Development and Validation of a Scale to Assess Trust in AI-Assisted English Writing Among Chinese University Students

Development and Validation of a Scale to Assess Trust in AI-Assisted English Writing Among Chinese University Students | Research Square window.SnipcartSettings = { analytics: { enabled: false } }; (function() { var accessVector = localStorage.getItem('access_vector') || ''; window.dataLayer = window.dataLayer || []; if (accessVector) { window.dataLayer.push({ user: { profile: { profileInfo: { snid: accessVector } } } }); } })(); (function(w,d,s,l,i){w[l]=w[l]||[];w[l].push({'gtm.start':new Date().getTime(),event:'gtm.js'});var f=d.getElementsByTagName(s)[0],j=d.createElement(s),dl=l!='dataLayer'?'&l='+l:'';j.async=true;j.src='https://www.googletagmanager.com/gtm.js?id='+i+dl;f.parentNode.insertBefore(j,f);})(window,document,'script','dataLayer','GTM-K279D39R'); Browse Preprints In Review Journals COVID-19 Preprints AJE Video Bytes Research Tools Research Promotion AJE Professional Editing AJE Rubriq About Preprint Platform In Review Editorial Policies Our Team Advisory Board Help Center Sign In Submit a Preprint Cite Share Download PDF Article Development and Validation of a Scale to Assess Trust in AI-Assisted English Writing Among Chinese University Students Ning Mi, Cunying Fan This is a preprint; it has not been peer reviewed by a journal. https://doi.org/ 10.21203/rs.3.rs-7939708/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 The purpose of this study is to develop and validate the Trust Scale for AI-assisted English writing targeted at Chinese university students (TS-AIEW). In accordance with well-established scientific protocols, we developed the preliminary version of the TS-AIEW and validated its performance by leveraging date derived from 527 valid survey questionnaires completed by Chinese university students. This dataset was split into two distinct subsamples: 257 responses were designated for the implementation of exploratory factor analysis (EFA), while the remaining 270 were allocated to conduct confirmatory factor analysis (CFA). During the EFA and CFA procedures, a total of 8 items were eliminated on account of low item-total correlation coefficients, values of corrected item-total correlations and factor loadings. The finalized TS-AIEW, consisting of 20 items, comprises four distinct dimensions: goal alignment, transparency, competency, and trust intention. For future research, scholars may opt to directly adopt this validated TS-AIEW or assess its applicability and validity across varied demographic groups and research contexts. Business and commerce/Information systems and information technology Biological sciences/Psychology Social science/Psychology trust AI-assisted English writing Chinese University students reliability validity scale development Figures Figure 1 Introduction The swift advancement of generative artificial intelligence (AI) has led to the increasing incorporation of AI-based writing tools into higher education, particularly to support English as a foreign language (EFL) learners in their writing tasks. These tools offer immediate feedback on aspects such as language use, text structure, and overall coherence 1,2 . Nevertheless, the mere availability of advanced technology does not guarantee improved learning outcomes. Users’ willingness to apply AI-generated suggestions is often contingent upon the degree of trust they place in the system 3-5 . Therefore, trust plays a pivotal role in mediating the link between technological capability and actual learning engagement. In the fields of behavioral science and organizational management, trust has often been described as a psychological state in which an individual is willing to accept vulnerability under conditions of uncertainty or potential risk 6 . This formulation has been widely adopted because it emphasizes the evaluative judgments people make about both ability and goodwill. Importantly, the same conceptualization has been extended beyond interpersonal settings to the study of human-system relationships in technology use 3 . A growing body of work suggests that users’ trust in automated systems substantially shapes their reliance on system outputs 7,8 . However, the implications are twofold: excessive trust may lead to uncritical acceptance, whereas insufficient trust can result in avoidance or underutilization. In both scenarios, the potential benefits of the technology are undermined to some extent. Recent conceptual advances have, to some extent, sparked growing scholarly attention to the question of how trust can and should be measured across different domains. Several influential frameworks illustrate this diversity, including the Muenster Epistemic Trustworthiness Inventory (METI) 9 , Model of Organizational Trust 6 , the Human-Computer Trust (HCT) model 4 , and the Multi-Dimensional Measure of Trust (MDMT) 10 . Although these instruments demonstrate a relative level of maturity, they remain, by and large, context-general rather than tailored to specific tasks. As a result, they do not fully reflect distinctive features of writing situations, including genre conventions, rhetorical tone, text organization, or the provision of explanatory feedback. Such omissions make it difficult to capture the nuanced ways in which EFL learners decide whether, and to what extent, to rely on AI-generated suggestions in authentic writing tasks. Against this backdrop, the present study targets EFL learners in Chinese universities with the goal of developing and verifying task-specific and dimensionally explicit scale for AI-assissted writing trust. By doing so, the study seeks to contribute not only to the improved design of AI writing tools but also to the enhancement of students’ AI literacy and to future research on the dynamics of AI-human collaborative learning. Methods The definition of trust put forth by Mayer, Davis, and Schoorman, centering on competence, benevolence, and integrity as the basis for one party's willingness to accept vulnerability, has been highly influential in organizational behavior 6 . The three qualities have been incorporated as central dimensions in a wide range of theoretical frameworks across disciplines. In this sense, trust is inherently relational and carries normative implications, shaping practical choices such as delegation, collaboration, and risk-taking within institutional settings. As the study of trust moved into technology-mediated contexts, conceptual refinements became necessary. Within human-technology interaction, trust has been reinterpreted more in terms of perceptions of a system’s reliability, functionality, and controllability 11 . This adjustment reflects users’ tendency to evaluate technical performance and output quality in addition to, or even instead of, presumed intentions. In the specific case of automation, Lee and See introduced the notion of calibrated trust, stressing that trust should evolve in line with system performance rather than being treated as a simple binary state 3 . Their analysis pointed to the risks of both excessive trust, leading to blind reliance, and insufficient trust, resulting in unwarranted avoidance. Building on these insights, Hoff and Bashir distinguished between performance-oriented cues including accuracy and reliability, and process-oriented cues including transparency and predictability, noting that both categories play a role in shaping users’ trust assessments 4 . Importantly, trust is shaped not only by perceptions of system accuracy but also by the intelligibility of its underlying decision processes. The emergence of artificial intelligence, particularly in generative and adaptive forms, adds an additional layer of complexity. As Glikson and Woolley observe, AI systems diverge from traditional automation in that they continually learn and adapt, often through opaque “black box” mechanisms 12 . In such contexts, user trust depends not only on perceived performance but also on factors, such as the ability to clarify how the system generates decisions or results, the degree to which the system’s objectives match users’ needs or values. Taken together, these developments indicate a paradigmatic transition from a mechanistic understanding of trust to one that also incorporates cognitive and alignment considerations in human-AI interaction. Trust constructs, identified in prior measurement tools that examine comparable areas, range from interpersonal trust to trust in e-commerce, information systems, technology adoption, and expert advice. With the growing role of AI in educational contexts, trust-related scales have been explored specifically designed for learning and instruction 13,14 . Finally, ten relevant instruments were presented (see Table 1). Table 1. Ten related scales Scale Abbreviation Reference Trust in Automated Systems Scale TAS Jian et al., 2000 Trust of Automated Systems Test TOAST Wojton et al., 2020 Multi-Dimensional Measure of Trust (v2) MDMT v2 Ullman & Malle, 2023 e-Commerce Trust Scale ECTS McKnight et al., 2002 Trust in Autonomous Vehicles Scale TAVS Choi & Ji, 2015 Laypeople’s Trust in Experts Scale / Muenster Epistemic Trustworthiness Inventory LTE / METI Hendriks et al., 2015 Teachers’ Trust in AI-Based Educational Technology Instrument TTAI-EdTech Nazaretsky et al., 2022 Trust in AI-Powered Educational Technology Scale TAI-EdTech Nazaretsky et al., 2025 Human-AI interaction HAII Shin, 2021 Human-like trust in AI and functionality trust in AI TAI Choung et al., 2022 Following a comprehensive review of these scales listed, the constructs of assessing trust in different situations were extracted and compared. Those deemed irrelevant to the research context were excluded, while overlapping categories were consolidated. This iterative refinement ultimately produced five central constructs: competency, integrity, transparency, trust intention and benevolence. Competency, as a trust construct, reflects the extent to which users believe a system or agent possesses the necessary ability to carry out its intended functions successfully and to satisfy user expectations 15 . Transparency, within the context of autonomous and intelligent systems, is commonly understood as the extent to which users are able to anticipate and interpret how the system operates 7 . Integrity was defined as a key component that reflects an individual’s adherence to moral and ethical principles, directly influencing the perception of their trustworthiness 16 . Benevolence, as a dimension of trust, has been articulated in prior scales through several complementary aspects. For instance, the Multi-Dimensional Measure of Trust (MDMT v2) emphasizes goodwill and considerateness 10 , the e-Commerce Trust Scale (ECTS) highlights ethical behavior and moral concern 11 . Trust intention, as identified in prior measurement frameworks, generally denotes an individual’s willingness or expressed likelihood to rely on a trustee. This includes both the readiness to depend and the subjective probability of doing so 11 . Results Item Analysis To evaluate the quality of the initial items, data from Sample 1 were analyzed using three procedures: independent-samples t-tests based on the upper and lower 27% groups, item-total correlation analysis, and internal consistency reliability. First, independent-samples t-tests were conducted by selecting the top 27% of respondents (high-score group) and the bottom 27% (low-score group) according to the total scores 17 . Results showed that all 28 items significantly differentiated between the two groups, indicating satisfactory item discrimination. Second, item-total correlation coefficients were computed to examine the degree to which each item was consistent with the overall construct. While correlations above .30 are generally considered acceptable 18 , values greater than .40 are recommended as a more stringent standard 19,20 . Results indicated that all items were significantly correlated with the total score (p < .001); however, Item Q28 showed a coefficient of only .307, failing to meet the .40 criterion. Finally, internal consistency reliability was assessed using Cronbach’s α 21 . The 28-item scale yielded an overall α of .948. Examination of the corrected item-total correlations (CITC) and the “α if item deleted” statistics suggested that three items, Q5 (CITC = .418, α = .949), Q23 (CITC = .454, α = .948), and Q28 (CITC = .247, α = .951), weakened the overall consistency of the scale. Considering both the low CITC values and the improvement of α when these items were removed, the deletion of Q5, Q23, and Q28 was deemed appropriate. The remaining 25 items demonstrated satisfactory psychometric properties and were retained for subsequent exploratory factor analysis. Exploratory Factor Analysis We conducted an exploratory factor analysis on the remaining 25 items using Sample 1 data. It was found that the rotation converged after seven iterations using principal component analysis and Kaiser normalization optimal oblique rotation. Statistical results showed that the KMO value was 0.945, exceeding the threshold of 0.7. Bartlett’s sphericity test yielded a significant result (approximate chi-square value of 4282.441, degrees of freedom 300, p < 0.001), indicating that the sample data were suitable for factor analysis. Based on the rotated factor matrix, factors with eigenvalues greater than 1 were extracted. In the first round of exploratory factor analysis, the factor loadings for items Q6, Q13, and Q22 were all less than 0.5, failing to meet the factor extraction criteria (Hair et al., 2014, set at 0.4). After multiple rounds of exploratory factor analysis, five items (Q6, Q13, Q14, Q22, and Q24) were removed, resulting in a four-factor structure comprising 20 items. The common factor variance extraction values for all items ranged from 0.563 to 0.839. The final structure had a KMO value of 0.936, exceeding 0.8, and Bartlett’s sphericity test was significant (approximate chi-square 3321.382, degrees of freedom 190, p < 0.001). Principal component analysis and Kaiser normalization were used, and the optimal oblique rotation converged after six iterations, extracting four factors with eigenvalues greater than 1, explaining a cumulative variance of 69.466%. The specific loadings, factor characteristics, and variance explained for each item are shown in Table 2. Table 2. Result of exploratory factor analysis. Factor loading coefficient Communality Item Mean SD Factor 1 Factor 2 Factor 3 Factor4 Item 8 3.76 0.822 0.931 0.706 Item 9 4.00 0.713 0.876 0.677 Item 10 3.84 0.783 0.819 0.74 Item 12 3.94 0.773 0.787 0.714 Item 11 3.88 0.758 0.776 0.706 Item 7 4.08 0.651 0.578 0.653 Item 17 3.80 0.769 0.899 0.612 Item 19 3.84 0.749 0.839 0.686 Item 20 3.85 0.758 0.793 0.663 Item 18 3.91 0.707 0.623 0.662 Item 15 3.73 0.816 0.621 0.608 Item 21 3.88 0.69 0.588 0.658 Item 16 3.93 0.728 0.515 0.563 Item 3 4.14 0.674 0.946 0.821 Item 1 4.12 0.675 0.926 0.839 Item 2 4.17 0.645 0.914 0.826 Item 4 4.06 0.715 0.688 0.674 Item 25 3.89 0.667 0.912 0.727 Item 27 3.89 0.728 0.773 0.687 Item 26 3.96 0.654 0.767 0.669 Eigenvalue 9.432 2.234 1.219 1.007 Contribution Rate % 47.16 11.172 6.097 5.037 Cumulative Contribution Rate % 47.16 58.332 64.429 69.466 Note: Factor 1=Goal alignment trust; Factor 2=Transparency; Factor 3=Competency; Factor 4=Trust intention As shown in Table 2, the factor loadings for all items range from 0.515 to 0.946 and are aggregated into four factors. Based on the content of the descriptive items included in each factor, the four factors are named F1 (Goal Alignment), F2 (Transparency ), F3 (Competency), and F4 (Trust intention). Confirmatory Factor Analysis To ensure robustness, we used AMOS 26.0 to conduct confirmatory factor analysis on Sample 2 (N=270) to measure the overall fit of the factor structure. The results demonstrated that the chi-square/df ratio reached 1.855, falling within the ideal range of 1 to 3 for adequate model fit. The Comparative Fit Index (CFI) yielded a value of 0.961, while the Incremental Fit Index (IFI) also stood at 0.961. Additionally, the Tucker-Lewis Index (TLI) was 0.955, and the Normed Fit Index (NFI) reached 0.950, all of these indices surpassing the acceptable threshold of 0.90. For the Root Mean Square Error of Approximation (RMSEA), the calculated value was 0.056. Though this figure exceeds the 0.05 threshold for excellent model fit, it still falls within the acceptable range of 0.05 to 0.08. The benchmarks for evaluating model fit were established in accordance with the guidelines proposed by Hair et al. and Hu and Bentler 21,22 . Taken together, these results verify that the model exhibits a good level of fit (see Table 3 and Figure 1). Table 3. Model fit Indicator Benchmark Result Interpretation Chi-square/df Between 1 to 3 1.855 Excellent CFI Larger than 0.9 0.961 Excellent IFI Larger than 0.9 0.961 Excellent TLI Larger than 0.9 0.955 Excellent NFI Larger than 0.9 0.950 Excellent RMSEA Smaller than 0.08 0.056 Good Reliability and Validity Testing Convergent validity was evaluated using values of average variance extracted (AVE) and composite reliability (CR), with the detailed findings presented in Table 4. For reliability analysis, Cronbach’ s α coefficient for the four dimensions were all above 0.70, specifically 0.917, 0.902, 0.895, and 0.824, demonstrating strong internal consistency. Composite reliability (CR) values ranged from 0.829 to 0.918, also exceeding the 0.70 threshold, further confirming the reliability of the trust scale for AI-assisted English writing. With respect to convergent validity, the AVE for Factor 2 was 0.58, slightly below the conventional cut-off of 0.60. However, as AVE is considered a conservative estimate, prior studies suggest that if CR exceeds 0.70, convergent validity may still be deemed acceptable even when AVE is below 0.60 23 . Given that the CR for this factor was 0.906, well above the recommended threshold, the dimension can reasonably be considered to demonstrate adequate convergent validity. Table 4 . Construct validity and convergent validity Construct Item Estimate S.E. Z p Factor loading Cronbach’s alpha AVE CR F1 F1_1 1 0.764 0.917 0.650 0.918 F1_2 1.076 0.081 13.344 *** 0.775 F1_3 0.945 0.069 13.681 *** 0.791 F1_4 1.147 0.078 14.669 *** 0.839 F1_5 1.114 0.074 15.004 *** 0.855 F1_6 1.09 0.078 14.049 *** 0.809 F2 F2_1 