Review Reports

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This paper explores the factors that influence teachers' intention to continue using artificial intelligence (AI) after its initial adoption. The study, which involved 549 Chinese participants, integrates the Technology Acceptance Model (TAM) and the Unified Theory of Acceptance and Use of Technology (UTAUT) to examine psychological, contextual, and perceptual variables. The methodology employed is particularly robust, combining structural equation modeling (SEM) with fuzzy-set qualitative comparative analysis (fsQCA), thus enabling the identification of multiple paths towards sustainable use of AI. This article addresses a current issue by shifting the focus from the initial acceptance of AI to its long-term integration into educational contexts. Its originality lies in going beyond traditional linear analyzes to provide a more nuanced understanding of human behavior through fsQCA. However, some methodological and terminological clarifications are necessary before publication.

Comments for author File: Comments.pdf

Author Response

Comment 1: The reference (Ng'ambi, Brown, Bozalek, Gachago, & Wood, 2016) is cited in the context of using ChatGPT by novice teachers for lesson planning, classroom management, and assignment grading. However, it is important to note that this research was published several years before ChatGPT was available.

Respond 1: We are grateful for the reviewer's careful reading on this point. In the original manuscript. In the revised version, we have replaced this reference to support a general claim about technology-enhanced teaching and learning in higher education, which accurately reflects the scope of that study. The specific claim regarding generative AI tools has been revised and supported with a more appropriate recent reference (van den Berg & du Plessis, 2023). These changes are highlighted in red in Section 1.1.

Comment 2: Additionally, while the title and abstract refer to "in-service teachers," other sections of the paper describe participants as "future professionals," "trainees," or "university students." The authors should clarify whether the sample consists of tenured teachers or pre-service teachers, as this distinction is crucial for interpreting the findings related to practical experience. Notably, the sample included 100 participants aged 18–20.

Respond 2: We thank the reviewer for this comment. We would like to clarify that terms such as "in-service teachers," "future professionals," "trainees," and "university students" are not used in our manuscript to describe participants. Throughout the title and abstract, we consistently use "pre-service teachers." The words "trainees" and "university students" only appear within the titles of cited references (References [6] and [16]), which we cannot modify. The phrases "future practitioners" and "future teachers" in Section 1.1 are commonly used expressions to describe the career path of pre-service teachers, not as participant labels.

Regarding participants aged 18–20, in China's teacher education system, teaching placements usually start in the second year. Therefore, these participants already had hands-on classroom experience. We have added a clarifying sentence in Section 3.2 to address this point (highlighted in red).

Comment 3: Furthermore, although the authors mention that infrastructure alone is insufficient to convince teachers of the pedagogical value of AI, the discussion should provide a deeper analysis of why the UTAUT model does not support this point specifically in the Chinese context.

Respond 3: We appreciate this insightful comment. In the revised manuscript, we have expanded the discussion of the non-significant Facilitating Condition to Perceived of usefulness path by providing a context-specific analysis. Specifically, we explain that current facilitating conditions in Chinese teacher education programs primarily emphasize technical access rather than pedagogical application, and that China's collectivist culture leads pre-service teachers to form beliefs about usefulness more through social recommendations than resource availability. This interpretation is empirically supported by the significant social influence to perceived usefulness path in our model. Please see Section 5, paragraph 3 for the revised text.

 

Comment 4: Lastly, the authors should explicitly discuss how social desirability bias may have influenced self-assessed responses, particularly in a culture that emphasizes technological innovation.

Respond 4: We recognize that social desirability bias may have inflated participants' self-reported attitudes, particularly within a cultural context that values technological innovation. This limitation has been explicitly acknowledged in the revised manuscript (see Conclusion, p. 19).  Reference (Graf-Vlachy,2018) has been cited accordingly.

