct benefit. This pattern is consistent with a resource-based interpretation: moderate reliance on GenAI enables efficient professional learning, while extensive judgment deferral creates moral strain that depletes the very resources that epistemic trust would otherwise mobilize.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec27\" class=\"Section2\"\u003e \u003ch2\u003e5.5 The Serial Mediation Pathway (H5, H6, H7): The Moral Stress Pathway\u003c/h2\u003e \u003cp\u003eThe confirmation of the full serial mediation pathway (ETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD; β = \u0026minus;.095, 95% BC CI [\u0026minus;\u0026thinsp;.133, \u0026minus;\u0026thinsp;.061]) constitutes the core theoretical contribution of this study. This finding reveals what we term the moral stress pathway: a sequential psychological process through which the epistemic disposition to trust GenAI\u0026mdash;when it catalyzes sufficient professional judgment reshaping\u0026mdash;ultimately undermines teacher professional development via the depletion of psychological resources associated with moral stress. To our knowledge, this is the first empirical study to identify and validate this complete serial pathway in an EFL teacher population, and one of the first in any educational context.\u003c/p\u003e \u003cp\u003eThe magnitude of the serial indirect effect (β = \u0026minus;.095), while modest relative to the direct effects, is entirely consistent with expectations for serial mediation models, in which the chaining of multiple mediation steps necessarily attenuates the overall indirect effect (Hayes, \u003cspan citationid=\"CR14\" class=\"CitationRef\"\u003e2018\u003c/span\u003e). The significance and directionality of the effect, combined with the precision of the bootstrap confidence interval, provide strong evidence that this pathway is real, meaningful, and theoretically interpretable. It adds a dimension of explanatory depth to the otherwise positive story of GenAI adoption: epistemic trust in GenAI can simultaneously facilitate direct professional growth and, through the moral stress mechanism, constrain it\u0026mdash;an irony that reflects the inherent complexity of integrating powerful autonomous AI systems into expert professional practice.\u003c/p\u003e \u003cp\u003eThe H5 finding\u0026mdash;that PJR mediates ETGAI \u0026rarr; MS (β\u0026thinsp;=\u0026thinsp;.233)\u0026mdash;establishes that professional judgment reshaping is not merely a benign cognitive adaptation but a psychologically consequential process that generates moral tension. The H6 finding\u0026mdash;that MS mediates PJR \u0026rarr; TPD (β = \u0026minus;.199)\u0026mdash;establishes that moral stress is a genuine mechanism rather than a mere correlate of constrained development. Together, H5 and H6 identify two nodes in the causal chain where intervention might interrupt the moral stress pathway: by supporting teachers in maintaining professional agency and identity amid AI integration (reducing PJR-induced MS), and by providing moral support resources that help teachers process and metabolize moral stress before it depletes developmental motivation.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec28\" class=\"Section2\"\u003e \u003ch2\u003e5.6 Theoretical Implications\u003c/h2\u003e \u003cp\u003eThe present study makes several theoretical contributions. First, it extends epistemic trust theory from its original clinical and developmental psychology contexts (Yirmiya \u0026amp; Fonagy, \u003cspan citationid=\"CR45\" class=\"CitationRef\"\u003e2025\u003c/span\u003ek\u0026eacute;s \u0026amp; Doorn, \u003cspan citationid=\"CR5\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) to the domain of professional education, demonstrating that the epistemic dimension of AI trust\u0026mdash;distinct from general technology acceptance\u0026mdash;has unique predictive power for professional psychological and developmental outcomes. Second, it advances moral stress theory beyond healthcare settings (Ansari et al., \u003cspan citationid=\"CR2\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Frush \u0026amp; Gaffney, \u003cspan citationid=\"CR12\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Trueblood et al., \u003cspan citationid=\"CR39\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) into educational contexts, documenting that the construct is theoretically coherent and empirically tractable in university EFL teacher populations. Third, the study contributes to COR-based models of professional development by identifying moral stress as a resource-depleting mechanism specifically generated by AI-induced professional judgment reshaping\u0026mdash;a novel application of COR theory to the digital transformation context.