1 0.622 0.902 0.58 0.906 F2_2 1.155 0.108 10.721 *** 0.807 F2_3 1.059 0.101 10.474 *** 0.781 F2_4 1.022 0.098 10.385 *** 0.772 F2_5 1.052 0.099 10.6 *** 0.794 F2_6 1.092 0.105 10.412 *** 0.774 F2_7 1.018 0.098 10.334 *** 0.766 F3 F3_1 1 0.897 0.895 0.704 0.904 F3_2 0.877 0.043 20.517 *** 0.883 F3_3 0.912 0.047 19.505 *** 0.86 F3_4 0.853 0.062 13.691 *** 0.701 F4 F4_1 1 0.718 0.824 0.618 0.829 F4_2 1.064 0.088 12.067 *** 0.844 F4_3 1.086 0.093 11.637 *** 0.792 Discriminant validity refers to the extent to which a construct is distinct from other constructs, typically demonstrated through low correlations between the latent traits they represent. According to the criterion 24 , discriminant validity is established when the square root of the AVE for a construct exceeds its correlations with other constructs. In this study, the results fulfilled this requirement as the square root of the AVE for each construct surpassed its inter-construct correlations, confirming that the dimensions exhibit adequate discriminant validity (see Table 5). Table 5 . AVE, the square root of AVE and the correlations between constructs. F4 F3 F2 F1 F4 0.786 F3 0.508 0.839 F2 0.653 0.541 0.762 F1 0.614 0.681 0.745 0.806 Discussion This study developed and validated a four-factor trust scale for AI-assisted English writing, encompassing Goal Alignment, Transparency, Competency, and Trust Intention. The instrument demonstrated sound psychometric qualities, with both exploratory and confirmatory factor analyses confirming its reliability and validity. These results resonate with classical trust frameworks while also extending recent empirical findings 6,11 . Firstly, introducing Goal Alignment as a distinct dimension extends existing trust models.While classical models emphasized benevolence and integrity 6 , this study conceptualizes trust in AI-assisted writing as strongly tied to whether system outputs align with students’ learning objectives.This perspective resonates with recent discussions of value and goal alignment in AI ethics and human-AI interaction. For instance, Gabriel highlighted value alignment as essential for trustworthy AI 25 , while Mechergui and Sreedharan reframed alignment in terms of human-aware goal matching 26 . Similarly, Shen et al. proposed a bidirectional perspective, emphasizing that alignment is dynamic and requires adaptation from both humans and AI 27 . Second, Transparency emerged as another critical dimension of trust. The findings suggest that students value not only accurate feedback but also insight into the logic and mechanisms that generate it. This observation is consistent with Malle and Ullman’s emphasis on explainability as central to human-AI trust and with Lee and See’s notion of calibrated reliance 10,3 . Recent empirical evidence reinforces this view which showed that higher levels of explainability significantly increased teachers’ trust in AI tools 28 . Similarly, Huang and Ball found that students with greater AI literacy developed more calibrated trust, underscoring the interdependence of transparency, literacy, and trust 29 . Collectively, these findings highlight transparency, coupled with user literacy, as essential for cultivating both confidence and critical engagement with AI in educational contexts. Third, Competency remains the cornerstone of trust. Students consistently evaluated AI tools by their capacity to generate high-quality texts, echoing Mayer et al.’s “ability” dimension and aligning with Jian et al.’s findings in automation 6,15 . Recent evidence also supports this view: Đerić et al. identified competency and transparency as the strongest predictors of student trust 30 , while Pratiwi et al. reported that Indonesian students valued AI for idea generation but questioned its limited revision and editing capabilities 31 . In both research and practice, teachers and students alike underscore effectiveness, reinforcing competency as the fundamental basis of trust. Finally, Trust Intention was confirmed as an independent dimension of the trust construct, capturing students’ willingness to rely on and continue using AI-assisted writing tools. This interpretation is consistent with McKnight et al.’ s model and parallels Choi and Ji’ s findings in the automation domain 11,7 , where intention to use was treated as a direct indicator of trust. Recent evidence further refines this perspective: Pitts et al. showed that students’ reliance behaviors were closely tied to their trust and satisfaction, but also revealed risks of overreliance when trust intention was uncalibrated 32 . These results align with Lee and See’ s principle of appropriate reliance, suggesting that trust intention should not be read as uniformly positive 3 . Pedagogically, fostering trust intention requires pairing it with transparency and literacy support to encourage calibrated rather than blind reliance. Taken together, the findings draw on classical trust models while incorporating recent evidence to clarify their relevance in educational contexts. Theoretically, it extends trust research into AI-assisted learning by underscoring the roles of goal alignment, transparency,competency and trust intention. Practically, it suggests that teachers can foster trust by aligning AI feedback with learning objectives, whereas developers should prioritize transparency and performance to encourage acceptance and sustained use. Future research should test the scale across cultural contexts and integrate behavioral and longitudinal data to enhance its ecological validity and explanatory strength. Conclusions This study developed and validated the Trust Scale of AI-Assisted English Writing (TS-AIEW) to measure Chinese university students’ trust in AI-assisted writing tools. Drawing on prior trust-related scales, four constructs were identified: Goal Alignment, Transparency, Competency, and Trust Intention. Following established scale development procedures, item analysis and reliability and validity testing were conducted, resulting in the removal of eight items 19,33 . The final instrument comprises 20 items across the four constructs, all of which demonstrated strong psychometric properties. The TS-AIEW offers a robust tool for assessing university students’ trust in AI-assisted English writing, enabling reliable evaluation and ongoing monitoring. Its main contribution lies in informing tailored instructional strategies and guiding the design of trustworthy educational technologies. Moreover, the scale provides a foundation for comparative and cross-contextual research, advancing understanding of learners’ trust dynamics in technology-enhanced education. One limitation of this study is that the Trust Scale of AI-Assisted English Writing (TS-AIEW), though developed and validated with rigorous psychometric procedures, was tested primarily with Chinese university students. This cultural specificity may restrict the generalizability of the findings. Future research should validate the TS-AIEW with more diverse cultural and linguistic groups to establish its cross-cultural applicability. The framework could also be adapted to other educational technologies by substituting the construct of “AI-assisted English writing” with domains such as “AI-assisted academic reading” or “AI-assisted translation”. In addition, pedagogically oriented research is needed to explore how trust interacts with learning outcomes, motivation, and critical thinking 34 . Finally, future studies should investigate contextual features of AI-assisted systems that may shape students’ trust calibration, such as autonomy, feedback form, and explainability 10,3 . Methods Research procedure This study follows a scale development approach consisting of several research stages, guided by the procedures outlined in DeVellis and Thorpe (2021), Luo (2024) and Luo (2023a) 19,33,35 . The first stage involved the identification of the scale’ s constructs. In this regard, constructs are described as the theoretical dimensions or latent factors that denote the phenomenon under investigation 19 . Drawing on related instruments, we integrated relevant constructs, removed those outside the scope of this study, refined their wording, and ultimately identified four dimensions: Competency, Goal Alignment, Transparency, and Trust Intention. Stage 2 focused on establishing the item pool. In scale development, items refer to the specific survey questions or statements designed to capture data about participants’ attitudes, behaviors, or perceptions. As DeVellis and Thorpe recommend, beginning with a broad pool of items helps ensure adequate coverage of the constructs 19 . Item generation can be achieved through several strategies, including adapting items from existing scales, drafting items from theoretical foundations, or converting qualitative insights into survey statements. In this study, we primarily relied on the first and the third approach, modifying items from related scales and transforming qualitative responses, to ensure alignment with the scope of AI-assisted English writing (see Table 6). The third stage involved the validation conducted by experts. The draft version of the TS-AIEW scale was reviewed by two professors with doctoral degrees and relevant research experience. They were asked to examine the scale for logical consistency, clarity of language, and alignment with the study’ s objectives and theoretical framework 19 . In addition, the experts evaluated each item in terms of relevance, necessity, and cultural adaptability. Based on their feedback, we finalized an initial version of the TS-AIEW consisting of 28 items, which could typically be completed within 5 to 8 minutes (see Table 6). Stage 4 centered on the validation of data. Following expert review, the scale of TS-AIEW was administered to participants. Surveys completed in less than 60 seconds or with identical responses across all items were excluded from analysis 36 . Using SPSS 26.0, we examined descriptive statistics for each item, calculated Cronbach’ s alpha for each construct, and tested the scale’ s suitability for factor analysis before conducting exploratory factor analysis (EFA). The process was iterative: when items were removed, the scale was re-evaluated to ensure reliability, which occasionally required further adjustments 33 . Through several rounds of refinement, we produced a validated version of the TS-AIEW scale, presented in the Findings section. Stage 5 addressed the evaluation of construct validity and the establishment of reliability for the TS-AIEW scale. Using AMOS 26.0, we conducted confirmatory factor analysis (CFA) to validate the factor structure identified by means of EFA. Model fit was evaluated with indices including the Chi-square statistic, RMSEA, and CFI, to examine the degree of consistency between the data and the hypothesized model. Factor loadings were also examined to confirm that each item loaded adequately on its intended construct, thereby strengthening evidence for construct validity. The detailed results of this analysis are reported in the Findings section. Scale development In constructing the Trust Scale for AI-assisted English Writing among Chinese university students (TS-AIEW), we systematically integrated four central dimensions that reflect both the contextualized features of AI-supported educational writing and insights derived from earlier measurement frameworks. These constructs are competency, alignment, transparency, and trust intention, which serve as the theoretical foundation of the scale (see Table 6). Firstly, we retained the construct of Competency as it is among the most widely used dimensions in trust research. Competency refers to the skills and expertise that enable effective performance 37 . In AI-assisted English writing, it can be operationalized as learners’ belief that such tools can identify problems, provide useful suggestions, and support task completion for Chinese EFL students. In the current study, the first four items were adapted from trust beliefs framework 11 . For example, the original item “LegalAdvice.com is competent and effective in providing legal advice” was adapted to “AI-assisted writing tool is competent and effective in providing writing advice.” Additional items (5-7) were derived from face-to-face depth interviews, including “All the feedback I received from the AI-assisted writing tool was correct” “I can rely on the advice given by AI-assisted writing tools” and “The suggestions provided by AI-assisted writing tools are suitable for learners of different English proficiency levels”. Secondly, we kept the significance of benevolence in trust measurement but reframed it as Goal Alignment to better fit the context of AI-assisted English writing. Benevolence has traditionally been defined as responsiveness and goodwill toward others 11 , and recent work highlights its importance in shaping user trust in AI-driven interactions 38 . In our study, this notion is extended to capture the extent to which AI writing tools align with learners’ educational goals and act in their best interest. Operationally, goal alignment reflects students’ perception that the system provides feedback supportive of their learning objectives and demonstrates care and responsibility for their academic development. For example, the item “I found the AI-assisted feedback are a great match to my English writing needs” was adapted from Shin to fit the current context of EFL writing 39 . Our study excluded the construct of Integrity because it primarily emphasizes moral and ethical aspects of AI behavior, such as honesty and fairness 40 . In the present context, the central concern lies in how effectively AI tools support writing tasks, rather than in the more subjective assessment of ethical standards 41 . For users of educational AI systems, sustained trust is more likely to stem from performance and goal alignment than from perceptions of integrity 42,43 . Accordingly, the integrity dimension was omitted from the final framework. Thirdly, we retained Transparency in the scale, following prior research that has emphasized its central role in shaping trust 44 . In AI systems, transparency refers to the clarity with which system operations are communicated, including the algorithms applied, the data processed, and the reasoning behind outputs. Empirical work has demonstrated its relevance in educational contexts, showing that higher transparency strengthens user trust in AI-driven learning tools 45 . Shin further conceptualized transparency through three interrelated aspects: understandability, explainability, and visibility 39 . In this study, transparency is operationalized as students’ perception that AI writing tools provide feedback in a manner that is understandable, interpretable, and predictable. Items developed for this dimension therefore capture whether learners consider the system’ s suggestions sufficiently clear and well justified to be trusted in authentic writing tasks. For example, “I have a general understanding of how AI generates writing feedback.” 39 Finally, the term “Trust Intention” has been retained as is. This construct encapsulates the willingness of users to engage with AI systems based on their perceptions of competence, alignment, and transparency. This willingness is essential for encouraging users to engage with technology, underlining the significance of trust intention as a predictor of actual use 46 . In the present study, trust intention is operationalized as students’ willingness to continue using AI writing tools for their English writing tasks, based on their perception of the competency, goal alignment and transparency of AI-assisted writing tool. In accordance with the prior related survey questions, we made modifications to fit the theme of AI-assisted English writing. For instance, the statement from McKnight et al. “I feel that I could count on LegalAdvice.com to help with a crucial legal problem” was revised to “AI writing tools can improve the quality of my writing.” 