Navigating the Dual-View Phenomenon: Social Ambivalence, Ambivalence Literacy, and Lecturer Role Transformation in AI-Integrated Transnational STEM Education
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsI would like to thank the editorial board for giving me the opportunity to review this very interesting and timely research.
I would also like to congratulate the authors for writing this article, which is easy to understand and read.
This research deals with the use of AI chatbots by STEM undergraduates across four transnational education programs. A survey of 467 students was conducted, allowing for a quantitative approach, without forgetting the analysis of the open-ended question at the end of the questionnaire. The results show a robust dual-view phenomenon in social perceptions on the part of students, which strongly impacts their expectations regarding the changing role of teachers.
Below, I share my questions and concerns in order to further improve the fluidity and relevance of the research presented.
In the introduction, line 108: the contributions are specified; they would benefit from being specified at the end of the article, as well as line 125: the research questions are summarized. They would also benefit from being presented at the end of the literature review.
Line 302: a result is announced, even though we are in the middle of the literature review. This result seems out of place.
Table 2, which highlights the specific features of the research, is very interesting, but it also presents results, which seems premature (the presentation of the methodology has not yet begun).
Tables 1 and 2 are not announced in the text.
Figure 1 is very interesting, as it presents the conceptual framework of the research. However, results are presented within the same figure. Again, this seems premature.
Section 2.1 deals with higher education in general and not STEM in particular. The literature review should be strengthened with specific references to STEM education.
Methodology section
The data comes mainly from institution 1, which appears to correspond to the first descriptive study carried out.
Did the data collection tools evolve between the first collection and the subsequent ones?
Could we have a little more information on the origin of the students surveyed?
Results section
Table 6 and Figure 3 are redundant, as they present the same results (only one of the two seems necessary). The same applies to Table 10 and the following figure (which is not numbered).
Line 505: how can these significant differences between institutions be explained? These results deserve further investigation.
Paragraph 4.2: what is the answer to the research question posed? It would be interesting to present the results by highlighting the research question.
Figure 5 deserves to be explained.
Discussion section
Lines 669/671: do you have any references to justify this?
In summary, this is a very interesting article, drawing on recent references, rigorous methodology, and data analysis. Two major areas for revision are suggested:
- strengthen the review of the literature on STEM education
- avoid repetition (particularly in the presentation of results): the article would benefit from being “lightened” by eliminating redundancies, without anticipating the presentation of results.
Best wishes for the future.
Author Response
We sincerely thank Reviewer 1 for the positive evaluation of the manuscript and for the thoughtful suggestions on structure, clarity, and reduction of redundancy. These comments were extremely helpful in improving the flow and presentation of the paper.
Comment 1: Contributions should be restated at the end of the article
Response: We have reorganised the contributions. The Introduction now contains a single sentence stating: This study makes five contributions, summarised in Table 2 at the end of Section 2. The full contributions appear in Table 2, which is positioned at the end of the Literature Review (Section 2), immediately before the Methods section. This structure moves the framing material out of the Introduction while keeping it accessible before the empirical sections. The Conclusion (Section 8) restates the key findings and their implications.
Comment 2: Research questions should be presented at the end of the literature review
Response: We have moved the four research questions to the end of Section 2 (after Section 2.4), immediately before the Methods section. We have also added a bridging paragraph explaining how the four questions form an integrated framework: RQ1 and RQ2 establish the core phenomena (role expectations and dual-view ambivalence), while RQ3 and RQ4 examine how these patterns relate to adoption readiness and contextual factors.
Comment 3: Line 302 announces a result in the literature review
Response: We reviewed the passage and confirm that Section 2 presents conceptual framing and prior literature rather than results from this study. The referenced content describes theoretical expectations and hypothesised relationships drawn from prior work, not our empirical findings. To further clarify this, we have revised Figure 1 to show only hypothesised relationships (see Comment 6 below).
Comment 4: Table 2 presents results prematurely
Response: We agree. We have revised the ‘This study’ row in Table 1. The cell now reads “Tests dual-view phenomenon” instead of reporting the correlation coefficient, ensuring that no empirical results appear before the Methods section. We have also softened the “Our extension” column to read “provides one of the first empirical operationalisations of ambivalence linked to role expectations” (in response to Reviewer 3’s concern about novelty claims).
Comment 5: Tables 1 and 2 are not announced in the text
Response: We have added introductory sentences for both tables. Table 2 is introduced at the end of the research questions section with the sentence: Table 2 summarises these contributions. Table 1 is introduced at the end of Section 2.3, and it positions this study relative to prior work.