\u003c/p\u003e \u003cp\u003eThe serial mediation architecture of the model also contributes to the methodological literature on AI adoption research in education. Most existing studies examine either attitudinal predictors of AI adoption or aggregate outcomes of adoption, without attending to the psychological mechanisms that connect them (Lee et al., \u003cspan citationid=\"CR21\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Fatalaki et al., \u003cspan citationid=\"CR10\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Mutlu, \u003cspan citationid=\"CR25\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). The present model demonstrates that the full causal story from epistemic trust to professional development requires attention to at least two intermediate psychological processes\u0026mdash;and that these processes carry opposite valences, generating both positive (direct) and negative (indirect) effects of ETGAI on TPD.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec29\" class=\"Section2\"\u003e \u003ch2\u003e5.7 Practical Implications\u003c/h2\u003e \u003cp\u003eThe findings carry several practical implications for institutions, teacher educators, and individual teachers. At the institutional level, the results suggest that AI integration policies for university EFL programs should extend beyond infrastructure and skills training to include explicit attention to the ethical dimensions of AI use and the psychological wellbeing of teachers navigating professional judgment reshaping. The finding that moral stress mediates negative effects on TPD implies that institutions that invest heavily in GenAI capability without providing parallel moral and emotional support risk undermining the very professional development gains they seek to promote. Regular structured forums for teachers to discuss ethical concerns about AI use\u0026mdash;analogous to the ethics rounds and moral case deliberation sessions documented in healthcare settings (cf. Tabata-Kelly et al., \u003cspan citationid=\"CR38\" class=\"CitationRef\"\u003e2026\u003c/span\u003e)\u0026mdash;may help to contain moral stress accumulation.\u003c/p\u003e \u003cp\u003eFor teacher educators and professional development designers, the study suggests that critical AI epistemic literacy\u0026mdash;the capacity to evaluate AI-generated content with appropriate skepticism and maintain professional judgment autonomy\u0026mdash;should be a core component of AI-era teacher education programs (Watts, \u003cspan citationid=\"CR41\" class=\"CitationRef\"\u003e2025\u003c/span\u003e; Sawyer et al., \u003cspan citationid=\"CR34\" class=\"CitationRef\"\u003e2025\u003c/span\u003e). Roe et al. (\u003cspan citationid=\"CR31\" class=\"CitationRef\"\u003e2025\u003c/span\u003e) found that higher education stakeholders were generally willing to engage productively with AI tools when provided with frameworks for evaluating AI reliability and maintaining critical distance. Developing such frameworks specifically for EFL teachers, and embedding them in pre-service and in-service professional development, appears warranted by the present findings.\u003c/p\u003e \u003cp\u003eFor individual teachers, the results may tentatively suggest the value of reflective practice focused on the ethical dimensions of GenAI use\u0026mdash;not merely on whether and how to use AI tools, but on how AI use is affecting their professional judgment, values, and sense of professional identity. Pham and Le (\u003cspan citationid=\"CR29\" class=\"CitationRef\"\u003e2026\u003c/span\u003e) found that structured co-reflection with AI, when approached ethically and critically, could generate meaningful professional learning without the moral costs of unreflective AI deference. Such reflective practices, particularly when conducted in professional learning communities, may provide a protective buffer against moral stress accumulation.\u003c/p\u003e \u003c/div\u003e"},{"header":"6. Conclusion","content":"\u003cp\u003eThis study investigated the psychological and developmental consequences of epistemic trust in generative AI among 420 university EFL teachers in China, proposing and validating a serial mediation model in which professional judgment reshaping and moral stress sequentially transmit the effects of epistemic trust on teacher professional development. All seven hypotheses were supported. Epistemic trust in GenAI positively predicted professional judgment reshaping, which in turn generated moral stress; moral stress negatively predicted teacher professional development; and epistemic trust exerted a significant positive direct effect on professional development. The full serial indirect pathway (ETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD) was significant, revealing a moral stress pathway through which epistemic trust in GenAI ultimately constrains professional development when mediated by extensive judgment reshaping.