11 In conclusion, the current research has identified four constructs linked to trust in AI-facilitated English writing among university students in China: competence, goal alignment, transparency, and trust intention. With reference to the phrasing of measuring items from existing related scales, we developed the initial version of the TS-AIEW scale, and its details are provided in Table 6. Table 6. Constructs and measuring items of the initial version of TS-AIEW (28 items). Constructs Measuring items Literature sources Competency Item 1. AI-assisted writing tool is competent and effective in providing writing advice. McKnight et al. (2002); Wojton et al. (2019) Choung et al. (2022); Item 2. AI-assisted writing tool performs its role of giving writing advice very well. Item 3. Overall, AI-assisted writing tool is a capable and proficient writing advice provider. Item 4. In general, AI-assisted writing tool is very knowledgeable about English writing. Item 5*. The suggestions provided by AI-assisted writing tool are suitable for learners of different English proficiency levels. Item 6*. All the feedback I received from the AI-assisted writing tool was correct. Item 7*. I can rely on the advice given by AI-assisted writing tool. Goal Alignment Item 8. AI feedback is basically consistent with what I want to express in my English writing. Nazaretsky et al. (2025); Choung et al. (2022); Shin (2021) Item 9. I found AI feedback is a great match to my English writing needs. Item 10. I feel that AI-assisted writing tool understands my writing needs. Item 11. AI-assisted writing tool helps me achieve the desired writing effect. Item 12*. AI-assisted writing tool helps me maintain clarity of my views throughout the writing process. Item 13*. AI suggestions can help me maintain a clear structure and train of thought in my writing. Item 14*. AI-assisted writing tool helps me focus on my writing goals. Transparency Item 15. I have a general understanding of how AI generates writing feedback. Shin (2021); Nazaretsky et al. (2025); Wojton et al. (2019) Item 16. AI-assisted writing tool can clearly explain the reasons behind their suggestions. Item 17. I can understand the reasoning behind a particular suggestion made by AI-assisted writing tool. Item 18. I find the presentation of AI feedback clear and understandable. Item 19*. AI-assisted writing tool provides clear and understandable explanations for language issues (such as grammar and word choice). Item 20*. AI-assisted writing tool helps me distinguish between correct and incorrect English expressions. Item 21*. The explanations provided by AI-assisted writing tool help me understand English writing rules. Trust Intention Item 22. The suggestions provided by AI-assisted writing tool influence how I revise my writing. McKnight et al. (2002); Nazaretsky et al. (2025) Choung et al. (2022); Shin (2021) Item 23*. I believe that AI-assisted writing tool can improve the quality of my writing. Item 24*. I believe that the suggestions provided by AI-assisted writing tool can improve my English writing skills. Item 25. I am willing to use AI-assisted tool as an aid in the process of English writing. Item 26. When completing English writing tasks, I tend to actively use AI-assisted writing tool. Item 27. Even if I occasionally find that the suggestions of AI-assisted writing tool are incorrect, I will still continue to use the tool. Item 28*. The suggestions provided by AI-assisted writing tool influence how I revise my writing. Note: Items marked with an asterisk (*) have been added to the scale after the face-to-face in-depth interview; participants were requested to finish the survey in accordance with their experiences over the past one semester. Questionnaire Design and Data Collection Although the TS-AIEW scale was developed through the procedures outlined above, it had to be adapted into a questionnaire format for administration. A scale is designed to measure a specific construct, whereas a questionnaire is a broader instrument that may include multiple constructs as well as background information. In this study, all items were measured on a 5-point Likert scale, ranging from 1 (“strongly disagree”) to 5 (“strongly agree”). In addition to the TS-AIEW items, questions on demographic variables (e.g., gender, grade) were included. Given that the study was conducted in China, all measurement items were translated into Chinese. To safeguard linguistic precision, the back-translation method was adopted: items were first translated from English to Chinese, and then a different translator independently translated them back into English. The original and its corresponding back-translated versions were compared to ensure semantic equivalence. Data were collected through an online survey platform (www.wjx.cn). The link was distributed mainly via QQ and WeChat among university students who were invited to participate voluntarily. The collection period lasted from late June to mid-July 2025. Before the main survey, a one-week pretest was conducted with twenty graduate students familiar with AI-assisted English writing to evaluate clarity and usability. Participants and Ethical Considerations Ethical approval for this study was obtained from the authors’ institutional review board prior to data collection. Convenience sampling was employed, yielding 728 completed questionnaires. After excluding 201 invalid cases due to insufficient response time or repetitive answering patterns, 527 valid responses were retained, resulting in an effective response rate of 72.4%. Data analysis was conducted using SPSS 27.0 and AMOS 26.0. To avoid the problem of capitalizing on chance by using the same dataset for both exploratory and confirmatory factor analyses, we randomly split the full sample into two approximately equal subsamples. Sample One (≈50%, N=257) was used for item analysis and EFA, while Sample Two (≈50%, N=270) was reserved for CFA. This approach has been recommended in methodological literature, which emphasizes that EFA and CFA should ideally be conducted on independent datasets and that one practical solution is to divide a large sample randomly into two subsamples of roughly equal size 47 . Comparisons indicated that the two subsamples were broadly similar in demographic variables such as gender, English proficiency, college, and grade, with no significant differences (see Table 7). All procedures adhered to established ethical principles. Participants were informed of the study objectives, assured of their rights and confidentiality, and reminded that participation was entirely voluntary. Anonymity was maintained throughout, and the research protocol successfully passed a rigorous ethical review process. Table 7. Demographic characteristics of participants for Sample 1 and Sample 2 (%) Sample 1 Sample 2 Category Subcategory N Percentage N Percentage Gender Male 85 33.1 73 27.0 Female 172 66.9 197 73.0 Major Humanities 99 38.5 127 47.0 Science 71 27.6 68 25.2 Engineering 87 33.9 75 27.8 Academic Year Freshman 106 41.2 128 47.4 Sophomore 141 54.9 121 44.8 Junior 9 3.5 19 7.0 Senior 1 0.4 2 0.7 English Proficiency Not Passed CET-4 140 54.5 162 60.0 Passed CET-4 102 39.7 76 28.1 Passed CET-6 15 5.8 32 11.9 Data Analysis, Reliability, and Validity During the phase of data analysis, we adhered to established guidelines 19,33,36 . Internal reliability was first evaluated for the full scale and each construct, with Cronbach’s alpha coefficients required to exceed the threshold of 0.70. Sampling adequacy was then assessed using the Kaiser-Meyer-Olkin (KMO) statistic, with values above 0.60 considered acceptable. In the subsequent step, exploratory factor analysis (EFA) was carried out through the use of principal component analysis, wherein Promax rotation and Kaiser normalization were implemented. Items were retained only when communalities were greater than 0.50 and factor loadings exceeded 0.50, with each item correctly associated with its intended construct 19,33 . For example, items designed to measure Goal Alignment were expected to load most strongly on the goal alignment factor and cluster consistently with other goal alignment items. Furthermore, average variance extracted (AVE) and composite reliability (CR) were assessed for each construct, with AVE values above 0.50 and CR values required to exceed 0.70, thereby providing evidence of convergent validity. Discriminant validity was assessed following the Fornell & Larcker criterion 48 , which requires that the square root of AVE for each construct exceed its correlations with other constructs. Subsequently, confirmatory factor analysis (CFA) was performed to test the measurement model. Model adequacy and stability were examined through key goodness-of-fit indices, including the Chi-square statistic, RMSEA, and CFI. Declarations Data availability The original contributions presented in the study are included in the article/Supplementary Materials. Further inquiries can be directed to the corresponding author. Acknowledgements Not applicable. Author contributions All authors contributed to the study conception and design. Writing - original draft preparation: Ning Mi, Cunying Fan; Writing - review and editing: Ning Mi; Conceptualization: Cunying Fan; Methodology: Ning Mi, Cunying Fan; Formal analysis and investigation: Ning Mi, Cunying Fan; Funding acquisition: Cunying Fan; Resources: Cunying Fan; Supervision: Ning Mi, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript. Funding sources This study was supported by the Shandong Social Sciences Planning Research Project of China (grant number 23CSDJ24). Competing interests The authors declare no competing interests. Ethics approval and consent to participate This study was reviewed and approved by the Institutional Review Board of Qufu Normal University (Approval No. 2025-106). All participants were informed about the purpose, procedures, and anonymity of the survey before participation. Completion of the online questionnaire was considered as provision of informed consent. The study was conducted in accordance with institutional guidelines and relevant national regulations. Additional information Supplementary Information The online version contains supplementary material available at https://doi.org/10.5281/zenodo.17432622 Correspondence and requests for materials should be addressed to N.M. Reprints and permissions information is available at www.nature.com/reprints. Publisher’ s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’ s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. References Li, J., Link, S., & Hegelheimer, V. Rethinking the Role of Automated Writing Evaluation (AWE) Feedback in ESL Writing Instruction. J. Second Lang. Writ. 27 , 1-18. https://doi.org/10.1016/j.jslw.2014.10.004 (2015). Ranalli, J. Automated written corrective feedback: how well can students make use of it? Comput. Assist. Lang. Learn. 31 , 653-674. https://doi.org/10.1080/09588221.2018.1428994 (2018). Lee, J. D., & See, K. A. Trust in automation: Designing for appropriate reliance. Hum. Factors 46 , 50-80. https://doi.org/10.1518/hfes.46.1.50_30392 (2004). Hoff, K. A., & Bashir, M. Trust in Automation: Integrating Empirical Evidence on Factors That Influence Trust. Hum. Factors 57 , 407-434. https://doi.org/10.1177/0018720814547570 (2015). Chiu, T. K. F., & Chai, C. Sustainable Curriculum Planning for Artificial Intelligence Education: A Self-Determination Theory Perspective. Sustainability 12 , 55-68. https://doi.org/10.3390/su12145568 (2020). Mayer, R. C., Davis, J. H., & Schoorman, F. D. An integrative model of organizational trust. Acad. Manag. Rev. 20 , 709-734. https://doi.org/10.5465/amr.1995.9508080335 (1995). Choi, J. K., & Ji, Y. G. Investigating the importance of trust on adopting an autonomous vehicle. Int. J. Hum.-Comput. Interact. 31 , 692-702. https://doi.org/10.1080/10447318.2015.1070549 (2015). Wojton, H. M., Porter, D., Lane, S. T., Bieber, C., & Madhavan, P. Initial validation of the trust of automated systems test (TOAST). J. Soc. Psychol. 160 , 735-750. https://doi.org/10.1080/00224545.2020.1749020 (2020). Hendriks, F., Kienhues, D., & Bromme, R. Measuring laypeople’s trust in experts in a digital age: The Muenster Epistemic Trustworthiness Inventory (METI). PLoS ONE 10 , e0139309. https://doi.org/10.1371/journal.pone.0139309 (2015). Malle, B. F., & Ullman, D. A multi-dimensional conception and measure of human-robot trust. In Nam, C. S., & Lyons, J. B. (eds.), Trust in Human-Robot Interaction: Research and Applications , 3-25. Elsevier (2021). McKnight, D. H., Choudhury, V., & Kacmar, C. Developing and validating trust measures for e-commerce: An integrative typology. Inf. Syst. Res. 13 , 334-359. https://doi.org/10.1287/isre.13.3.334.81 (2002). Glikson, E., & Woolley, A. W. Human Trust in Artificial Intelligence: Review of empirical research. Acad. Manag. Ann. 14 , 627-660 (2020). Nazaretsky, T., Ariely, M., Cukurova, M., & Alexandron, G. Teachers’ trust in AI-powered educational technology and a professional development program to improve it. Br. J. Educ. Technol. 53 , 914-931. https://doi.org/10.1111/bjet.13232 (2022a). Nazaretsky, T., Mejia-Domenzain, P., Swamy, V., Frej, J., & Käser, T. The critical role of trust in adopting AI-powered educational technology for learning: An instrument for measuring student perceptions. Comput. Educ.: Artif. Intell. 8 , 100368. https://doi.org/10.1016/j.caeai.2025.100368 (2025). Jian, J. Y., Bisantz, A. M., & Drury, C. G. Foundations for an empirically determined scale of trust in automated systems. Int. J. Cogn. Ergon. 4 , 53-71. https://doi.org/10.1207/S15327566IJCE0401_04 (2000). Clark, M., & Payne, R. Character-based determinants of trust in leaders. Risk Anal. 26 , 1161-1173. https://doi.org/10.1111/j.1539-6924.2006.00823.x (2006). Kelley, T. L. The selection of upper and lower groups for the validation of test items. J. Educ. Psychol. 30 , 17-24. https://doi.org/10.1037/h0057123 (1939). Nunnally, J. C., & Bernstein, I. H. Psychometric Theory (3rd ed.). New York: McGraw-Hill (1994). DeVellis, R. F., & Thorpe, C. T. Scale Development: Theory and Applications . Sage Publications (2021). Hair, J. F., Babin, B. J., Anderson, R. E., & Black, W. C. Multivariate Data Analysis (8th ed.). Cengage Learning (2019). Cronbach, L. J. Coefficient alpha and the internal structure of tests. Psychometrika 16 , 297-334 (1951). Hu, L., & Bentler, P. M. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Struct. Equ. Model. 6 , 1-55. https://doi.org/10.1080/10705519909540118 (1999). Lam, L. W. Impact of competitiveness on salespeople’s commitment and performance. J. Bus. Res. 65 , 1328-1334. https://doi.org/10.1016/j.jbusres.2011.10.026 (2012). Fornell, C., & Larcker, D. F. Evaluating Structural Equation Models with Unobservable Variables and Measurement Error. J. Mark. Res. 18 , 39-50. https://doi.org/10.1177/002224378101800104 (1981). Gabriel, I. Artificial intelligence, values, and alignment. Minds Mach. 30 , 411-437. https://doi.org/10.1007/s11023-020-09539-2 (2020). Mechergui, M., & Sreedharan, S. Goal alignment: Reanalyzing value alignment problems using human-aware AI. Proc. AAAI Conf. Artif. Intell. 38 , 14393-14401. https://doi.org/10.1609/aaai.v38i13.29371 (2024). Shen, H., Knearem, T., Ghosh, R., Alkiek, K., Krishna, K., Liu, Y., & Jurgens, D. Towards bidirectional human-AI alignment: A systematic review for clarifications, framework, and future directions. arXiv preprint arXiv:2406.09264 (2024). Feldman-Maggor, Y., Cukurova, M., Kent, C., et al. The Impact of Explainable AI on Teachers’ Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations. Int. J. Artif. Intell. Educ. (2025). https://doi.org/10.1007/s40593-025-00486-6 Huang, K., & Ball, C. The Influence of AI Literacy on User's Trust in AI in Practical Scenarios: A Digital Divide Pilot Study. Proc. Assoc. Inf. Sci. Technol. 61 , 937-939. https://doi.org/10.1002/pra2.1146 (2024). Đerić, E., Frank, D., & Milković, M. Trust in Generative AI Tools: A Comparative Study of Higher Education Students, Teachers, and Researchers. Information 16 , 622. https://doi.org/10.3390/info16070622 (2025). Pratiwi, S., Andayani, A., Winarni, R., & Sumarwati, S. Utilization of artificial intelligence technology in an academic writing class: How do Indonesian students perceive? Contemp. Educ. Technol. 15 , ep450. https://doi.org/10.30935/cedtech/13419 (2023). Pitts, G., Rani, N., Mildort, W., & Cook, E. M. Students’ reliance on AI in higher education: Identifying contributing factors. arXiv: 2506.13845v1 (2025). Luo, Z. Factors contributing to teachers’ acceptance intention to gamified EFL tools: A scale development study. Educ. Technol. Res. Dev. 72 , 447-477 (2024). Amoozadeh, M., et al. Trust in Generative AI among Students: An exploratory study. Proc. 55th ACM Tech. Symp. Comput. Sci. Educ. 1 , 67-73. https://doi.org/10.1145/3626252.3630842 (2024). Luo, Z. Determinants of the perceived usefulness (PU) in the context of using gamification for classroom-based ESL teaching: A scale development study. Educ. Inf. Technol. 28 , 4741-4768 (2023a). Huang, J. L., Curran, P. G., Keeney, J., et al. Detecting and deterring insufficient effort responding to surveys. J. Bus. Psychol. 27 , 99-114. https://doi.org/10.1007/s10869-011-9231-8 (2012). Caldwell, C., & Hansen, M. H. Trustworthiness, governance, and wealth creation. J. Bus. Ethics 97 , 173-188 (2010). Choudhury, A., & Shamszare, H. Investigating the impact of user trust on the adoption and use of ChatGPT: Survey analysis. J. Med. Internet Res. 25 , e47184. https://doi.org/10.2196/47184 (2023). Shin, D. The effects of explainability and causability on perception, trust, and acceptance: Implications for explainable AI. Int. J. Hum.-Comput. Stud. 146 , 102551. https://doi.org/10.1016/j.ijhcs.2020.102551 (2021). Lalot, F., & Bertram, A. When the bot walks the talk: Investigating the foundations of trust in an artificial intelligence (AI) chatbot. J. Exp. Psychol. Gen. 154 , 533-551. https://doi.org/10.1037/xge0001696 (2025). Vice, J., & Khan, M. M. Toward accountable and explainable artificial intelligence part two: The framework implementation. IEEE Access 10 , 36091-36105. https://doi.org/10.1109/ACCESS.2022.3163523 (2022). Stanley, J. C., & Dorton, S. L. Exploring trust with the AI incident database. Proc. Hum. Factors Ergon. Soc. Annu. Meet. 67 , 489-494 (2023). Kaplan, A. D., Kessler, T. T., Brill, J. C., & Hancock, P. A. Trust in artificial intelligence: Meta-analytic findings. Hum. Factors 65 , 337-359 (2023). Mitchell, T. Trust and transparency in artificial intelligence. Philos. Technol. 38 , 87. https://doi.org/10.1007/s13347-025-00916-2 (2025). Kizilcec, R. F. How much information? Effects of transparency on trust in an algorithmic interface. In Proc. 2016 CHI Conf. Hum. Factors Comput. Syst. , 2390-2395. New York, USA (2016). Dimitriadis, S., & Kyrezis, N. Linking trust to use intention for technology-enabled bank channels: The role of trusting intentions. Psychol. Mark. 27 , 799-820. https://doi.org/10.1002/mar.20358 (2010). Worthington, R. L., & Whittaker, T. A. Scale development research: A content analysis and recommendations for best practices. Couns. Psychol. 34 , 806-838. https://doi.org/10.1177/0011000006288127 (2006). Fornell, C., & Larcker, D. F. Evaluating structural equation models with unobservable variables and measurement error. J. Mark. Res. 18 , 39-50 (1981). Additional Declarations No competing interests reported. Supplementary Files Supplementaryfile.zip 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-7939708","acceptedTermsAndConditions":true,"allowDirectSubmit":false,"archivedVersions":[],"articleType":"Article","associatedPublications":[],"authors":[{"id":589606006,"identity":"46ca6e0d-cd93-4823-af07-07a2b88d7d5e","order_by":0,"name":"Ning Mi","email":"data:image/png;base64,iVBORw0KGgoAAAANSUhEUgAAAZAAAAAyAQMAAABI0h/eAAAABlBMVEX///8AAABVwtN+AAAACXBIWXMAAA7EAAAOxAGVKw4bAAAA3UlEQVRIiWNgGAWjYFCCBIYDDAYMDPwMjA9AXMYGorVINjAbEK8FDAwOEKtFvj078XBBwR27zceT2R7zMNjIbjjA/OwBPi0GZ95uODzD4FnytjOP2Y15GNKMNxxgMzfAq0Uid8NhHoPDyWY38o9J8zAcTtxwgIdNAq/DZkC1GM9IZgNq+U9YC8MNiBY7AwmwlgOEtYD9AtSSIHHmMZvkHINk45mH2czwO6w9d/Nnnj+H7fnbk9kk3lTYyfYdb36G32FQkNgAjiBQUDETox4I7OFxOgpGwSgYBaMAHQAAtzVLQdCMS3IAAAAASUVORK5CYII=","orcid":"","institution":"Qufu Normal University","correspondingAuthor":true,"prefix":"","firstName":"Ning","middleName":"","lastName":"Mi","suffix":""},{"id":589606007,"identity":"06692276-9ae3-4029-bec0-937fad01bde9","order_by":1,"name":"Cunying Fan","email":"","orcid":"","institution":"Qufu Normal University","correspondingAuthor":false,"prefix":"","firstName":"Cunying","middleName":"","lastName":"Fan","suffix":""}],"badges":[],"createdAt":"2025-10-24 17:32:13","currentVersionCode":1,"declarations":"","doi":"10.21203/rs.3.rs-7939708/v1","doiUrl":"https://doi.org/10.21203/rs.3.rs-7939708/v1","draftVersion":[],"editorialEvents":[],"editorialNote":"","failedWorkflow":false,"files":[{"id":104283220,"identity":"19799028-b9d0-415f-b320-4cd37696ba30","added_by":"auto","created_at":"2026-03-10 03:58:06","extension":"png","order_by":1,"title":"Figure 1","display":"","copyAsset":false,"role":"figure","size":262089,"visible":true,"origin":"","legend":"\u003cp\u003eThe results of the confirmatory factor analysis.