Reviewer 2 Report

Comments and Suggestions for Authors

The manuscript is a reasonable contribution to a growing literature, but it needs tightening — both in terms of internal consistency and in the depth of its interpretive discussion — before it is ready for publication. The authors position the Expectation Confirmation Model (ECM) as a central theoretical lens for understanding continuance intention, devoting considerable space to it in Section 2.1. Yet the research model presented in Figure 1 includes neither confirmation nor satisfaction — the two constructs that give ECM its distinctive explanatory logic. What remains is essentially a TAM-UTAUT extension, with ECM functioning more as rhetorical backdrop than as a structural contributor. The authors should either incorporate these core ECM constructs into the measurement instrument, or scale back the theoretical claims made for ECM and clarify its actual role in the study.

As reported in Table 10, configurations C1, C2, and C3 carry unique coverage values of 0.010, 0.014, and 0.025 respectively — figures low enough to question how much independent explanatory work each configuration is actually doing. Grouping all three under a single "intrinsic-value driven" pattern without acknowledging their substantial overlap, or discussing what distinguishes them from one another, weakens the analytical contribution. The prevalence of "don't care" conditions across multiple configurations compounds this concern. A more critical reading of what the configurations do and do not tell us would considerably strengthen this section.

A minor but telling detail: Section 3.3 reports the use of IBM SPSS 26.0, while Section 6 refers to SPSS 27.0. This discrepancy suggests the manuscript was not carefully proofread before submission. On a more substantive note, continuance intention is by definition a longitudinal construct — it concerns what users choose to do after sustained experience with a technology. The authors acknowledge the single time-point limitation briefly in the conclusion, but do not reflect on how this might affect the validity of their inferences. Given that Bhattacherjee (2001), whom the authors cite, explicitly frames continuance as a post-adoptive process, this tension deserves more than a passing mention.

Author Response

Comment 1: The authors position the Expectation Confirmation Model (ECM) as a central theoretical lens for understanding continuance intention, devoting considerable space to it in Section 2.1. Yet the research model presented in Figure 1 includes neither confirmation nor satisfaction — the two constructs that give ECM its distinctive explanatory logic. What remains is essentially a TAM-UTAUT extension, with ECM functioning more as rhetorical backdrop than as a structural contributor. The authors should either incorporate these core ECM constructs into the measurement instrument, or scale back the theoretical claims made for ECM and clarify its actual role in the study.

Respond 1: We appreciate this observation and have reworded section 2.1 to more accurately reflect how we have included aspects of ECM into the model

 

Comment 2: As reported in Table 10, configurations C1, C2, and C3 carry unique coverage values of 0.010, 0.014, and 0.025 respectively — figures low enough to question how much independent explanatory work each configuration is actually doing. Grouping all three under a single "intrinsic-value driven" pattern without acknowledging their substantial overlap, or discussing what distinguishes them from one another, weakens the analytical contribution.

Respond 2: In the revised manuscript, we have distinguished C1, C2, and C3 by identifying their shared core conditions (low GAI anxiety, high AI self-efficacy, high perceived usefulness) and their different contextual drivers (facilitating conditions in C1, social influence in C2, perceived ease of use in C3). We have also acknowledged that the low unique coverage (0.010–0.025) reflects case overlap, while the moderate raw coverage (0.481–0.521) confirms each pathway's relevance. Please see revised Section 4.4.

 

Comment 3: The prevalence of "don't care" conditions across multiple configurations compounds this concern. A more critical reading of what the configurations do and do not tell us would considerably strengthen this section.

Respond 3: We appreciate this comment. In the revised manuscript, we have clarified that "don't care" conditions do not indicate irrelevance. Rather, within each specific configuration, these conditions are neither necessary nor negating, as the other present conditions are already sufficient to produce high continuance intention. This interpretation is consistent with the configurational logic of fsQCA (Ragin, 2008). Please see revised Section 4.4.

 

Comment 4: Section 3.3 reports the use of IBM SPSS 26.0, while Section 6 refers to SPSS 27.0. This discrepancy suggests the manuscript was not carefully proofread before submission.

Respond 4: Thank you for pointing out this inconsistency. We have carefully checked the manuscript and confirmed that the correct software version is IBM SPSS Statistics 27.0. We have revised the relevant text in Section 3.3 and Section 6, and the software version has now been consistently reported as SPSS 27.0 throughout the manuscript. All corresponding changes have been marked in red.