Comment 6: Figure 1 presents results within the conceptual framework
Response: We agree this was problematic. We have substantially revised Figure 1 to remove all empirical statistics. The green box title has been changed from “Key Empirical Finding” to “Hypothesised Relationship,” and all statistical values (ρ = .547, χ² = 44.12, V = 0.38) have been removed. The figure now shows only the conceptual constructs and their hypothesised relationships, with no results appearing before the Methods section.
Comment 7: Section 2.1 should be strengthened with STEM-specific references
Response: We agree this is an important addition. We have strengthened Section 2.1 with STEM-specific references in two ways. First, we added Prince (2004) on iterative problem solving and feedback in engineering education within the body of Section 2.1. Second, we added a new closing paragraph addressing STEM-specific challenges, citing Freeman et al. (2014) on active learning in science and engineering and Felder & Brent (2016) on STEM pedagogy. All three references have been added to the reference list in MDPI format.
Comment 8: Did data collection tools evolve between institutions?
Response: We have clarified in Section 3.1 that the same survey instrument was used across all four institutions, with no modifications made between data-collection waves. This ensures comparability across the sample.
Comment 9: More information on student origins
Response: Participants were predominantly Chinese nationals enrolled in English-medium Sino-foreign partnership programmes. Section 3.2 and Table 3 provide the available demographic information: institution, programme, and year of study. More detailed demographic data (e.g., specific nationality composition, prior AI experience) were not collected in order to maintain survey brevity. We acknowledge this as a limitation in Section 7.
Comment 10: Table 6 and Figure are redundant; Table 10 and following figure are also redundant
Response: We agree and have removed all redundant figures. Specifically, we deleted: (1) the old Figure 2 (participant distribution bar chart), which duplicated Table 3; (2) the old Figure 3 (endorsement rates for lecturer role changes), which duplicated Table 6; and (3) the unnumbered support-needs bar chart that followed Table 10. For the support-needs data, we have replaced the figure with explanatory text after Table 10 (see our response to Comment 13 below). The manuscript now contains three figures, all presenting unique information: Figure 1 (conceptual framework), Figure 2 (dual-view scatter plot), and Figure 3 (ambivalence literacy model).
Comment 11: Institutional differences deserve further investigation
Response: We have added interpretive text following the reporting of institutional differences in Section 4.2. We note that these differences may reflect variation in programme culture, prior AI exposure, or the degree to which institutions have already communicated expectations about AI use. We also acknowledge that the cross-sectional design does not allow causal attribution.
Comment 12: Section 4.2 should highlight the answer to the research question
Response: We have added an explicit summary paragraph at the end of Section 4.2 stating: In summary, addressing RQ1, students expect AI chatbots to redistribute lecturer effort from routine explanation toward facilitation, higher-order thinking, and motivational support. The dominant expectation is not that lecturers become less important, but that their role shifts toward activities requiring human judgment, relationship, and orchestration.
Comment 13: Figure 5 (support needs) deserves to be explained
Response: We agree that the support-needs figure required a clearer explanation. In the revised manuscript, we removed the support-needs bar chart because it duplicated Table 10. We instead added explanatory text after Table 10 describing two key patterns: (1) most students report needing substantial support, indicating that adoption readiness does not imply self-sufficiency; and (2) the positive associations between support needs and adoption indicators suggest that engaged students recognise the complexity of responsible AI use.
Comment 14: Lines 669/671 in the Discussion need references
Response: In the original manuscript, lines 669–671 discussed the lecturer’s distinctive contribution in curating disciplinary ways of thinking. In the revised manuscript, this passage appears in Section 5.1 and is directly supported by citations to Mishra & Koehler (2006) and Mishra et al. (2023) in the same sentence. We have reviewed this section and believe the citations adequately support the claims made.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThank you for this contribution to the literature. Your work is very thorough, I only have a few suggestions that will improve the manuscript.
- Four research questions is a lot for one paper, and though you address the points, there is a lot! I wonder if it would be helpful to divide things out into more than one manuscript so you do not lose the readers along the way.
- Make sure to describe and reference in text all figures, tables and appendices. Also make sure that the figure and table legends are sufficient to aid the reader in interpreting them
- I am not sure all of your figures are necessary, for instance, all of the information conveyed in Figure 2 is also present in Table 3 just above. Make sure you are not being redundant in your presentation.
- In your implications section, it is rather jarring how you move back and forth between full paragraph form and the numbered list suggestions. I would suggest adding some smoother transition or considering making everything in paragraph form. Sometimes a figure or table can help summarize these points as well.
Author Response
We are grateful to you for the encouraging feedback and helpful suggestions regarding organisation, readability, and the presentation of figures, tables, and implications. These comments helped us improve the coherence and readability of the revised manuscript.