\u003c/p\u003e","description":"","filename":"1.png","url":"https://assets-eu.researchsquare.com/files/rs-7939708/v1/e6475031cfdf2dd82a915ad9.png"},{"id":104779397,"identity":"4b916667-5e45-47ba-9d25-5f4116e386f9","added_by":"auto","created_at":"2026-03-17 07:39:46","extension":"pdf","order_by":0,"title":"","display":"","copyAsset":false,"role":"manuscript-pdf","size":1326722,"visible":true,"origin":"","legend":"","description":"","filename":"manuscript.pdf","url":"https://assets-eu.researchsquare.com/files/rs-7939708/v1/afd77304-fecf-465d-bae1-8e2f7dc9defd.pdf"},{"id":104283221,"identity":"5f4714be-9d8d-447d-b01d-19991870a284","added_by":"auto","created_at":"2026-03-10 03:58:06","extension":"zip","order_by":0,"title":"","display":"","copyAsset":false,"role":"supplement","size":980946,"visible":true,"origin":"","legend":"","description":"","filename":"Supplementaryfile.zip","url":"https://assets-eu.researchsquare.com/files/rs-7939708/v1/cec0e97e455a1a4b2514e5a6.zip"}],"financialInterests":"No competing interests reported.","formattedTitle":"Development and Validation of a Scale to Assess Trust in AI-Assisted English Writing Among Chinese University Students","fulltext":[{"header":"Introduction","content":"\u003cp\u003eThe swift advancement of generative artificial intelligence (AI) has led to the increasing incorporation of AI-based writing tools into higher education, particularly to support English as a foreign language (EFL) learners in their writing tasks. These tools offer immediate feedback on aspects such as language use, text structure, and overall coherence\u003csup\u003e1,2\u003c/sup\u003e. Nevertheless, the mere availability of advanced technology does not guarantee improved learning outcomes. Users\u0026rsquo; willingness to apply AI-generated suggestions is often contingent upon the degree of trust they place in the system\u003csup\u003e3-5\u003c/sup\u003e. Therefore, trust plays a pivotal role in mediating the link between technological capability and actual learning engagement.\u003c/p\u003e\n\u003cp\u003eIn the fields of behavioral science and organizational management, trust has often been described as a psychological state in which an individual is willing to accept vulnerability under conditions of uncertainty or potential risk\u003csup\u003e6\u003c/sup\u003e. This formulation has been widely adopted because it emphasizes the evaluative judgments people make about both ability and goodwill. Importantly, the same conceptualization has been extended beyond interpersonal settings to the study of human-system relationships in technology use\u003csup\u003e3\u003c/sup\u003e. A growing body of work suggests that users\u0026rsquo; trust in automated systems substantially shapes their reliance on system outputs\u003csup\u003e7,8\u003c/sup\u003e. However, the implications are twofold: excessive trust may lead to uncritical acceptance, whereas insufficient trust can result in avoidance or underutilization. In both scenarios, the potential benefits of the technology are undermined to some extent.\u003c/p\u003e\n\u003cp\u003eRecent conceptual advances have, to some extent, sparked growing scholarly attention to the question of how trust can and should be measured across different domains. Several influential frameworks illustrate this diversity, including the Muenster Epistemic Trustworthiness Inventory (METI)\u003csup\u003e9\u003c/sup\u003e, Model of Organizational Trust\u003csup\u003e6\u003c/sup\u003e, the Human-Computer Trust (HCT) model\u003csup\u003e4\u003c/sup\u003e, and the Multi-Dimensional Measure of Trust (MDMT)\u003csup\u003e10\u003c/sup\u003e. Although these instruments demonstrate a relative level of maturity, they remain, by and large, context-general rather than tailored to specific tasks. As a result, they do not fully reflect distinctive features of writing situations, including genre conventions, rhetorical tone, text organization, or the provision of explanatory feedback. Such omissions make it difficult to capture the nuanced ways in which EFL learners decide whether, and to what extent, to rely on AI-generated suggestions in authentic writing tasks.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eAgainst this backdrop, the present study targets EFL learners in Chinese universities with the goal of developing and verifying task-specific and dimensionally explicit scale for AI-assissted writing trust. By doing so, the study seeks to contribute not only to the improved design of AI writing tools but also to the enhancement of students\u0026rsquo; AI literacy and to future research on the dynamics of AI-human collaborative learning.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003eThe definition of trust put forth by Mayer, Davis, and Schoorman, centering on competence, benevolence, and integrity as the basis for one party\u0026apos;s willingness to accept vulnerability, has been highly influential in organizational behavior\u003csup\u003e6\u003c/sup\u003e. The three qualities have been incorporated as central dimensions in a wide range of theoretical frameworks across disciplines. In this sense, trust is inherently relational and carries normative implications, shaping practical choices such as delegation, collaboration, and risk-taking within institutional settings.\u003c/p\u003e\n\u003cp\u003eAs the study of trust moved into technology-mediated contexts, conceptual refinements became necessary. Within human-technology interaction, trust has been reinterpreted more in terms of perceptions of a system\u0026rsquo;s reliability, functionality, and controllability\u003csup\u003e11\u003c/sup\u003e. This adjustment reflects users\u0026rsquo; tendency to evaluate technical performance and output quality in addition to, or even instead of, presumed intentions. In the specific case of automation, Lee and See introduced the notion of calibrated trust, stressing that trust should evolve in line with system performance rather than being treated as a simple binary state\u003csup\u003e3\u003c/sup\u003e. Their analysis pointed to the risks of both excessive trust, leading to blind reliance, and insufficient trust, resulting in unwarranted avoidance.\u003c/p\u003e\n\u003cp\u003eBuilding on these insights, Hoff and Bashir distinguished between performance-oriented cues including accuracy and reliability, and process-oriented cues including transparency and predictability, noting that both categories play a role in shaping users\u0026rsquo; trust assessments\u003csup\u003e4\u003c/sup\u003e. Importantly, trust is shaped not only by perceptions of system accuracy but also by the intelligibility of its underlying decision processes.\u003c/p\u003e\n\u003cp\u003eThe emergence of artificial intelligence, particularly in generative and adaptive forms, adds an additional layer of complexity. As Glikson and Woolley observe, AI systems diverge from traditional automation in that they continually learn and adapt, often through opaque \u0026ldquo;black box\u0026rdquo; mechanisms\u003csup\u003e12\u003c/sup\u003e. In such contexts, user trust depends not only on perceived performance but also on factors, such as the ability to clarify how the system generates decisions or results, the degree to which the system\u0026rsquo;s objectives match users\u0026rsquo; needs or values. Taken together, these developments indicate a paradigmatic transition from a mechanistic understanding of trust to one that also incorporates cognitive and alignment considerations in human-AI interaction.\u003c/p\u003e\n\u003cp\u003eTrust constructs, identified in prior measurement tools that examine comparable areas, range from interpersonal trust to trust in e-commerce, information systems, technology adoption, and expert advice. With the growing role of AI in educational contexts, trust-related scales have been explored specifically designed for learning and instruction\u003csup\u003e13,14\u003c/sup\u003e. Finally, ten relevant instruments were presented (see Table 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 1.\u003c/strong\u003e Ten related scales\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"571\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 270px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eScale\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eAbbreviation\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 173px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eReference\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eTrust in Automated Systems Scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTAS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eJian et al., 2000\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eTrust of Automated Systems Test\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTOAST\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eWojton et al., 2020\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eMulti-Dimensional Measure of Trust (v2)\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eMDMT v2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eUllman \u0026amp; Malle, 2023\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003ee-Commerce Trust Scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eECTS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eMcKnight et al., 2002\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eTrust in Autonomous Vehicles Scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTAVS\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eChoi \u0026amp; Ji, 2015\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eLaypeople\u0026rsquo;s Trust in Experts Scale / Muenster Epistemic Trustworthiness Inventory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eLTE / METI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eHendriks et al., 2015\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eTeachers\u0026rsquo; Trust in AI-Based Educational Technology Instrument\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTTAI-EdTech\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eNazaretsky et al., 2022\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eTrust in AI-Powered Educational Technology Scale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTAI-EdTech\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eNazaretsky et al., 2025\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eHuman-AI interaction\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eHAII\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eShin, 2021\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 270px;\"\u003e\n \u003cp\u003eHuman-like trust in AI and functionality trust in AI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 129px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eTAI\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 173px;\"\u003e\n \u003cp\u003eChoung et al., 2022\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eFollowing a comprehensive review of these scales listed, the constructs of assessing trust in different situations were extracted and compared. Those deemed irrelevant to the research context were excluded, while overlapping categories were consolidated. This iterative refinement ultimately produced five central constructs: competency, integrity, transparency, trust intention and benevolence. Competency, as a trust construct, reflects the extent to which users believe a system or agent possesses the necessary ability to carry out its intended functions successfully and to satisfy user expectations\u003csup\u003e15\u003c/sup\u003e. Transparency, within the context of autonomous and intelligent systems, is commonly understood as the extent to which users are able to anticipate and interpret how the system operates\u003csup\u003e7\u003c/sup\u003e. Integrity was defined as a key component that reflects an individual\u0026rsquo;s adherence to moral and ethical principles, directly influencing the perception of their trustworthiness\u003csup\u003e16\u003c/sup\u003e. Benevolence, as a dimension of trust, has been articulated in prior scales through several complementary aspects. For instance, the Multi-Dimensional Measure of Trust (MDMT v2) emphasizes goodwill and considerateness\u003csup\u003e10\u003c/sup\u003e, the e-Commerce Trust Scale (ECTS) highlights ethical behavior and moral concern\u003csup\u003e11\u003c/sup\u003e. Trust intention, as identified in prior measurement frameworks, generally denotes an individual\u0026rsquo;s willingness or expressed likelihood to rely on a trustee. This includes both the readiness to depend and the subjective probability of doing so\u003csup\u003e11\u003c/sup\u003e. \u0026nbsp; \u0026nbsp;\u003c/p\u003e"},{"header":"Results","content":"\u003cp\u003e\u003cstrong\u003eItem Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo evaluate the quality of the initial items, data from Sample 1 were analyzed using three procedures: independent-samples t-tests based on the upper and lower 27% groups, item-total correlation analysis, and internal consistency reliability.\u003c/p\u003e\n\u003cp\u003eFirst, independent-samples t-tests were conducted by selecting the top 27% of respondents (high-score group) and the bottom 27% (low-score group) according to the total scores\u003csup\u003e17\u003c/sup\u003e. Results showed that all 28 items significantly differentiated between the two groups, indicating satisfactory item discrimination.\u003c/p\u003e\n\u003cp\u003eSecond, item-total correlation coefficients were computed to examine the degree to which each item was consistent with the overall construct. While correlations above .30 are generally considered acceptable\u003csup\u003e18\u003c/sup\u003e, values greater than .40 are recommended as a more stringent standard\u003csup\u003e19,20\u003c/sup\u003e. Results indicated that all items were significantly correlated with the total score (p \u0026lt; .001); however, Item Q28 showed a coefficient of only .307, failing to meet the .40 criterion.\u003c/p\u003e\n\u003cp\u003eFinally, internal consistency reliability was assessed using Cronbach\u0026rsquo;s \u0026alpha;\u003csup\u003e21\u003c/sup\u003e. The 28-item scale yielded an overall \u0026alpha; of .948. Examination of the corrected item-total correlations (CITC) and the \u0026ldquo;\u0026alpha; if item deleted\u0026rdquo; statistics suggested that three items, Q5 (CITC = .418, \u0026alpha; = .949), Q23 (CITC = .454, \u0026alpha; = .948), and Q28 (CITC = .247, \u0026alpha; = .951), weakened the overall consistency of the scale. Considering both the low CITC values and the improvement of \u0026alpha; when these items were removed, the deletion of Q5, Q23, and Q28 was deemed appropriate. The remaining 25 items demonstrated satisfactory psychometric properties and were retained for subsequent exploratory factor analysis.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eExploratory Factor Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eWe conducted an exploratory factor analysis on the remaining 25 items using Sample 1 data. It was found that the rotation converged after seven iterations using principal component analysis and Kaiser normalization optimal oblique rotation. Statistical results showed that the KMO value was 0.945, exceeding the threshold of 0.7. Bartlett\u0026rsquo;s sphericity test yielded a significant result (approximate chi-square value of 4282.441, degrees of freedom 300, p \u0026lt; 0.001), indicating that the sample data were suitable for factor analysis. Based on the rotated factor matrix, factors with eigenvalues greater than 1 were extracted. In the first round of exploratory factor analysis, the factor loadings for items Q6, Q13, and Q22 were all less than 0.5, failing to meet the factor extraction criteria (Hair et al., 2014, set at 0.4). After multiple rounds of exploratory factor analysis, five items (Q6, Q13, Q14, Q22, and Q24) were removed, resulting in a four-factor structure comprising 20 items. The common factor variance extraction values for all items ranged from 0.563 to 0.839. The final structure had a KMO value of 0.936, exceeding 0.8, and Bartlett\u0026rsquo;s sphericity test was significant (approximate chi-square 3321.382, degrees of freedom 190, p \u0026lt; 0.001). Principal component analysis and Kaiser normalization were used, and the optimal oblique rotation converged after six iterations, extracting four factors with eigenvalues greater than 1, explaining a cumulative variance of 69.466%. The specific loadings, factor characteristics, and variance explained for each item are shown in Table 2.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 2.\u003c/strong\u003e Result of exploratory factor analysis.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"590\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"4\" style=\"width: 277px;\"\u003e\n \u003cp\u003eFactor loading coefficient\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"2\" style=\"width: 96px;\"\u003e\n \u003cp\u003eCommunality\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003eMean\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003eSD\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003eFactor 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003eFactor 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003eFactor 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003eFactor4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.822\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e0.931\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.706\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e4.00\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.713\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e0.876\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.677\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 10\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.783\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e0.819\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.74\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.94\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e0.787\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.714\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 11\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.758\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e0.776\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.706\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e4.