\u003c/p\u003e \u003cp\u003eThese findings advance understanding of the psychological dynamics of AI integration in professional teaching contexts in three ways. Theoretically, they extend epistemic trust theory, Conservation of Resources theory, and moral stress frameworks to university EFL education, proposing a novel explanatory model that captures both the beneficial and constraining effects of epistemic trust in GenAI. Empirically, they provide the first validated structural evidence for the moral stress pathway in EFL teacher professional development, drawing on a well-powered sample with rigorously validated instrumentation. Practically, they identify actionable leverage points\u0026mdash;professional judgment autonomy support, ethical AI literacy training, and moral support resources\u0026mdash;for mitigating the negative developmental consequences of moral stress in AI-integrated teaching environments.\u003c/p\u003e \u003cdiv id=\"Sec31\" class=\"Section2\"\u003e \u003ch2\u003e6.1 Limitations\u003c/h2\u003e \u003cp\u003eSeveral limitations should be noted. First, the cross-sectional design precludes causal inference; longitudinal or experimental designs are needed to establish directionality with greater confidence. Second, data were collected via self-report, which introduces potential social desirability bias; future studies might supplement surveys with behavioral measures or experience sampling. Third, the sample, while sizable, was drawn through non-probabilistic methods and is limited to Chinese university EFL contexts; generalizability to other national, institutional, or disciplinary contexts requires empirical verification. Fourth, the newly adapted PJR and MS scales, while demonstrating adequate psychometric properties in the present study, require further cross-cultural validation before use in other populations. Fifth, the present model does not include potential moderating variables\u0026mdash;such as institutional support, AI self-efficacy, or professional identity strength\u0026mdash;that may condition the proposed pathways.\u003c/p\u003e \u003c/div\u003e \u003cdiv id=\"Sec32\" class=\"Section2\"\u003e \u003ch2\u003e6.2 Future Directions\u003c/h2\u003e \u003cp\u003eFuture research might profitably pursue several lines of inquiry. Longitudinal studies tracking teachers across AI adoption phases would strengthen causal claims and illuminate the temporal dynamics of moral stress accumulation and professional development. Cross-cultural comparative designs\u0026mdash;examining, for example, whether the moral stress pathway operates similarly in Western higher education contexts\u0026mdash;would address generalizability constraints. Qualitative research exploring the lived experience of moral stress in GenAI-integrated EFL teaching would provide narrative depth to the structural patterns identified here, and might yield richer insight into the specific professional value conflicts that drive moral stress. Intervention studies testing the efficacy of critical AI epistemic literacy programs, ethical reflection frameworks, and institutional moral support structures would translate the present findings into evidence-based professional development practice. Finally, future models might investigate moderating variables\u0026mdash;including institutional AI governance quality, teacher professional identity strength, and peer community support\u0026mdash;that may buffer or amplify the moral stress pathway identified in this study.\u003c/p\u003e \u003c/div\u003e"},{"header":"Declarations","content":" \u003cp\u003e \u003cstrong\u003eEthics approval and consent to participate.\u003c/strong\u003e \u003cp\u003eThis study was approved by the Yinchuan Municipal Education Bureau Social Science Ethics Approval Committee (YMEBSSEAC), Yinchuan, Ningxia Province, Approval No. 86. Informed consent was obtained from all participants prior to data collection. All procedures complied with the Declaration of Helsinki. YMEBSSEAC-0086. (2025.9.18)\u003c/p\u003e \u003c/p\u003e \u003cp\u003e \u003cstrong\u003eConsent for publication.\u003c/strong\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eCompeting interests.\u003c/h2\u003e \u003cp\u003eThe authors declare no competing interests.\u003c/p\u003e \u003c/p\u003e\u003cp\u003e \u003ch2\u003eClinical trial number.\u003c/h2\u003e \u003cp\u003eNot applicable.\u003c/p\u003e \u003c/p\u003e\u003ch2\u003eFunding.\u003c/h2\u003e \u003cp\u003eNo funding was received for this study.\u003c/p\u003e\u003ch2\u003eAuthor Contribution\u003c/h2\u003e\u003cp\u003eThe author is solely responsible for the conception, design, data collection, analysis, interpretation of the results, and writing of this manuscript. No other individuals contributed to the research.\u003c/p\u003e\u003ch2\u003eData Availability\u003c/h2\u003e\u003cp\u003eThe datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\u003cli\u003e\u003cspan\u003eAlreiahi NJ, Alrwaished N. Integrating AI tools into preservice mathematics teacher education: A qualitative study of lesson planning practices. Contemp Educational Technol. 2025;17(4):ep617. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.30935/cedtech/17549\u003c/span\u003e\u003cspan address=\"10.30935/cedtech/17549\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAnsari N, Wilson R, Warner E, Taylor-Swanson L, Van Epps J, Iacob E, Supiano K. We are not heroes: Examining nurses' experiences of moral distress during COVID-19. Teach Learn Nurs. 2025;20(2):e368\u0026ndash;75. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.teln.2024.11.021\u003c/span\u003e\u003cspan address=\"10.1016/j.teln.2024.11.021\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eAwad NHA, Ashour HMA, Abouelala RASA, Yaseen MM, Awad WHA. When distress meets technology: The mediating role of AI integration in the link between nurses' moral distress and moral integrity in critical care settings. Appl Nurs Res. 2026;88:152065. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.apnr.2026.152065\u003c/span\u003e\u003cspan address=\"10.1016/j.apnr.2026.152065\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBarr P. Mediation effects of burnout measure dimensions on the relationship between moral distress and having considered leaving: A regression-based correlational study in neonatal intensive care unit nurses. Australian Crit Care. 2025;38(6):101446. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.aucc.2025.101446\u003c/span\u003e\u003cspan address=\"10.1016/j.aucc.2025.101446\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eB\u0026eacute;k\u0026eacute;s V, Doorn KAV. The most vulnerable are prone to use AI therapists: The role of attachment, epistemic trust, and mental health symptoms in acceptance of digital mental health interventions. Psychother Res. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10503307.2026.2615388\u003c/span\u003e\u003cspan address=\"10.1080/10503307.2026.2615388\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eBoyd RL, Markowitz DM. Artificial intelligence and the psychology of human connection. Perspect Psychol Sci. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/17456916251404394\u003c/span\u003e\u003cspan address=\"10.1177/17456916251404394\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen QH, Tawilapakul U, Lin AMY. Teacherness in the age of GenAI: An ethnographic exploration of pedagogical decision-making and AI integration in english language teaching. System. 2026;136:103877. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.system.2025.103877\u003c/span\u003e\u003cspan address=\"10.1016/j.system.2025.103877\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eChen XX, Pi XJ. Assessing EFL students' GenAI-assisted writing: Teachers' pains, perceptions and practices. Assess Writ. 2026;67:100996. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.asw.2025.100996\u003c/span\u003e\u003cspan address=\"10.1016/j.asw.2025.100996\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eErbay-\u0026Ccedil;etinkaya S. A triadic support framework for reflective teacher identity: Peer, AI, and faculty collaboration within a community of practice. Teach Teacher Educ. 2025;168:105239. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.tate.2025.105239\u003c/span\u003e\u003cspan address=\"10.1016/j.tate.2025.105239\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFatalaki JA, Saeedi Z, Marand ES, Firuz ARN. Artificial intelligence in education and EFL teachers' attitudes: Scale development and validation. Innov Lang Learn Teach. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/17501229.2025.2494823\u003c/span\u003e\u003cspan address=\"10.1080/17501229.2025.2494823\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFornell C, Larcker DF. Evaluating structural equation models with unobservable variables and measurement error. J Mark Res. 1981;18(1):39\u0026ndash;50. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/002224378101800104\u003c/span\u003e\u003cspan address=\"10.1177/002224378101800104\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eFrush BW, Gaffney S. Scarcity, futility, and moral distress: A case analysis of the challenges of sound ethical description. J Pain Symptom Manag. 2025;70(5). \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jpainsymman.2025.07.025\u003c/span\u003e\u003cspan address=\"10.1016/j.jpainsymman.2025.07.025\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHair JF, Risher JJ, Sarstedt M, Ringle CM. When to use and how to report the results of PLS-SEM. Eur Bus Rev. 2019;31(1):2\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/EBR-11-2018-0203\u003c/span\u003e\u003cspan address=\"10.1108/EBR-11-2018-0203\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHayes AF. Introduction to mediation, moderation, and conditional process analysis. 2nd ed. Guilford Press; 2018.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHobfoll SE. Conservation of resources: A new attempt at conceptualizing stress. Am Psychol. 1989;44(3):513\u0026ndash;24. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/0003-066X.44.3.513\u003c/span\u003e\u003cspan address=\"10.1037/0003-066X.44.3.513\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHoang NH. Cultural-linguistic congruence as a moderator of teacher emotions: Examining AI implementation in Vietnamese EFL classrooms. RELC J. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1177/00336882251367785\u003c/span\u003e\u003cspan address=\"10.1177/00336882251367785\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu HZ, Zhou Q, Hashim H. Negotiating identity in the age of ChatGPT: Non-native english researchers' experiences with AI-assisted academic writing. Humanit Social Sci Commun. 2025;12(1):965. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1057/s41599-025-05351-4\u003c/span\u003e\u003cspan address=\"10.1057/s41599-025-05351-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eHu L, Bentler PM. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct Equ Model. 1999;6(1):1\u0026ndash;55. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10705519909540118\u003c/span\u003e\u003cspan address=\"10.1080/10705519909540118\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eKline RB. Principles and practice of structural equation modeling. 4th ed. Guilford Press; 2016.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eK\u0026uuml;\u0026ccedil;\u0026uuml;kuncular A, Ertugan A. Teaching in today's world: A Marxian critique of AI in education. J Acad Ethics. 2025;24(1):18. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10805-025-09692-2\u003c/span\u003e\u003cspan address=\"10.1007/s10805-025-09692-2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLee S, Jeon J, McKinley J, Rose H. Generative AI and english language teaching: A global Englishes perspective. Annu Rev Appl Linguist. 2025;45:85\u0026ndash;108. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1017/S0267190525100184\u003c/span\u003e\u003cspan address=\"10.1017/S0267190525100184\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLi BX, Bhattarai A, Ding ZQ. In search of humanness: Professional identities of qualitative research educators in the age of generative AI. Learn Media Technol. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/17439884.2025.2521547\u003c/span\u003e\u003cspan address=\"10.1080/17439884.2025.2521547\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eLipschuetz M, Topaz M. (2026). Prioritizing nurse-patient relationships in the digital transformation of nursing: A discussion paper. \u003cem\u003eInternational Journal of Nursing Studies\u003c/em\u003e, \u003cem\u003e177\u003c/em\u003e, 105369. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijnurstu.2026.105369\u003c/span\u003e\u003cspan address=\"10.1016/j.ijnurstu.2026.105369\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMcCarthy DR. A journey to trust for AI: Civil-military relations and epistemic authority in American socio-technical imaginaries of artificial intelligence. Contemp Secur Policy. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/13523260.2025.2607593\u003c/span\u003e\u003cspan address=\"10.1080/13523260.2025.2607593\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMutlu AK. Unveiling Turkish EFL teachers' generative artificial intelligence (GenAI) literacy: Voices from low-resource contexts. J Educ Students Placed Risk. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/10824669.2025.2579972\u003c/span\u003e\u003cspan address=\"10.1080/10824669.2025.2579972\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eMuth\u0026eacute;n LK, Muth\u0026eacute;n BO. (1998\u0026ndash;2017). \u003cem\u003eMplus user's guide\u003c/em\u003e (8th ed.). Muth\u0026eacute;n \u0026amp; Muth\u0026eacute;n.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePandey CS, Mishra P, Pandey SR, Pandey S. Epistemic trust in generative AI for higher education scale (ETGAI-HE scale). AI Soc. 2026;41(2):1387\u0026ndash;400. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s00146-025-02566-6\u003c/span\u003e\u003cspan address=\"10.1007/s00146-025-02566-6\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePandey SR, Pandey CS, Mishra P, Pandey S, Kumar M. Mapping the evolution of epistemic trust in artificial intelligence: A bibliometric analysis of key themes, influences, and global trends. Aslib J Inform Manage. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/AJIM-11-2024-0898\u003c/span\u003e\u003cspan address=\"10.1108/AJIM-11-2024-0898\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePham TT, Le TT. Negotiating reflection and ethics in co-reflection with AI: A narrative inquiry of Vietnamese EFL teachers' experiences. Reflective Pract. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/14623943.2026.2631417\u003c/span\u003e\u003cspan address=\"10.1080/14623943.2026.2631417\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003ePodsakoff PM, MacKenzie SB, Lee JY, Podsakoff NP. Common method biases in behavioral research: A critical review of the literature and recommended remedies. J Appl Psychol. 