 

Comment 5: Continuance intention is by definition a longitudinal construct — it concerns what users choose to do after sustained experience with a technology. The authors acknowledge the single time-point limitation briefly in the conclusion, but do not reflect on how this might affect the validity of their inferences. Given that Bhattacherjee (2001), whom the authors cite, explicitly frames continuance as a post-adoptive process, this tension deserves more than a passing mention.

Respond 5: We acknowledge that continuance intention, as conceptualised by Bhattacherjee (2001), is a post-adoptive construct, and a longitudinal design would better capture its temporal dynamics. However, our study measures continuance intention (i.e., stated willingness to continue using the technology) rather than actual behaviour. This approach is consistent with established cross-sectional research in the field (e.g., Yang et al., 2023; Abdullahi et al., 2021). We have expanded the limitations section accordingly to discuss this constraint more explicitly and recommend longitudinal designs for future research.

Reviewer 3 Report

Comments and Suggestions for Authors

This manuscript addresses a timely topic (pre-service teachers’ continuance intention to use generative AI) and uses a sizable sample (N=549) with complementary analyses (SEM and fsQCA). The overall model is reasonable and the results are presented in a structured manner. However, I recommend major revisions before the manuscript can be considered further, due to important reporting inconsistencies, citation accuracy concerns, and English-language clarity issues. 

The manuscript states that the questionnaire “did not collect personal demographic information such as age or gender,” yet Table 2 reports both gender and age categories. Please correct and clarify exactly what demographic variables were collected and how anonymity was preserved.  The methods describe a “random sampling approach” via an electronically distributed survey. Please provide sufficient detail to justify the claim of random sampling (sampling frame, recruitment process, institutions/regions, and how randomness was implemented). If the sampling was convenience/voluntary response, please state this explicitly and discuss implications for generalizability. The manuscript mixes terminology (e.g., discussion of “PLS-SEM” while the analysis is described with SPSS/AMOS SEM). Please clearly specify the analytic approach (CB-SEM with AMOS vs PLS-SEM), ensure consistency throughout, and standardize software/version reporting (e.g., SPSS is reported differently in different sections).  In the introduction, a key claim about ChatGPT being widely used by novice teachers is supported by Ng’ambi et al. (2016). Please verify that the cited source actually supports the specific claim about generative AI/ChatGPT; if not, replace with appropriate, recent references and ensure all citations accurately reflect the statements they support. Hypotheses are listed, but explicit research questions are not clearly stated. Consider adding 1–3 research questions at the end of the introduction to strengthen alignment between aims, hypotheses, and analyses. Also standardize naming/abbreviations of the dependent construct (e.g., “teaching continuous intention (TCI)” vs “AI continuous intention (AI-CI)”) and use one label consistently. Significant editing is needed for grammar, phrasing, capitalization, and formatting (including spacing after punctuation and clearer headings). Where causal language is used (e.g., “effects”), consider cautious wording given the cross-sectional design (you already note key limitations—please ensure the narrative aligns with them).

Comments on the Quality of English Language

Significant editing is needed for grammar, phrasing, capitalization, and formatting (including spacing after punctuation and clearer headings). Where causal language is used (e.g., “effects”), consider cautious wording given the cross-sectional design (you already note key limitations—please ensure the narrative aligns with them).

Author Response

Comment 1: The manuscript states that the questionnaire "did not collect personal demographic information such as age or gender," yet Table 2 reports both gender and age categories. Please correct and clarify exactly what demographic variables were collected and how anonymity was preserved.

Respond 1: Thank you for this important comment. We agree that the original wording was inaccurate and may have caused confusion. We have revised the relevant text to clarify that the questionnaire collected basic demographic information, including gender and age, as reported in Table 2, while no personally identifiable information was obtained. We have also clarified that the questionnaire was completed anonymously and that all responses were used solely for research purposes. The corresponding changes have been marked in red.

 

Comment 2: The methods describe a "random sampling approach" via an electronically distributed survey. Please provide sufficient detail to justify the claim of random sampling (sampling frame, recruitment process, institutions/regions, and how randomness was implemented). If the sampling was convenience/voluntary response, please state this explicitly and discuss implications for generalizability.