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.651\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e0.578\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.653\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.80\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.769\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.899\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.612\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.84\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.749\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.839\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.686\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 20\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.758\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.793\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.663\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 18\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.91\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.707\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.623\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.662\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.816\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.621\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.608\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 21\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.88\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.69\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.588\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.658\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.93\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.728\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.515\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.563\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e4.14\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.674\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.946\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.821\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e4.12\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.675\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.926\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.839\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e4.17\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.645\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.914\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.826\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e4.06\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.715\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.688\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.674\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 25\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.667\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.912\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.727\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 27\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.89\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.728\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.773\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.687\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 76px;\"\u003e\n \u003cp\u003eItem 26\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 64px;\"\u003e\n \u003cp\u003e3.96\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 77px;\"\u003e\n \u003cp\u003e0.654\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e0.767\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e0.669\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 217px;\"\u003e\n \u003cp\u003eEigenvalue\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e9.432\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e2.234\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e1.219\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e1.007\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 217px;\"\u003e\n \u003cp\u003eContribution Rate %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e47.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e11.172\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e6.097\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e5.037\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"3\" style=\"width: 217px;\"\u003e\n \u003cp\u003eCumulative Contribution Rate %\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 67px;\"\u003e\n \u003cp\u003e47.16\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e58.332\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e64.429\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 70px;\"\u003e\n \u003cp\u003e69.466\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 96px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd colspan=\"8\" style=\"width: 590px;\"\u003e\n \u003cp\u003eNote: Factor 1=Goal alignment trust; Factor 2=Transparency; Factor 3=Competency; Factor 4=Trust intention\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eAs shown in Table 2, the factor loadings for all items range from 0.515 to 0.946 and are aggregated into four factors. Based on the content of the descriptive items included in each factor, the four factors are named F1 (Goal Alignment), F2 (Transparency ), F3 (Competency), and F4 (Trust intention).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eConfirmatory Factor Analysis\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eTo ensure robustness, we used AMOS 26.0 to conduct confirmatory factor analysis on Sample 2 (N=270) to measure the overall fit of the factor structure. The results demonstrated that the chi-square/df ratio reached 1.855, falling within the ideal range of 1 to 3 for adequate model fit. The Comparative Fit Index (CFI) yielded a value of 0.961, while the Incremental Fit Index (IFI) also stood at 0.961. Additionally, the Tucker-Lewis Index (TLI) was 0.955, and the Normed Fit Index (NFI) reached 0.950, all of these indices surpassing the acceptable threshold of 0.90. For the Root Mean Square Error of Approximation (RMSEA), the calculated value was 0.056. Though this figure exceeds the 0.05 threshold for excellent model fit, it still falls within the acceptable range of 0.05 to 0.08. The benchmarks for evaluating model fit were established in accordance with the guidelines proposed by Hair et al. and Hu and Bentler\u003csup\u003e21,22\u003c/sup\u003e. Taken together, these results verify that the model exhibits a good level of fit (see Table 3 and Figure 1).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 3.\u0026nbsp;\u003c/strong\u003eModel fit\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eIndicator\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eBenchmark\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003eResult\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eInterpretation\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eChi-square/df\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eBetween 1 to 3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003e1.855\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eExcellent\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eCFI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eLarger than 0.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003e0.961\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eExcellent\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eIFI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eLarger than 0.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003e0.961\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eExcellent\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eTLI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eLarger than 0.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003e0.955\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eExcellent\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eNFI\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eLarger than 0.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003e0.950\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eExcellent\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 25px;\"\u003e\n \u003cp\u003eRMSEA\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 31px;\"\u003e\n \u003cp\u003eSmaller than 0.08\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 16px;\"\u003e\n \u003cp\u003e0.056\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 26px;\"\u003e\n \u003cp\u003eGood\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003c/br\u003e\n\u003cp\u003e\u003cstrong\u003eReliability and Validity Testing\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eConvergent validity was evaluated using values of average variance extracted (AVE) and composite reliability (CR),\u0026nbsp;with the detailed findings presented in Table 4.\u003c/p\u003e\n\u003cp\u003eFor reliability analysis, Cronbach\u0026rsquo; s\u0026nbsp;\u0026alpha;\u0026nbsp;coefficient for the four dimensions were all above 0.70, specifically 0.917, 0.902, 0.895, and 0.824, demonstrating strong internal consistency. Composite reliability (CR) values ranged from 0.829 to 0.918, also exceeding the 0.70 threshold, further confirming the reliability of the trust scale for AI-assisted English writing.\u003c/p\u003e\n\u003cp\u003eWith respect to convergent validity, the AVE for Factor 2 was 0.58, slightly below the conventional cut-off of 0.60. However, as AVE is considered a conservative estimate, prior studies suggest that if CR exceeds 0.70, convergent validity may still be deemed acceptable even when AVE is below 0.60\u003csup\u003e23\u003c/sup\u003e. Given that the CR for this factor was 0.906, well above the recommended threshold, the dimension can reasonably be considered to demonstrate adequate convergent validity.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 4\u003c/strong\u003e. Construct validity and convergent validity\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eConstruct\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eItem\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eEstimate\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eS.E.\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eZ\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003e\u003cem\u003ep\u003c/em\u003e\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eFactor loading\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCronbach\u0026rsquo;s alpha\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eAVE\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u003cstrong\u003eCR\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1_1 \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.764\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"6\" valign=\"top\"\u003e\n \u003cp\u003e0.917\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"6\" valign=\"top\"\u003e\n \u003cp\u003e0.650\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"6\" valign=\"top\"\u003e\n \u003cp\u003e0.918\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1_2 \u0026nbsp;\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.076\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.081\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e13.344\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.775\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1_3 \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.945\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.069\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e13.681\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.791\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1_4 \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.147\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.078\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e14.669\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.839\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1_5 \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.114\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.074\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e15.004\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.855\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF1_6 \u0026nbsp; \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.09\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.078\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e14.049\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.809\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.622\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\"\u003e\n \u003cp\u003e0.902\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\"\u003e\n \u003cp\u003e0.58\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\"\u003e\n \u003cp\u003e0.906\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.155\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.108\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10.721\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.807\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.059\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.101\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10.474\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.781\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.022\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10.385\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.772\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.052\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.099\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.794\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.092\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.105\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10.412\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.774\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF2_7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.018\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.098\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e10.334\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.766\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF3_1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.897\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003e0.895\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003e0.704\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"4\" valign=\"top\"\u003e\n \u003cp\u003e0.904\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF3_2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.877\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.043\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e20.517\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.883\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF3_3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.912\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.047\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e19.505\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.86\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF3_4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.853\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.062\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e13.691\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.701\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF4_1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.718\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" valign=\"top\"\u003e\n \u003cp\u003e0.824\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" valign=\"top\"\u003e\n \u003cp\u003e0.618\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"3\" valign=\"top\"\u003e\n \u003cp\u003e0.829\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF4_2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.064\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.088\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e12.067\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.844\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003eF4_3\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e1.086\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.093\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e11.637\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e***\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\"\u003e\n \u003cp\u003e0.792\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eDiscriminant validity refers to the extent to which a construct is distinct from other constructs, typically demonstrated through low correlations between the latent traits they represent. According to the criterion\u003csup\u003e24\u003c/sup\u003e, discriminant validity is established when the square root of the AVE for a construct exceeds its correlations with other constructs. In this study, the results fulfilled this requirement as the square root of the AVE for each construct surpassed its inter-construct correlations, confirming that the dimensions exhibit adequate discriminant validity (see Table 5).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 5\u003c/strong\u003e. AVE, the square root of AVE and the correlations between constructs.