2003;88(5):879\u0026ndash;903. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1037/0021-9010.88.5.879\u003c/span\u003e\u003cspan address=\"10.1037/0021-9010.88.5.879\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eRoe J, Perkins M, Somoray K, Miller D, Furze L. Can synthetic avatars replace lecturers? An exploratory international study of higher education stakeholder perceptions. Int J Educational Technol High Educ. 2025;22(1):71. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41239-025-00568-4\u003c/span\u003e\u003cspan address=\"10.1186/s41239-025-00568-4\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSacco F, Manzi F, Di Dio C, Schepman A, Rodway P, Massaro D, Marchetti A. Validation of the general attitudes towards artificial intelligence scale in the Italian context and the role of trust. Computers Hum Behav Rep. 2025;19:100751. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.chbr.2025.100751\u003c/span\u003e\u003cspan address=\"10.1016/j.chbr.2025.100751\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eS\u0026aacute;nchez-Trujillo MD, Bernab\u0026eacute;-S\u0026aacute;nchez E, S\u0026aacute;enz-Eg\u0026uacute;squiza FD. (2025). Artificial intelligence and teacher training: Perspectives of education students. \u003cem\u003eEduweb \u0026ndash; Revista de Tecnolog\u0026iacute;a de Informaci\u0026oacute;n y Comunicaci\u0026oacute;n en Educaci\u0026oacute;n\u003c/em\u003e, \u003cem\u003e19\u003c/em\u003e(3), 22\u0026ndash;34. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.46502/issn.1856-7576/2025.19.03.2\u003c/span\u003e\u003cspan address=\"10.46502/issn.1856-7576/2025.19.03.2\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSawyer K, Dave VS, Gonyeau MJ, Cain J. Teaching tomorrow's pharmacists in an AI world: Risk, responsibility, and reflection. Am J Pharm Educ. 2025;89(12):101905. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ajpe.2025.101905\u003c/span\u003e\u003cspan address=\"10.1016/j.ajpe.2025.101905\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSedlakova J, Lucivero F, Pavarini G, Kerasidou A. Human-like epistemic trust? A conceptual and normative analysis of conversational AI in mental healthcare. Am J Bioeth. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/15265161.2025.2526734\u003c/span\u003e\u003cspan address=\"10.1080/15265161.2025.2526734\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSharma K, Rai P. Integrating ChatGPT into online teaching: An interpretative phenomenological study of Indian higher education educators. Am J Distance Educ. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1080/08923647.2026.2627737\u003c/span\u003e\u003cspan address=\"10.1080/08923647.2026.2627737\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eSu Y, Yu ZY, Chen S, Chen J, Wang L. Exploring digital resilience and its related factors among nursing students in the era of AI. Nurse Educ Pract. 2026;90:104636. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.nepr.2025.104636\u003c/span\u003e\u003cspan address=\"10.1016/j.nepr.2025.104636\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTabata-Kelly M, Bulger AL, Reich AJ, Gazarian PK, Izumi S, Gray TF, Wichmann L, Rinaldi S, Bernacki RE. Nurses' experiences of serious illness conversations with marginalized patients in COVID-19. J Pain Symptom Manag. 2026;71(1):24\u0026ndash;42. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.jpainsymman.2025.09.001\u003c/span\u003e\u003cspan address=\"10.1016/j.jpainsymman.2025.09.001\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eTrueblood AE, Vogelsmeier A, Scott SD, Rantz M. Psychological distress of nurse managers in acute care: A scoping review. Nurs Outlook. 2025;73(4):102447. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.outlook.2025.102447\u003c/span\u003e\u003cspan address=\"10.1016/j.outlook.2025.102447\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eVassallo D. The AI guilt complex: Moral emotions and ethical dilemmas in academic technology adoption. J Acad Ethics. 2026;24(1):53. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10805-026-09726-3\u003c/span\u003e\u003cspan address=\"10.1007/s10805-026-09726-3\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWatts KJ. Paying the cognitive debt: An experiential learning framework for integrating AI in social work education. Educ Sci. 2025;15(10):1304. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.3390/educsci15101304\u003c/span\u003e\u003cspan address=\"10.3390/educsci15101304\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eWoo BFY, Cato K, Cho H, You B, S., Song J. The use of large language models in clinical documentation: A scoping review. Int J Nurs Stud. 2026;176:105322. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.ijnurstu.2025.105322\u003c/span\u003e\u003cspan address=\"10.1016/j.ijnurstu.2025.105322\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eXia Q, Yang YM, Weng XJ, Cilsalar-Sagnak H, Cheng WK, Chiu TKF. Generative artificial intelligence competency self-efficacy scale for teachers and researchers (GAICS-TR) in higher education. J Computers Educ. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s40692-025-00373-y\u003c/span\u003e\u003cspan address=\"10.1007/s40692-025-00373-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYang RH. Between epistemic empowerment and moral anxiety: Chinese patients' ambivalence toward AI-assisted diagnosis. Soc Sci Med. 2026;391:118929. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1016/j.socscimed.2026.118929\u003c/span\u003e\u003cspan address=\"10.1016/j.socscimed.2026.118929\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eYirmiya K, Fonagy P. Mentalizing without a mind: Psychotherapeutic potential of generative AI. J Med Internet Res. 2025;27:e79156. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.2196/79156\u003c/span\u003e\u003cspan address=\"10.2196/79156\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZainuddin Z, Churiyah M, Rahayu WP, Pratikto H, Halili SH. E-portfolios as authentic assessment in pre-service teacher education: Enhancing reflection, professional identity, and digital preparedness. Educ Inform Technol. 2026. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1007/s10639-025-13869-y\u003c/span\u003e\u003cspan address=\"10.1007/s10639-025-13869-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhang H, Wang Q. How could GenAI work on in-service teachers' knowledge building process? An empirical study based on epistemic network analysis. Int J Educational Technol High Educ. 2025;22(1):47. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1186/s41239-025-00544-y\u003c/span\u003e\u003cspan address=\"10.1186/s41239-025-00544-y\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e \u003cli\u003e\u003cspan\u003eZhao ZJ, Li JJ, Hong YJ, Yun TY. Lesson study as an approach to facilitate the integration of Gen-AI into EFL curriculum design in higher education. Int J Lesson Learn Stud. 2025. \u003cspan class=\"ExternalRef\"\u003e\u003cspan class=\"RefSource\"\u003ehttps://doi.org/10.1108/IJLLS-01-2025-0029\u003c/span\u003e\u003cspan address=\"10.1108/IJLLS-01-2025-0029\" targettype=\"DOI\" class=\"RefTarget\"\u003e\u003c/span\u003e\u003c/span\u003e.\u003c/span\u003e\u003c/li\u003e\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":"epistemic trust, generative AI, professional judgment reshaping, moral stress, teacher professional development, EFL, China, structural equation modeling","lastPublishedDoi":"10.21203/rs.3.rs-9071480/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-9071480/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"\u003cp\u003eAs generative artificial intelligence (GenAI) becomes deeply embedded in university language teaching contexts, EFL teachers are increasingly confronted with a novel epistemic challenge: whether and how to trust AI-generated knowledge in their professional practice. This study examines the psychological and professional consequences of epistemic trust in GenAI (ETGAI) among 420 university EFL teachers in China, proposing a serial mediation model in which professional judgment reshaping (PJR) and moral stress (MS) sequentially transmit the effects of ETGAI on teacher professional development (TPD). Drawing on epistemic trust theory, Conservation of Resources (COR) theory, and professional agency frameworks, a covariance-based structural equation model (CBSEM) was estimated using Mplus 8.3 with 5,000 bootstrap replications. Results supported all seven hypotheses: ETGAI positively predicted PJR (β\u0026thinsp;=\u0026thinsp;.476) and TPD (β\u0026thinsp;=\u0026thinsp;.346), PJR positively predicted MS (β\u0026thinsp;=\u0026thinsp;.490), and MS negatively predicted TPD (β = \u0026minus;.405). The serial indirect effect (ETGAI \u0026rarr; PJR \u0026rarr; MS \u0026rarr; TPD) was significant (β = \u0026minus;.095, 95% BC CI [\u0026minus;\u0026thinsp;.133, \u0026minus;\u0026thinsp;.061]), revealing a moral stress pathway through which epistemic trust in GenAI ultimately undermines professional development when mediated by judgment reshaping. Theoretical and practical implications for AI-integrated EFL teacher education are discussed.\u003c/p\u003e","manuscriptTitle":"Epistemic Trust in Generative AI and the Reshaping of Professional Judgment:","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-04-17 20:13:50","doi":"10.21203/rs.3.rs-9071480/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":"72061ff6-6594-42ed-a585-76847be20c29","owner":[],"postedDate":"April 17th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[],"tags":[],"updatedAt":"2026-04-22T08:27:37+00:00","versionOfRecord":[],"versionCreatedAt":"2026-04-17 20:13:50","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-9071480","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-9071480","identity":"rs-9071480","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}