Respond 2: We acknowledge that the original description was inaccurate. The sampling method has been corrected from "random sampling" to "convenience sampling" in Section 3.2, with added details on recruitment. The implications for generalizability have also been addressed in Section 6.

 

Comment 3: The manuscript mixes terminology (e.g., discussion of "PLS-SEM" while the analysis is described with SPSS/AMOS SEM). Please clearly specify the analytic approach (CB-SEM with AMOS vs PLS-SEM), ensure consistency throughout, and standardize software/version reporting (e.g., SPSS is reported differently in different sections).

Respond 3: We sincerely thank the reviewer for this important observation. Upon careful review, we acknowledge the terminological inconsistency. To clarify: the present study employed covariance-based structural equation modeling (CB-SEM) using IBM SPSS AMOS 26.0. IBM SPSS Statistics 27.0 was used for descriptive analysis, reliability testing, and exploratory factor analysis. The inadvertent reference to PLS-SEM has been removed from the revised manuscript.

 

Comment 4: In the introduction, a key claim about ChatGPT being widely used by novice teachers is supported by Ng'ambi et al. (2016). Please verify that the cited source actually supports the specific claim about generative AI/ChatGPT; if not, replace with appropriate, recent references and ensure all citations accurately reflect the statements they support.

Respond 4: We are grateful for the reviewer's careful reading on this point. In the original manuscript. In the revised version, we have replaced this reference to support a general claim about technology-enhanced teaching and learning in higher education, which accurately reflects the scope of that study. The specific claim regarding generative AI tools has been revised and supported with a more appropriate recent reference (van den Berg & du Plessis, 2023). These changes are highlighted in red in Section 1.1.

 

Comment 5: Hypotheses are listed, but explicit research questions are not clearly stated. Consider adding 1–3 research questions at the end of the introduction to strengthen alignment between aims, hypotheses, and analyses.

Respond 5: We thank the reviewer for this suggestion. We have added a research question at the end of Section 1.3 (Lines 116–118): "What factors influence Chinese pre-service teachers' intentions to use GAI?" This overarching question encompasses hypotheses H1–H7, which examine how perceptual, psychological, and contextual factors shape continuance intention.

 

Comment 6: Standardize naming/abbreviations of the dependent construct (e.g., "teaching continuous intention (TCI)" vs "AI continuous intention (AI-CI)") and use one label consistently.

Respond 6: Upon careful revision of the manuscript, we confirm that the dependent construct investigated in this study is GAI continuous intention (GAI-CI). To avoid ambiguity, we have standardized the terminology throughout the manuscript. In particular, the inconsistent labels in Section 3 (Methods), especially Section 3.1, and Section 4 (Findings), especially Section 4.1, have been revised to AI continuous intention (AI-CI). All corresponding changes have been marked in red.

 

Comment 7: Significant editing is needed for grammar, phrasing, capitalization, and formatting (including spacing after punctuation and clearer headings).

Respond 7: Thank you – co-authors with English as a first language have read through and edited and corrected.

 

Comment 8: Where causal language is used (e.g., "effects"), consider cautious wording given the cross-sectional design (you already note key limitations — please ensure the narrative aligns with them).

Respond 8: Thank you – co-authors with English as a first language have read through and edited and corrected.

 

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

/

Comments on the Quality of English Language

Significant editing is needed for grammar, phrasing, capitalization, and formatting (including spacing after punctuation and clearer headings). Where causal language is used (e.g., “effects”), consider cautious wording given the cross-sectional design (you already note key limitations—please ensure the narrative aligns with them).

Author Response

Comment: Significant editing is needed for grammar, phrasing, capitalization, and formatting (including spacing after punctuation and clearer headings). Where causal language is used (e.g., “effects”), consider cautious wording given the cross-sectional design (you already note key limitations—please ensure the narrative aligns with them).

Author response: Revision has been made as suggested. Please kindly check the attached manuscript for details. Thank you very much.