\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\" width=\"100%\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 18px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF4\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 19px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF4\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.786\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 19px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF3\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.508\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 19px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.839\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF2\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.653\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 19px;\"\u003e\n \u003cp\u003e0.541\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.762\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 18px;\"\u003e\n \u003cp\u003e\u003cstrong\u003eF1\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 20px;\"\u003e\n \u003cp\u003e0.614\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 19px;\"\u003e\n \u003cp\u003e0.681\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e0.745\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 21px;\"\u003e\n \u003cp\u003e\u003cstrong\u003e0.806\u003c/strong\u003e\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e"},{"header":"Discussion","content":"\u003cp\u003eThis study developed and validated a four-factor trust scale for AI-assisted English writing, encompassing Goal Alignment, Transparency, Competency, and Trust Intention. The instrument demonstrated sound psychometric qualities, with both exploratory and confirmatory factor analyses confirming its reliability and validity. These results resonate with classical trust frameworks while also extending recent empirical findings\u003csup\u003e6,11\u003c/sup\u003e.\u003c/p\u003e\n\u003cp\u003eFirstly, introducing Goal Alignment as a distinct dimension extends existing trust models.While classical models emphasized benevolence and integrity\u003csup\u003e6\u003c/sup\u003e, this study conceptualizes trust in AI-assisted writing as strongly tied to whether system outputs align with students\u0026rsquo; learning objectives.This perspective resonates with recent discussions of value and goal alignment in AI ethics and human-AI interaction. For instance, Gabriel highlighted value alignment as essential for trustworthy AI\u003csup\u003e25\u003c/sup\u003e, while Mechergui and Sreedharan reframed alignment in terms of human-aware goal matching\u003csup\u003e26\u003c/sup\u003e. Similarly, Shen et al. proposed a bidirectional perspective, emphasizing that alignment is dynamic and requires adaptation from both humans and AI\u003csup\u003e27\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eSecond, Transparency emerged as another critical dimension of trust. The findings suggest that students value not only accurate feedback but also insight into the logic and mechanisms that generate it. This observation is consistent with Malle and Ullman\u0026rsquo;s emphasis on explainability as central to human-AI trust and with Lee and See\u0026rsquo;s notion of calibrated reliance\u003csup\u003e10,3\u003c/sup\u003e. Recent empirical evidence reinforces this view which showed that higher levels of explainability significantly increased teachers\u0026rsquo; trust in AI tools\u003csup\u003e28\u003c/sup\u003e. Similarly, Huang and Ball found that students with greater AI literacy developed more calibrated trust, underscoring the interdependence of transparency, literacy, and trust\u003csup\u003e29\u003c/sup\u003e. Collectively, these findings highlight transparency, coupled with user literacy, as essential for cultivating both confidence and critical engagement with AI in educational contexts.\u003c/p\u003e\n\u003cp\u003eThird, Competency remains the cornerstone of trust. Students consistently evaluated AI tools by their capacity to generate high-quality texts, echoing Mayer et al.\u0026rsquo;s \u0026ldquo;ability\u0026rdquo; dimension and aligning with Jian et al.\u0026rsquo;s findings in automation\u003csup\u003e6,15\u003c/sup\u003e. Recent evidence also supports this view: Đerić et al. identified competency and transparency as the strongest predictors of student trust\u003csup\u003e30\u003c/sup\u003e, while Pratiwi et al. reported that Indonesian students valued AI for idea generation but questioned its limited revision and editing capabilities\u003csup\u003e31\u003c/sup\u003e. In both research and practice, teachers and students alike underscore effectiveness, reinforcing competency as the fundamental basis of trust.\u003c/p\u003e\n\u003cp\u003eFinally, Trust Intention was confirmed as an independent dimension of the trust construct, capturing students\u0026rsquo; willingness to rely on and continue using AI-assisted writing tools. This interpretation is consistent with McKnight et al.\u0026rsquo; s model and parallels Choi and Ji\u0026rsquo; s findings in the automation domain\u003csup\u003e11,7\u003c/sup\u003e, where intention to use was treated as a direct indicator of trust. Recent evidence further refines this perspective: Pitts et al. showed that students\u0026rsquo; reliance behaviors were closely tied to their trust and satisfaction, but also revealed risks of overreliance when trust intention was uncalibrated\u003csup\u003e32\u003c/sup\u003e. These results align with Lee and See\u0026rsquo; s principle of appropriate reliance, suggesting that trust intention should not be read as uniformly positive\u003csup\u003e3\u003c/sup\u003e. Pedagogically, fostering trust intention requires pairing it with transparency and literacy support to encourage calibrated rather than blind reliance.\u003c/p\u003e\n\u003cp\u003eTaken together, the findings draw on classical trust models while incorporating recent evidence to clarify their relevance in educational contexts. Theoretically, it extends trust research into AI-assisted learning by underscoring the roles of goal alignment, transparency,competency and trust intention. Practically, it suggests that teachers can foster trust by aligning AI feedback with learning objectives, whereas developers should prioritize transparency and performance to encourage acceptance and sustained use. Future research should test the scale across cultural contexts and integrate behavioral and longitudinal data to enhance its ecological validity and explanatory strength.\u003c/p\u003e"},{"header":"Conclusions","content":"\u003cp\u003eThis study developed and validated the Trust Scale of AI-Assisted English Writing (TS-AIEW) to measure Chinese university students\u0026rsquo; trust in AI-assisted writing tools. Drawing on prior trust-related scales, four constructs were identified: Goal Alignment, Transparency, Competency, and Trust Intention. Following established scale development procedures, item analysis and reliability and validity testing were conducted, resulting in the removal of eight items\u003csup\u003e19,33\u003c/sup\u003e. The final instrument comprises 20 items across the four constructs, all of which demonstrated strong psychometric properties.\u003c/p\u003e\n\u003cp\u003eThe TS-AIEW offers a robust tool for assessing university students\u0026rsquo; trust in AI-assisted English writing, enabling reliable evaluation and ongoing monitoring. Its main contribution lies in informing tailored instructional strategies and guiding the design of trustworthy educational technologies. Moreover, the scale provides a foundation for comparative and cross-contextual research, advancing understanding of learners\u0026rsquo; trust dynamics in technology-enhanced education.\u003c/p\u003e\n\u003cp\u003eOne limitation of this study is that the Trust Scale of AI-Assisted English Writing (TS-AIEW), though developed and validated with rigorous psychometric procedures, was tested primarily with Chinese university students. This cultural specificity may restrict the generalizability of the findings. Future research should validate the TS-AIEW with more diverse cultural and linguistic groups to establish its cross-cultural applicability. The framework could also be adapted to other educational technologies by substituting the construct of \u0026ldquo;AI-assisted English writing\u0026rdquo; with domains such as \u0026ldquo;AI-assisted academic reading\u0026rdquo; or \u0026ldquo;AI-assisted translation\u0026rdquo;. In addition, pedagogically oriented research is needed to explore how trust interacts with learning outcomes, motivation, and critical thinking\u003csup\u003e34\u003c/sup\u003e. Finally, future studies should investigate contextual features of AI-assisted systems that may shape students\u0026rsquo; trust calibration, such as autonomy, feedback form, and explainability\u003csup\u003e10,3\u003c/sup\u003e.\u003c/p\u003e"},{"header":"Methods","content":"\u003cp\u003e\u003cstrong\u003eResearch procedure\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study follows a scale development approach consisting of several research stages, guided by the procedures outlined in DeVellis and Thorpe (2021), Luo (2024) and Luo (2023a)\u003csup\u003e19,33,35\u003c/sup\u003e.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe first stage involved the identification of the scale\u0026rsquo; s constructs. In this regard, constructs are described as the theoretical dimensions or latent factors that denote the phenomenon under investigation\u003csup\u003e19\u003c/sup\u003e. Drawing on related instruments, we integrated relevant constructs, removed those outside the scope of this study, refined their wording, and ultimately identified four dimensions: Competency, Goal Alignment, Transparency, and Trust Intention.\u003c/p\u003e\n\u003cp\u003eStage 2 focused on establishing the item pool. In scale development, items refer to the specific survey questions or statements designed to capture data about participants\u0026rsquo; attitudes, behaviors, or perceptions. As DeVellis and Thorpe recommend, beginning with a broad pool of items helps ensure adequate coverage of the constructs\u003csup\u003e19\u003c/sup\u003e. Item generation can be achieved through several strategies, including adapting items from existing scales, drafting items from theoretical foundations, or converting qualitative insights into survey statements. In this study, we primarily relied on the first and the third approach, modifying items from related scales and transforming qualitative responses, to ensure alignment with the scope of AI-assisted English writing (see Table 6).\u003c/p\u003e\n\u003cp\u003eThe third stage involved the validation conducted by experts. The draft version of the TS-AIEW scale was reviewed by two professors with doctoral degrees and relevant research experience. They were asked to examine the scale for logical consistency, clarity of language, and alignment with the study\u0026rsquo; s objectives and theoretical framework\u003csup\u003e19\u003c/sup\u003e. In addition, the experts evaluated each item in terms of relevance, necessity, and cultural adaptability. Based on their feedback, we finalized an initial version of the TS-AIEW consisting of 28 items, which could typically be completed within 5 to 8 minutes (see Table 6).\u003c/p\u003e\n\u003cp\u003eStage 4 centered on the validation of data. Following expert review, the scale of TS-AIEW was administered to participants. Surveys completed in less than 60 seconds or with identical responses across all items were excluded from analysis\u003csup\u003e36\u003c/sup\u003e. Using SPSS 26.0, we examined descriptive statistics for each item, calculated Cronbach\u0026rsquo; s alpha for each construct, and tested the scale\u0026rsquo; s suitability for factor analysis before conducting exploratory factor analysis (EFA). The process was iterative: when items were removed, the scale was re-evaluated to ensure reliability, which occasionally required further adjustments\u003csup\u003e33\u003c/sup\u003e. Through several rounds of refinement, we produced a validated version of the TS-AIEW scale, presented in the Findings section.\u003c/p\u003e\n\u003cp\u003eStage 5 addressed the evaluation of construct validity and the establishment of reliability for the TS-AIEW scale. Using AMOS 26.0, we conducted confirmatory factor analysis (CFA) to validate the factor structure identified by means of EFA. Model fit was evaluated with indices including the Chi-square statistic, RMSEA, and CFI, to examine the degree of consistency between the data and the hypothesized model. Factor loadings were also examined to confirm that each item loaded adequately on its intended construct, thereby strengthening evidence for construct validity. The detailed results of this analysis are reported in the Findings section.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eScale development\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eIn constructing the Trust Scale for AI-assisted English Writing among Chinese university students (TS-AIEW), we systematically integrated four central dimensions that reflect both the contextualized features of AI-supported educational writing and insights derived from earlier measurement frameworks. These constructs are competency, alignment, transparency, and trust intention, which serve as the theoretical foundation of the scale (see Table 6).\u003c/p\u003e\n\u003cp\u003eFirstly, we retained the construct of Competency as it is among the most widely used dimensions in trust research. Competency refers to the skills and expertise that enable effective performance\u003csup\u003e37\u003c/sup\u003e. In AI-assisted English writing, it can be operationalized as learners\u0026rsquo; belief that such tools can identify problems, provide useful suggestions, and support task completion for Chinese EFL students. In the current study, the first four items were adapted from trust beliefs framework\u003csup\u003e11\u003c/sup\u003e. For example, the original item \u0026ldquo;LegalAdvice.com is competent and effective in providing legal advice\u0026rdquo; was adapted to \u0026ldquo;AI-assisted writing tool is competent and effective in providing writing advice.\u0026rdquo; Additional items (5-7) were derived from face-to-face depth interviews, including \u0026ldquo;All the feedback I received from the AI-assisted writing tool was correct\u0026rdquo; \u0026ldquo;I can rely on the advice given by AI-assisted writing tools\u0026rdquo; and \u0026ldquo;The suggestions provided by AI-assisted writing tools are suitable for learners of different English proficiency levels\u0026rdquo;.\u003c/p\u003e\n\u003cp\u003eSecondly, we kept the significance of benevolence in trust measurement but reframed it as Goal Alignment to better fit the context of AI-assisted English writing. Benevolence has traditionally been defined as responsiveness and goodwill toward others\u003csup\u003e11\u003c/sup\u003e, and recent work highlights its importance in shaping user trust in AI-driven interactions\u003csup\u003e38\u003c/sup\u003e. In our study, this notion is extended to capture the extent to which AI writing tools align with learners\u0026rsquo; educational goals and act in their best interest. Operationally, goal alignment reflects students\u0026rsquo; perception that the system provides feedback supportive of their learning objectives and demonstrates care and responsibility for their academic development. For example, the item \u0026ldquo;I found the AI-assisted feedback are a great match to my English writing needs\u0026rdquo; was adapted from Shin to fit the current context of EFL writing\u003csup\u003e39\u003c/sup\u003e. \u0026nbsp;\u003c/p\u003e\n\u003cp\u003eOur study excluded the construct of Integrity because it primarily emphasizes moral and ethical aspects of AI behavior, such as honesty and fairness\u003csup\u003e40\u003c/sup\u003e. In the present context, the central concern lies in how effectively AI tools support writing tasks, rather than in the more subjective assessment of ethical standards\u003csup\u003e41\u003c/sup\u003e. For users of educational AI systems, sustained trust is more likely to stem from performance and goal alignment than from perceptions of integrity\u003csup\u003e42,43\u003c/sup\u003e. Accordingly, the integrity dimension was omitted from the final framework.\u003c/p\u003e\n\u003cp\u003eThirdly, we retained Transparency in the scale, following prior research that has emphasized its central role in shaping trust\u003csup\u003e44\u003c/sup\u003e. In AI systems, transparency refers to the clarity with which system operations are communicated, including the algorithms applied, the data processed, and the reasoning behind outputs. Empirical work has demonstrated its relevance in educational contexts, showing that higher transparency strengthens user trust in AI-driven learning tools\u003csup\u003e45\u003c/sup\u003e. Shin further conceptualized transparency through three interrelated aspects: understandability, explainability, and visibility\u003csup\u003e39\u003c/sup\u003e. In this study, transparency is operationalized as students\u0026rsquo; perception that AI writing tools provide feedback in a manner that is understandable, interpretable, and predictable. Items developed for this dimension therefore capture whether learners consider the system\u0026rsquo; s suggestions sufficiently clear and well justified to be trusted in authentic writing tasks. For example, \u0026ldquo;I have a general understanding of how AI generates writing feedback.\u0026rdquo;\u003csup\u003e39\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003eFinally, the term \u0026ldquo;Trust Intention\u0026rdquo; has been retained as is. This construct encapsulates the willingness of users to engage with AI systems based on their perceptions of competence, alignment, and transparency. This willingness is essential for encouraging users to engage with technology, underlining the significance of trust intention as a predictor of actual use\u003csup\u003e46\u003c/sup\u003e. In the present study, trust intention is operationalized as students\u0026rsquo; willingness to continue using AI writing tools for their English writing tasks, based on their perception of the competency, goal alignment and transparency of AI-assisted writing tool. In accordance with the prior related survey questions, we made modifications to fit the theme of AI-assisted English writing. For instance, the statement from McKnight et al. \u0026ldquo;I feel that I could count on LegalAdvice.com to help with a crucial legal problem\u0026rdquo; was revised to \u0026ldquo;AI writing tools can improve the quality of my writing.