Comment 1: Four research questions is a lot for one paper
Response: We appreciate this observation. The four research questions form an integrated framework rather than four separate studies: RQ1 and RQ2 establish the core phenomena (role expectations and dual-view ambivalence), while RQ3 and RQ4 examine how these patterns relate to adoption readiness and contextual factors. Importantly, the link between ambivalence (RQ2) and role expectations (RQ1), the core finding of the paper, would be lost if these were separated into different manuscripts. We have added a paragraph after the RQs (now at the end of Section 2) to clarify this coherence.
Comment 2: Describe and reference all figures, tables, and appendices in text
Response: We have audited the manuscript and ensured all figures, tables, and appendices are introduced in the text before they appear. Specific additions include introductions to Tables 1 and 2; explicit reference to Appendix A in Section 3.3 (The full item wording is provided in Appendix A); and explanatory text for all figures.
Comment 3: Figures duplicate tables; avoid redundancy
Response: We have removed three redundant figures: (1) the participant distribution bar chart (which duplicated Table 3), (2) the endorsement rates figure (which duplicated Table 6), and (3) the support-needs bar chart (which duplicated Table 10). The manuscript now contains three figures, each presenting unique information: Figure 1 (conceptual framework), Figure 2 (a dual-view scatter plot showing the association between enhancement and reduction indices), and Figure 3 (the ambivalence literacy conceptual model).
Comment 4: Section 6 transitions are jarring between paragraphs and numbered lists
Response: We have revised Section 6 to improve transitions and consistency. The section is now organised into three clearly labelled subsections: 6.1 (lecturers and teaching teams), 6.2 (institutions and programme leadership), and 6.3 (TNE partnerships). The opening paragraph explains the two-level organisation, and each subsection begins with a transition sentence before the numbered recommendations (e.g., The following practices address student expectations for facilitation, epistemic vigilance, and support).
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe research study investigates an important current research issue through its implementation of a multi-institutional survey across international STEM programs. The dual-view phenomenon finding (ρ = 0.547, robust across specifications) proves to be an exceptional discovery because it challenges existing unidimensional technology acceptance models. The link between social ambivalence intensity and expectations for lecturer governance and facilitation (Cramér's V up to 0.38) has considerable practical significance. The conceptual contribution of ambivalence literacy opens a promising research direction.
The psychometric foundations need to undergo extensive development especially in the area of validity evidence for the multi-select items which serve as the primary measurement tools of the study. The novelty claims need reduction and the causal language requires complete correction and the ambivalence construct needs to establish its foundations through accepted measurement techniques and the 92.2% prevalence demands statistical contextualization. The required changes can be executed without the necessity of gathering extra information.
The manuscript would enhance the literature on AI integration in higher education after revisions solve the existing issues.
Major Revisions Required
1. The multi-select items need validation evidence which goes beyond their initial face validation. At minimum from the pilot testing procedures the sample size and all feedback-based changes need to be reported. The expert review process requires documentation which should include the number of reviewers together with their qualifications and the methods used to reach agreement. The multi-select format needs evaluation to determine its ability to measure specific constructs because the current method combines endorsement frequency with belief strength.
2. Address baseline co-occurrence probability
The 92.2% dual-view prevalence needs statistical contextualization:
• Calculate and report the expected co-occurrence rate under independence, given the marginal endorsement rates for individual items
• If the expected rate is also high (e.g., >85%), acknowledge that the prevalence figure partially reflects statistical properties of the multi-select format
• Emphasize the correlation (ρ = 0.547) as the primary evidence that enhancement and reduction systematically co-activate within individuals
• Consider presenting the 92.2% figure as a descriptive result rather than as evidence of a "phenomenon" per se
3. Systematically correct causal language
The discussion requires language which shows causal connections to be replaced with language that matches the requirements of the cross-sectional research design. The statement "ambivalence drives expectations" should be changed to "ambivalence was associated with stronger expectations" The terms "leads to" and "enables" and "produces" should be replaced with "was associated with" and "co-occurred with" The construction "students who perceive X also perceive Y" should be replaced with "perception of X was positively correlated with perception of Y" The mechanisms should be presented as interpretations instead of established processes (e.g., "One interpretation is that... " or "This pattern is consistent with...").