\u0026rdquo;\u003csup\u003e11\u0026nbsp;\u003c/sup\u003e\u003c/p\u003e\n\u003cp\u003eIn conclusion, the current research has identified four constructs linked to trust in AI-facilitated English writing among university students in China: competence, goal alignment, transparency, and trust intention. With reference to the phrasing of measuring items from existing related scales, we developed the initial version of the TS-AIEW scale, and its details are provided in Table 6.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 6.\u0026nbsp;\u003c/strong\u003eConstructs and measuring items of the initial version of TS-AIEW (28 items).\u003c/p\u003e\n\u003ctable border=\"1\" cellspacing=\"0\" cellpadding=\"0\" width=\"587\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 95px;\"\u003e\n \u003cp\u003eConstructs\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eMeasuring items\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 155px;\"\u003e\n \u003cp\u003eLiterature sources\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eCompetency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 1. AI-assisted writing tool is competent and effective in providing writing advice.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 155px;\"\u003e\n \u003cp\u003eMcKnight et al. (2002); Wojton et al. (2019) \u0026nbsp; \u0026nbsp; Choung et al. (2022);\u0026nbsp;\u003c/p\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 2. AI-assisted writing tool performs its role of giving writing advice very well. \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 3. Overall, AI-assisted writing tool is a capable and proficient writing advice provider.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 4. In general, AI-assisted writing tool is very knowledgeable about English writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 5*. The suggestions provided by AI-assisted writing tool are suitable for learners of different English proficiency levels.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 6*. All the feedback I received from the AI-assisted writing tool was correct.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 7*. I can rely on the advice given by AI-assisted writing tool.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eGoal Alignment\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 8. AI feedback is basically consistent with what I want to express in my English writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 155px;\"\u003e\n \u003cp\u003eNazaretsky et al. (2025); Choung et al. (2022); Shin (2021)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 9. I found AI feedback is a great match to my English writing needs.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 10. I feel that AI-assisted writing tool understands my writing needs.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 11. AI-assisted writing tool helps me achieve the desired writing effect.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 12*. AI-assisted writing tool helps me maintain clarity of my views throughout the writing process.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 13*. AI suggestions can help me maintain a clear structure and train of thought in my writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 14*. AI-assisted writing tool helps me focus on my\u0026nbsp;\u003c/p\u003e\n \u003cp\u003ewriting goals.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eTransparency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 15. I have a general understanding of how AI\u0026nbsp;\u003c/p\u003e\n \u003cp\u003egenerates writing feedback.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 155px;\"\u003e\n \u003cp\u003eShin (2021);\u0026nbsp;\u003c/p\u003e\n \u003cp\u003eNazaretsky et al. (2025);\u003c/p\u003e\n \u003cp\u003eWojton et al. (2019)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 16. AI-assisted writing tool can clearly explain the\u0026nbsp;\u003c/p\u003e\n \u003cp\u003ereasons behind their suggestions.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 17. I can understand the reasoning behind a particular suggestion made by AI-assisted writing tool. \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 18. I find the presentation of AI feedback clear and understandable.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 19*. AI-assisted writing tool provides clear and understandable explanations for language issues (such as grammar and word choice). \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 20*. AI-assisted writing tool helps me distinguish between correct and incorrect English expressions. \u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 21*. The explanations provided by AI-assisted writing tool help me understand English writing rules.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 95px;\"\u003e\n \u003cp\u003eTrust Intention\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 22. The suggestions provided by AI-assisted writing tool influence how I revise my writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd rowspan=\"7\" valign=\"top\" style=\"width: 155px;\"\u003e\n \u003cp\u003eMcKnight et al. (2002); Nazaretsky et al. (2025)\u003cbr\u003e\u0026nbsp;Choung et al. (2022); Shin (2021)\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 23*. I believe that AI-assisted writing tool can improve the quality of my writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 24*. I believe that the suggestions provided by AI-assisted writing tool can improve my English writing skills.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 25. I am willing to use AI-assisted tool as an aid in the process of English writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 26. When completing English writing tasks, I tend to actively use AI-assisted writing tool.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 27. Even if I occasionally find that the suggestions of AI-assisted writing tool are incorrect, I will still continue to use the tool.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 337px;\"\u003e\n \u003cp\u003eItem 28*. The suggestions provided by AI-assisted writing tool influence how I revise my writing.\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003eNote: Items marked with an asterisk (*) have been\u0026nbsp;added to the scale after the face-to-face in-depth interview;\u0026nbsp;participants were requested to finish the survey in accordance with their experiences over the past one semester.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eQuestionnaire Design and Data Collection\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAlthough the TS-AIEW scale was developed through the procedures outlined above, it had to be adapted into a questionnaire format for administration. A scale is designed to measure a specific construct, whereas a questionnaire is a broader instrument that may include multiple constructs as well as background information. In this study, all items were measured on a 5-point Likert scale, ranging from 1 (\u0026ldquo;strongly disagree\u0026rdquo;) to 5 (\u0026ldquo;strongly agree\u0026rdquo;). In addition to the TS-AIEW items, questions on demographic variables (e.g., gender, grade) were included.\u003c/p\u003e\n\u003cp\u003eGiven that the study was conducted in China, all measurement items were translated into Chinese. To safeguard linguistic precision, the back-translation method was adopted: items were first translated from English to Chinese, and then a different translator independently translated them back into English. The original and its corresponding back-translated versions were compared to ensure semantic equivalence.\u003c/p\u003e\n\u003cp\u003eData were collected through an online survey platform (www.wjx.cn). The link was distributed mainly via QQ and WeChat among university students who were invited to participate voluntarily. The collection period lasted from late June to mid-July 2025. Before the main survey, a one-week pretest was conducted with twenty graduate students familiar with AI-assisted English writing to evaluate clarity and usability.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eParticipants and Ethical Considerations\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eEthical approval for this study was obtained from the authors\u0026rsquo; institutional review board prior to data collection. Convenience sampling was employed, yielding 728 completed questionnaires. After excluding 201 invalid cases due to insufficient response time or repetitive answering patterns, 527 valid responses were retained, resulting in an effective response rate of 72.4%.\u003c/p\u003e\n\u003cp\u003eData analysis was conducted using SPSS 27.0 and AMOS 26.0. To avoid the problem of capitalizing on chance by using the same dataset for both exploratory and confirmatory factor analyses, we randomly split the full sample into two approximately equal subsamples. Sample One (\u0026asymp;50%, N=257) was used for item analysis and EFA, while Sample Two (\u0026asymp;50%, N=270) was reserved for CFA. This approach has been recommended in methodological literature, which emphasizes that EFA and CFA should ideally be conducted on independent datasets and that one practical solution is to divide a large sample randomly into two subsamples of roughly equal size\u003csup\u003e47\u003c/sup\u003e. Comparisons indicated that the two subsamples were broadly similar in demographic variables such as gender, English proficiency, college, and grade, with no significant differences (see Table 7).\u003c/p\u003e\n\u003cp\u003eAll procedures adhered to established ethical principles. Participants were informed of the study objectives, assured of their rights and confidentiality, and reminded that participation was entirely voluntary. Anonymity was maintained throughout, and the research protocol successfully passed a rigorous ethical review process.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eTable 7.\u0026nbsp;\u003c/strong\u003eDemographic characteristics of participants for Sample 1 and Sample 2 (%)\u003c/p\u003e\n\u003ctable border=\"0\" cellspacing=\"0\" cellpadding=\"0\"\u003e\n \u003ctbody\u003e\n \u003ctr\u003e\n \u003ctd style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd style=\"width: 133px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 143px;\"\u003e\n \u003cp\u003eSample 1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd colspan=\"2\" style=\"width: 150px;\"\u003e\n \u003cp\u003eSample 2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003eCategory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eSubcategory\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003eN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003ePercentage\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 55px;\"\u003e\n \u003cp\u003eN\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003ePercentage\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003eGender\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eMale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e85\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e33.1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e73\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e27.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eFemale\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e172\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e66.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e197\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e73.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003eMajor\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eHumanities\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e99\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e38.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e127\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e47.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eScience\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e71\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e27.6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e68\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e25.2\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eEngineering\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e87\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e33.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e75\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e27.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003eAcademic Year\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eFreshman\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e106\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e41.2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e128\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e47.4\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eSophomore\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e141\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e54.9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e121\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e44.8\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eJunior\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e9\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e3.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e19\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e7.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003eSenior\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e1\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e0.4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e2\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e0.7\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"top\" style=\"width: 142px;\"\u003e\n \u003cp\u003eEnglish Proficiency\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 133px;\"\u003e\n \u003cp\u003eNot Passed CET-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 59px;\"\u003e\n \u003cp\u003e140\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 83px;\"\u003e\n \u003cp\u003e54.5\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 55px;\"\u003e\n \u003cp\u003e162\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"top\" style=\"width: 96px;\"\u003e\n \u003cp\u003e60.0\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003ePassed CET-4\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e102\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e39.7\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e76\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e28.1\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003ctr\u003e\n \u003ctd valign=\"bottom\" style=\"width: 142px;\"\u003e\n \u003cp\u003e\u0026nbsp;\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 133px;\"\u003e\n \u003cp\u003ePassed CET-6\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 59px;\"\u003e\n \u003cp\u003e15\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 83px;\"\u003e\n \u003cp\u003e5.8\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 55px;\"\u003e\n \u003cp\u003e32\u003c/p\u003e\n \u003c/td\u003e\n \u003ctd valign=\"bottom\" style=\"width: 96px;\"\u003e\n \u003cp\u003e11.9\u003c/p\u003e\n \u003c/td\u003e\n \u003c/tr\u003e\n \u003c/tbody\u003e\n\u003c/table\u003e\n\u003cp\u003e\u003cstrong\u003e\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eData Analysis, Reliability, and Validity\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eDuring the phase of data analysis, we adhered to established guidelines\u003csup\u003e19,33,36\u003c/sup\u003e. Internal reliability was first evaluated for the full scale and each construct, with Cronbach\u0026rsquo;s alpha coefficients required to exceed the threshold of 0.70. Sampling adequacy was then assessed using the Kaiser-Meyer-Olkin (KMO) statistic, with values above 0.60 considered acceptable.\u003c/p\u003e\n\u003cp\u003eIn the subsequent step, exploratory factor analysis (EFA) was carried out through the use of principal component analysis, wherein Promax rotation and Kaiser normalization were implemented. Items were retained only when communalities were greater than 0.50 and factor loadings exceeded 0.50, with each item correctly associated with its intended construct\u003csup\u003e19,33\u003c/sup\u003e. For example, items designed to measure Goal Alignment were expected to load most strongly on the goal alignment factor and cluster consistently with other goal alignment items.\u003c/p\u003e\n\u003cp\u003eFurthermore, average variance extracted (AVE) and composite reliability (CR) were assessed for each construct, with AVE values above 0.50 and CR values required to exceed 0.70, thereby providing evidence of convergent validity. Discriminant validity was assessed following the Fornell \u0026amp; Larcker criterion\u003csup\u003e48\u003c/sup\u003e, which requires that the square root of AVE for each construct exceed its correlations with other constructs. Subsequently, confirmatory factor analysis (CFA) was performed to test the measurement model. Model adequacy and stability were examined through key goodness-of-fit indices, including the Chi-square statistic, RMSEA, and CFI.\u003c/p\u003e"},{"header":"Declarations","content":"\u003cp\u003e\u003cstrong\u003eData availability\u003c/strong\u003e\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eThe original contributions presented in the study are included in the article/Supplementary Materials. Further inquiries can be directed to the corresponding author.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAcknowledgements\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eNot applicable.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAuthor contributions\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eAll authors contributed to the study conception and design. Writing - original draft preparation: Ning Mi, Cunying Fan; Writing - review and editing: Ning Mi; Conceptualization: Cunying Fan; Methodology: Ning Mi, Cunying Fan; Formal analysis and investigation: Ning Mi, Cunying Fan; Funding acquisition: Cunying Fan; Resources: Cunying Fan; Supervision: Ning Mi, and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eFunding sources\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was supported by the Shandong Social Sciences Planning Research Project of China (grant number 23CSDJ24).\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCompeting interests\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThe authors declare no competing interests.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eEthics approval and consent to participate\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003eThis study was reviewed and approved by\u0026nbsp;the Institutional Review Board of Qufu Normal University (Approval No. 2025-106). All participants were informed about the purpose, procedures, and anonymity of the survey before participation. Completion of the online questionnaire was considered as provision of informed consent. The study was conducted in accordance with institutional guidelines and relevant national regulations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eAdditional information\u0026nbsp;\u003c/strong\u003e\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eSupplementary Information\u003c/strong\u003e The online version contains supplementary material available at https://doi.org/10.5281/zenodo.17432622\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eCorrespondence\u003c/strong\u003e and requests for materials should be addressed to N.M.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eReprints and permissions information\u003c/strong\u003e is available at www.nature.com/reprints.\u0026nbsp;\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003ePublisher\u0026rsquo; s note\u003c/strong\u003e Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.\u003c/p\u003e\n\u003cp\u003e\u003cstrong\u003eOpen Access\u003c/strong\u003e This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and\u0026nbsp;\u003c/p\u003e\n\u003cp\u003eindicate if changes were made. The images or other third party material in this article are included in the article\u0026rsquo; s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article\u0026rsquo;s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.\u003c/p\u003e"},{"header":"References","content":"\u003col\u003e\n\u003cli\u003eLi, J., Link, S., \u0026amp; Hegelheimer, V. Rethinking the Role of Automated Writing Evaluation (AWE) Feedback in ESL Writing Instruction. \u003cem\u003eJ. Second Lang. Writ.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 1-18. https://doi.org/10.1016/j.jslw.2014.10.004 (2015).\u003c/li\u003e\n\u003cli\u003eRanalli, J. Automated written corrective feedback: how well can students make use of it? \u003cem\u003eComput. Assist. Lang. Learn.\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 653-674. https://doi.org/10.1080/09588221.2018.1428994 (2018).\u003c/li\u003e\n\u003cli\u003eLee, J. D., \u0026amp; See, K. A. Trust in automation: Designing for appropriate reliance. \u003cem\u003eHum. Factors\u003c/em\u003e \u003cstrong\u003e46\u003c/strong\u003e, 50-80. https://doi.org/10.1518/hfes.46.1.50_30392 (2004).\u003c/li\u003e\n\u003cli\u003eHoff, K. A., \u0026amp; Bashir, M. Trust in Automation: Integrating Empirical Evidence on Factors That Influence Trust. \u003cem\u003eHum. Factors\u003c/em\u003e \u003cstrong\u003e57\u003c/strong\u003e, 407-434. https://doi.org/10.1177/0018720814547570 (2015).\u003c/li\u003e\n\u003cli\u003eChiu, T. K. F., \u0026amp; Chai, C. Sustainable Curriculum Planning for Artificial Intelligence Education: A Self-Determination Theory Perspective. \u003cem\u003eSustainability\u003c/em\u003e \u003cstrong\u003e12\u003c/strong\u003e, 55-68. https://doi.org/10.3390/su12145568 (2020).\u003c/li\u003e\n\u003cli\u003eMayer, R. C., Davis, J. H., \u0026amp; Schoorman, F. D. An integrative model of organizational trust. \u003cem\u003eAcad. Manag. Rev.\u003c/em\u003e \u003cstrong\u003e20\u003c/strong\u003e, 709-734. https://doi.org/10.5465/amr.1995.9508080335 (1995).\u003c/li\u003e\n\u003cli\u003eChoi, J. K., \u0026amp; Ji, Y. G. Investigating the importance of trust on adopting an autonomous vehicle. \u003cem\u003eInt. J. Hum.-Comput. Interact.\u003c/em\u003e \u003cstrong\u003e31\u003c/strong\u003e, 692-702. https://doi.org/10.1080/10447318.2015.1070549 (2015).\u003c/li\u003e\n\u003cli\u003eWojton, H. M., Porter, D., Lane, S. T., Bieber, C., \u0026amp; Madhavan, P. Initial validation of the trust of automated systems test (TOAST). \u003cem\u003eJ. Soc. Psychol.\u003c/em\u003e \u003cstrong\u003e160\u003c/strong\u003e, 735-750. https://doi.org/10.1080/00224545.2020.1749020 (2020).\u003c/li\u003e\n\u003cli\u003eHendriks, F., Kienhues, D., \u0026amp; Bromme, R. Measuring laypeople\u0026rsquo;s trust in experts in a digital age: The Muenster Epistemic Trustworthiness Inventory (METI). \u003cem\u003ePLoS ONE\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, e0139309. https://doi.org/10.1371/journal.pone.0139309 (2015).\u003c/li\u003e\n\u003cli\u003eMalle, B. F., \u0026amp; Ullman, D. A multi-dimensional conception and measure of human-robot trust. In Nam, C. S., \u0026amp; Lyons, J. B. (eds.), \u003cem\u003eTrust in Human-Robot Interaction: Research and Applications\u003c/em\u003e, 3-25. Elsevier (2021).\u003c/li\u003e\n\u003cli\u003eMcKnight, D. H., Choudhury, V., \u0026amp; Kacmar, C. Developing and validating trust measures for e-commerce: An integrative typology. \u003cem\u003eInf. Syst. Res.\u003c/em\u003e \u003cstrong\u003e13\u003c/strong\u003e, 334-359. https://doi.org/10.1287/isre.13.3.334.81 (2002).\u003c/li\u003e\n\u003cli\u003eGlikson, E., \u0026amp; Woolley, A. W. Human Trust in Artificial Intelligence: Review of empirical research. \u003cem\u003eAcad. Manag. Ann.\u003c/em\u003e \u003cstrong\u003e14\u003c/strong\u003e, 627-660 (2020).\u003c/li\u003e\n\u003cli\u003eNazaretsky, T., Ariely, M., Cukurova, M., \u0026amp; Alexandron, G. Teachers\u0026rsquo; trust in AI-powered educational technology and a professional development program to improve it. \u003cem\u003eBr. J. Educ. Technol.\u003c/em\u003e \u003cstrong\u003e53\u003c/strong\u003e, 914-931. https://doi.org/10.1111/bjet.13232 (2022a).\u003c/li\u003e\n\u003cli\u003eNazaretsky, T., Mejia-Domenzain, P., Swamy, V., Frej, J., \u0026amp; K\u0026auml;ser, T. The critical role of trust in adopting AI-powered educational technology for learning: An instrument for measuring student perceptions. \u003cem\u003eComput. Educ.: Artif. Intell.\u003c/em\u003e \u003cstrong\u003e8\u003c/strong\u003e, 100368. https://doi.org/10.1016/j.caeai.2025.100368 (2025).\u003c/li\u003e\n\u003cli\u003eJian, J. Y., Bisantz, A. M., \u0026amp; Drury, C. G. Foundations for an empirically determined scale of trust in automated systems. \u003cem\u003eInt. J. Cogn. Ergon.\u003c/em\u003e \u003cstrong\u003e4\u003c/strong\u003e, 53-71. https://doi.org/10.1207/S15327566IJCE0401_04 (2000).\u003c/li\u003e\n\u003cli\u003eClark, M., \u0026amp; Payne, R. Character-based determinants of trust in leaders. \u003cem\u003eRisk Anal.\u003c/em\u003e \u003cstrong\u003e26\u003c/strong\u003e, 1161-1173. https://doi.org/10.1111/j.1539-6924.2006.00823.x (2006).\u003c/li\u003e\n\u003cli\u003eKelley, T. L. The selection of upper and lower groups for the validation of test items. \u003cem\u003eJ. Educ. Psychol.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 17-24. https://doi.org/10.1037/h0057123 (1939).\u003c/li\u003e\n\u003cli\u003eNunnally, J. C., \u0026amp; Bernstein, I. H. \u003cem\u003ePsychometric Theory\u003c/em\u003e (3rd ed.). New York: McGraw-Hill (1994).\u003c/li\u003e\n\u003cli\u003eDeVellis, R. F., \u0026amp; Thorpe, C. T. \u003cem\u003eScale Development: Theory and Applications\u003c/em\u003e. Sage Publications (2021).\u003c/li\u003e\n\u003cli\u003eHair, J. F., Babin, B. J., Anderson, R. E., \u0026amp; Black, W. C. \u003cem\u003eMultivariate Data Analysis\u003c/em\u003e (8th ed.). Cengage Learning (2019).\u003c/li\u003e\n\u003cli\u003eCronbach, L. J. Coefficient alpha and the internal structure of tests. \u003cem\u003ePsychometrika\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 297-334 (1951).\u003c/li\u003e\n\u003cli\u003eHu, L., \u0026amp; Bentler, P. M. Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. \u003cem\u003eStruct. Equ. Model.\u003c/em\u003e \u003cstrong\u003e6\u003c/strong\u003e, 1-55. https://doi.org/10.1080/10705519909540118 (1999).\u003c/li\u003e\n\u003cli\u003eLam, L. W. Impact of competitiveness on salespeople\u0026rsquo;s commitment and performance. \u003cem\u003eJ. Bus. Res.\u003c/em\u003e \u003cstrong\u003e65\u003c/strong\u003e, 1328-1334. https://doi.org/10.1016/j.jbusres.2011.10.026 (2012).\u003c/li\u003e\n\u003cli\u003eFornell, C., \u0026amp; Larcker, D. F. Evaluating Structural Equation Models with Unobservable Variables and Measurement Error. \u003cem\u003eJ. Mark. Res.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 39-50. https://doi.org/10.1177/002224378101800104 (1981).\u003c/li\u003e\n\u003cli\u003eGabriel, I. Artificial intelligence, values, and alignment. \u003cem\u003eMinds Mach.\u003c/em\u003e \u003cstrong\u003e30\u003c/strong\u003e, 411-437. https://doi.org/10.1007/s11023-020-09539-2 (2020).\u003c/li\u003e\n\u003cli\u003eMechergui, M., \u0026amp; Sreedharan, S. Goal alignment: Reanalyzing value alignment problems using human-aware AI. \u003cem\u003eProc. AAAI Conf. Artif. Intell.\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 14393-14401. https://doi.org/10.1609/aaai.v38i13.29371 (2024).\u003c/li\u003e\n\u003cli\u003eShen, H., Knearem, T., Ghosh, R., Alkiek, K., Krishna, K., Liu, Y., \u0026amp; Jurgens, D. Towards bidirectional human-AI alignment: A systematic review for clarifications, framework, and future directions. \u003cem\u003earXiv preprint arXiv:2406.09264\u003c/em\u003e (2024).\u003c/li\u003e\n\u003cli\u003eFeldman-Maggor, Y., Cukurova, M., Kent, C., et al. The Impact of Explainable AI on Teachers\u0026rsquo; Trust and Acceptance of AI EdTech Recommendations: The Power of Domain-specific Explanations. \u003cem\u003eInt. J. Artif. Intell. Educ.\u003c/em\u003e (2025). https://doi.org/10.1007/s40593-025-00486-6\u003c/li\u003e\n\u003cli\u003eHuang, K., \u0026amp; Ball, C. The Influence of AI Literacy on User\u0026apos;s Trust in AI in Practical Scenarios: A Digital Divide Pilot Study. \u003cem\u003eProc. Assoc. Inf. Sci. Technol.\u003c/em\u003e \u003cstrong\u003e61\u003c/strong\u003e, 937-939. https://doi.org/10.1002/pra2.1146 (2024).\u003c/li\u003e\n\u003cli\u003eĐerić, E., Frank, D., \u0026amp; Milković, M. Trust in Generative AI Tools: A Comparative Study of Higher Education Students, Teachers, and Researchers. \u003cem\u003eInformation\u003c/em\u003e \u003cstrong\u003e16\u003c/strong\u003e, 622. https://doi.org/10.3390/info16070622 (2025).\u003c/li\u003e\n\u003cli\u003ePratiwi, S., Andayani, A., Winarni, R., \u0026amp; Sumarwati, S. Utilization of artificial intelligence technology in an academic writing class: How do Indonesian students perceive? \u003cem\u003eContemp. Educ. Technol.\u003c/em\u003e \u003cstrong\u003e15\u003c/strong\u003e, ep450. https://doi.org/10.30935/cedtech/13419 (2023).\u003c/li\u003e\n\u003cli\u003ePitts, G., Rani, N., Mildort, W., \u0026amp; Cook, E. M. Students\u0026rsquo; reliance on AI in higher education: Identifying contributing factors. \u003cem\u003earXiv: 2506.13845v1\u003c/em\u003e (2025).\u003c/li\u003e\n\u003cli\u003eLuo, Z. Factors contributing to teachers\u0026rsquo; acceptance intention to gamified EFL tools: A scale development study. \u003cem\u003eEduc. Technol. Res. Dev.\u003c/em\u003e \u003cstrong\u003e72\u003c/strong\u003e, 447-477 (2024).\u003c/li\u003e\n\u003cli\u003eAmoozadeh, M., et al. Trust in Generative AI among Students: An exploratory study. \u003cem\u003eProc. 55th ACM Tech. Symp. Comput. Sci. Educ.\u003c/em\u003e \u003cstrong\u003e1\u003c/strong\u003e, 67-73. https://doi.org/10.1145/3626252.3630842 (2024).\u003c/li\u003e\n\u003cli\u003eLuo, Z. Determinants of the perceived usefulness (PU) in the context of using gamification for classroom-based ESL teaching: A scale development study. \u003cem\u003eEduc. Inf. Technol.\u003c/em\u003e \u003cstrong\u003e28\u003c/strong\u003e, 4741-4768 (2023a).\u003c/li\u003e\n\u003cli\u003eHuang, J. L., Curran, P. G., Keeney, J., et al. Detecting and deterring insufficient effort responding to surveys. \u003cem\u003eJ. Bus. Psychol.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 99-114. https://doi.org/10.1007/s10869-011-9231-8 (2012).\u003c/li\u003e\n\u003cli\u003eCaldwell, C., \u0026amp; Hansen, M. H. Trustworthiness, governance, and wealth creation. \u003cem\u003eJ. Bus. Ethics\u003c/em\u003e \u003cstrong\u003e97\u003c/strong\u003e, 173-188 (2010).\u003c/li\u003e\n\u003cli\u003eChoudhury, A., \u0026amp; Shamszare, H. Investigating the impact of user trust on the adoption and use of ChatGPT: Survey analysis. \u003cem\u003eJ. Med. Internet Res.\u003c/em\u003e \u003cstrong\u003e25\u003c/strong\u003e, e47184. https://doi.org/10.2196/47184 (2023).\u003c/li\u003e\n\u003cli\u003eShin, D. The effects of explainability and causability on perception, trust, and acceptance: Implications for explainable AI. \u003cem\u003eInt. J. Hum.-Comput. Stud.\u003c/em\u003e \u003cstrong\u003e146\u003c/strong\u003e, 102551. https://doi.org/10.1016/j.ijhcs.2020.102551 (2021).\u003c/li\u003e\n\u003cli\u003eLalot, F., \u0026amp; Bertram, A. When the bot walks the talk: Investigating the foundations of trust in an artificial intelligence (AI) chatbot. \u003cem\u003eJ. Exp. Psychol. Gen.\u003c/em\u003e \u003cstrong\u003e154\u003c/strong\u003e, 533-551. https://doi.org/10.1037/xge0001696 (2025).\u003c/li\u003e\n\u003cli\u003eVice, J., \u0026amp; Khan, M. M. Toward accountable and explainable artificial intelligence part two: The framework implementation. \u003cem\u003eIEEE Access\u003c/em\u003e \u003cstrong\u003e10\u003c/strong\u003e, 36091-36105. https://doi.org/10.1109/ACCESS.2022.3163523 (2022).\u003c/li\u003e\n\u003cli\u003eStanley, J. C., \u0026amp; Dorton, S. L. Exploring trust with the AI incident database. \u003cem\u003eProc. Hum. Factors Ergon. Soc. Annu. Meet.\u003c/em\u003e \u003cstrong\u003e67\u003c/strong\u003e, 489-494 (2023).\u003c/li\u003e\n\u003cli\u003eKaplan, A. D., Kessler, T. T., Brill, J. C., \u0026amp; Hancock, P. A. Trust in artificial intelligence: Meta-analytic findings. \u003cem\u003eHum. Factors\u003c/em\u003e \u003cstrong\u003e65\u003c/strong\u003e, 337-359 (2023).\u003c/li\u003e\n\u003cli\u003eMitchell, T. Trust and transparency in artificial intelligence. \u003cem\u003ePhilos. Technol.\u003c/em\u003e \u003cstrong\u003e38\u003c/strong\u003e, 87. https://doi.org/10.1007/s13347-025-00916-2 (2025).\u003c/li\u003e\n\u003cli\u003eKizilcec, R. F. How much information? Effects of transparency on trust in an algorithmic interface. In \u003cem\u003eProc. 2016 CHI Conf. Hum. Factors Comput. Syst.\u003c/em\u003e, 2390-2395. New York, USA (2016).\u003c/li\u003e\n\u003cli\u003eDimitriadis, S., \u0026amp; Kyrezis, N. Linking trust to use intention for technology-enabled bank channels: The role of trusting intentions. \u003cem\u003ePsychol. Mark.\u003c/em\u003e \u003cstrong\u003e27\u003c/strong\u003e, 799-820. https://doi.org/10.1002/mar.20358 (2010).\u003c/li\u003e\n\u003cli\u003eWorthington, R. L., \u0026amp; Whittaker, T. A. Scale development research: A content analysis and recommendations for best practices. \u003cem\u003eCouns. Psychol.\u003c/em\u003e \u003cstrong\u003e34\u003c/strong\u003e, 806-838. https://doi.org/10.1177/0011000006288127 (2006).\u003c/li\u003e\n\u003cli\u003eFornell, C., \u0026amp; Larcker, D. F. Evaluating structural equation models with unobservable variables and measurement error. \u003cem\u003eJ. Mark. Res.\u003c/em\u003e \u003cstrong\u003e18\u003c/strong\u003e, 39-50 (1981).\u003c/li\u003e\n\u003c/ol\u003e"}],"fulltextSource":"","fullText":"","funders":[],"hasAdminPriorityOnWorkflow":false,"hasManuscriptDocX":true,"hasOptedInToPreprint":true,"hasPassedJournalQc":"","hasAnyPriority":true,"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":"trust, AI-assisted English writing, Chinese University students, reliability, validity, scale development","lastPublishedDoi":"10.21203/rs.3.rs-7939708/v1","lastPublishedDoiUrl":"https://doi.org/10.21203/rs.3.rs-7939708/v1","license":{"name":"CC BY 4.0","url":"https://creativecommons.org/licenses/by/4.0/"},"manuscriptAbstract":"The purpose of this study is to develop and validate the Trust Scale for AI-assisted English writing targeted at Chinese university students (TS-AIEW). In accordance with well-established scientific protocols, we developed the preliminary version of the TS-AIEW and validated its performance by leveraging date derived from 527 valid survey questionnaires completed by Chinese university students. This dataset was split into two distinct subsamples: 257 responses were designated for the implementation of exploratory factor analysis (EFA), while the remaining 270 were allocated to conduct confirmatory factor analysis (CFA). During the EFA and CFA procedures, a total of 8 items were eliminated on account of low item-total correlation coefficients, values of corrected item-total correlations and factor loadings. The finalized TS-AIEW, consisting of 20 items, comprises four distinct dimensions: goal alignment, transparency, competency, and trust intention. For future research, scholars may opt to directly adopt this validated TS-AIEW or assess its applicability and validity across varied demographic groups and research contexts.","manuscriptTitle":"Development and Validation of a Scale to Assess Trust in AI-Assisted English Writing Among Chinese University Students","msid":"","msnumber":"","nonDraftVersions":[{"code":1,"date":"2026-03-10 03:58:02","doi":"10.21203/rs.3.rs-7939708/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":"3d6619bc-ec73-430b-96e2-26686dbbbbc4","owner":[],"postedDate":"March 10th, 2026","published":true,"recentEditorialEvents":[],"rejectedJournal":[],"revision":"","amendment":"","status":"posted","subjectAreas":[{"id":62734954,"name":"Business and commerce/Information systems and information technology"},{"id":62734955,"name":"Biological sciences/Psychology"},{"id":62734956,"name":"Social science/Psychology"}],"tags":[],"updatedAt":"2026-03-10T03:58:02+00:00","versionOfRecord":[],"versionCreatedAt":"2026-03-10 03:58:02","video":"","vorDoi":"","vorDoiUrl":"","workflowStages":[]},"version":"v1","identity":"rs-7939708","journalConfig":"researchsquare"},"__N_SSP":true},"page":"/article/[identity]/[[...version]]","query":{"redirect":"/article/rs-7939708","identity":"rs-7939708","version":["v1"]},"buildId":"XKTyCvWXoU3ODBz1xrDgd","isFallback":false,"isExperimentalCompile":false,"dynamicIds":[84888],"gssp":true,"scriptLoader":[]}