4. Clarify the status of ambivalence literacy
The present status of ambivalence literacy needs to be explained better. The paper requires you to make different distinctions between two different types of evidence throughout the document. The first category describes factual evidence which shows the dual-view pattern with a value of ρ = 0.547 and its link to role expectations which appears in Table 9. The second category explains the teaching method which will develop ambivalence literacy as a learnable skill. The following modifications need to be made. The abstract should receive the replacement of "we therefore advance ambivalence literacy as a practical-theoretical target" with "we therefore propose ambivalence literacy as a conceptual framework for responsible AI integration" The contribution 5 entry needs a qualifier which should state that the proposed construct remains untested through empirical measurement. The section 5.8 needs to present "orchestration of ambivalence" as a research hypothesis instead of a final conclusion through the statement "We propose that teacher knowledge frameworks might usefully include... The limitations paragraph needs to explain that this study did not operationalize or measure ambivalence literacy.
Minor Revisions Recommended
5. Report response rate and discuss non-response bias
Include in Section 3.2:
The total student invitations for all four institutions should be presented.
The response rate needs to be presented as a percentage with total results and institutional results.
The study needs to explain how non-response bias occurs because students who use AI more or who show greater interest in the subject matter will respond more which leads to an overestimation of positive results.
Author Response
We would like to thank you for the careful and rigorous review of the manuscript and for the detailed methodological and conceptual suggestions. These comments were especially valuable in helping us strengthen the study's framing, clarify the status of ambivalence literacy, and improve the interpretation of the findings.
Revision 1: Validation evidence for multi-select items
Response: We have added validation information to Section 3.3. No formal pilot study was conducted; instead, the instrument drew on established item formats from prior chatbot adoption research (Bond et al., 2024; Wu & Yu, 2024). Items were reviewed independently by three faculty members with expertise in educational technology research and survey methodology. Reviewers assessed clarity, relevance, and appropriateness for STEM and TNE contexts; discrepancies were resolved through discussion, and feedback led to minor wording revisions to improve clarity. Where scaled items were used (perceived learning enhancement), internal consistency was acceptable (Cronbach’s α = 0.79). We also acknowledge that multi-select items capture endorsement frequency rather than belief intensity and note that future research could use scaled items to assess the strength of agreement.
Revision 2: Address baseline co-occurrence probability
Response: We have added the requested calculation to Section 4.3. Using the marginal endorsement rates from Tables 7 and 8: P(at least one enhancement) = 1 − [0.415 × 0.484 × 0.514 × 0.679] = 93.0%; P(at least one reduction) = 1 − [0.512 × 0.522 × 0.578 × 0.591 × 0.737] = 93.3%. The expected co-occurrence rate under independence is therefore 86.7% (0.930 × 0.933). The observed 92.2% exceeds this baseline by 5.5 percentage points, representing modest additional co-occurrence beyond what would be expected from the multi-select response structure alone. We emphasise that the positive correlation (ρ = 0.547) is the primary evidence that enhancement and reduction perceptions systematically co-occur within individuals and present the 92.2% figure as a descriptive result rather than as standalone evidence of a phenomenon.
Revision 3: Systematically correct causal language
Response: We have systematically reviewed the Discussion and Conclusion for causal language. Key changes include: (1) replacing “design driver” with associational language such as “associated with expectations” and “a relevant consideration for learning design”; (2) replacing “drives,” “leads to,” and “produces” with “was associated with,” “co-occurred with,” and “was linked to” throughout; and (3) framing mechanisms as interpretations rather than established processes (e.g., “One interpretation is that…” and “This pattern is consistent with…”).
Revision 4: Clarify the status of ambivalence literacy
Response: We have made the following changes to clarify that ambivalence literacy is a proposed construct, not an empirical finding:
(1) Abstract: Changed “practical-theoretical target” to “conceptual framework” and used “we therefore propose ambivalence literacy.”
(2) Table 2, Contribution 5: Added qualifier (proposed conceptual construct; not directly measured in this study).
(3) Section 5.8: Reframed as a hypothesis: “We therefore propose that teacher knowledge frameworks may usefully include ‘orchestration of ambivalence’… This hypothesis warrants empirical testing in future research.”
(4) Section 7 (Limitations): Explicitly acknowledges that we “did not directly measure [ambivalence literacy] as a psychometric scale” and calls for future development and validation of an ambivalence literacy instrument.
Revision 5: Report response rate and discuss non-response bias
Response: We have added text to Section 3.2: Because distribution occurred through programme communication channels rather than individualised invitations, precise counts of total recipients are unavailable and formal response rates cannot be calculated. We acknowledge potential non-response bias: students with greater interest in AI or more frequent chatbot use may have been more likely to participate, potentially inflating estimates of familiarity and positive attitudes.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors have provided a comprehensive and meticulous response to the previous review cycle. The five essential revisions of this study demonstrate academic research excellence through their complete solution of all methodological and conceptual problems which had been previously identified
