Beyond Ease of Use: Dynamics of Technology Adoption and Cognitive Load in AI-Assisted Programming for Non-Technical Students

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

As a reviewer, the manuscript is timely and relevant, but several weak areas need improvement:

1. The paper repeatedly discusses GenAI benefits, but the exact novelty over prior TAM–GenAI studies should be stated more directly.
2. Although the study claims random assignment, it is described as quasi-experimental. The authors should clarify why it is not treated as a randomized experiment.
3. Google Gemini is justified by accessibility, but the manuscript mentions “Gemini 3.5,” which may need verification and clearer version reporting.
4. The study focuses only on perception, excluding academic performance. This limits claims about learning effectiveness.
5. Cognitive friction” and task complexity are central concepts, but no direct cognitive load scale is reported.
6.  Effect sizes and confidence intervals should be added alongside p-values.
7. Claims about educational equity and GenAI as an “instrumental equalizer” need stronger empirical support.

Author Response

Comment 1:  The paper repeatedly discusses GenAI benefits, but the exact novelty over prior TAM–GenAI studies should be stated more directly.

Response 1:  We fully agree that the growing volume of literature on GenAI adoption makes it essential to articulate our specific theoretical contribution more sharply. To address this, we have revised the introductory paragraphs leading up to our research questions to explicitly distinguish our work from prior studies. We emphasize that while most TAM-GenAI research focuses on low-friction tasks characterized by immediate ease of use, our study explores environments of high cognitive friction, specifically within non-STEM programming contexts. Consequently, we have reframed 'Perceived Ease of Use' not as a static precursor, but as an acquired proficiency that students must actively conquer to unlock the tool's intrinsic value, which we believe clearly defines the novel scope of our research.

Comment 2:  Although the study claims random assignment, it is described as quasi-experimental. The authors should clarify why it is not treated as a randomized experiment.

Response 2: We are grateful to the reviewer for highlighting this taxonomic inconsistency, as you are absolutely correct. Given that we implemented individual random assignment through our learning management system, our study qualifies as a true experiment rather than a quasi-experiment. Our earlier use of "quasi-experimental" was an overly conservative misnomer, stemming from the naturalistic classroom setting and the use of a convenience sample rather than any lack of randomization. We have updated the manuscript—specifically the Abstract and Section 3—to redefine the methodology as a "randomized field experiment." This change better aligns our terminology with standard experimental frameworks and eliminates the internal contradiction regarding our research design.

Comment 3:  Google Gemini is justified by accessibility, but the manuscript mentions “Gemini 3.5,” which may need verification and clearer version reporting.

Response 3: We sincerely thank the reviewer for identifying this clerical error, as the reference to 'Gemini 3.5' was indeed a typographical oversight. As detailed in our methodology, the experimental group used the standard Google Gemini interface via university-provided Google Workspace for Education accounts, which ran on the Gemini 1.5 model architecture throughout the experimental period. We have updated Section 3 to specify that Google Gemini (powered by the Gemini 1.5 model) was the tool used in the study, thereby ensuring accurate version reporting and maintaining the technical credibility of our research.

Comment 4:  The study focuses only on perception, excluding academic performance. This limits claims about learning effectiveness.

Response 4. We fully agree with the reviewer regarding these methodological boundaries; because our study relies exclusively on self-reported perception data, it cannot claim to measure actual cognitive skill acquisition or objective learning effectiveness. To ensure this limitation is transparent, we have expanded Section 5.5 to clarify that our findings reflect perceived pedagogical value and behavioral intentions, and to advocate for future research to triangulate these results with objective metrics such as exam scores. Additionally, we have carefully audited the manuscript—particularly within the Discussion and Conclusions—to remove any terminology implying objective instructional efficacy, ensuring that our claims are strictly confined to observed shifts in student attitudes, intrinsic motivation, and behavioral intention.

Comment 5:  Cognitive friction” and task complexity are central concepts, but no direct cognitive load scale is reported.

Response 5:  We fully acknowledge the methodological boundary identified by the reviewer. It is correct that we did not employ a direct psychometric instrument, such as the NASA-TLX, to measure cognitive load. In this study, cognitive friction was inferred from the 'Perceived Ease of Use' construct, and task complexity was defined a priori based on the programming objectives. To ensure full transparency regarding our instrumentation, we have added a dedicated paragraph to the Limitations section (5.5) that explicitly states the absence of a direct cognitive load scale and recommends its integration in future research.

Comment 6:  Effect sizes and confidence intervals should be added alongside p-values.

Response 6: We thank the reviewer for highlighting the need to report effect sizes, as we fully agree that p-values alone do not provide a complete picture of the magnitude and precision of our results. In response, we have comprehensively updated the statistical reporting throughout Section 4. For all students’ t-tests, we have now calculated and included Cohen’s d along with the corresponding 95% confidence intervals. Similarly, for all non-parametric tests—specifically the Wilcoxon rank-sum and signed-rank tests—we have incorporated the effect size r and its 95% confidence intervals. These additions are now integrated alongside every inferential statistic reported in Sections 4.1 and 4.2, which we believe substantially enhances the transparency and rigor of our quantitative analysis and aligns our manuscript with contemporary reporting standards.

Comment 7:  Claims about educational equity and GenAI as an “instrumental equalizer” need stronger empirical support.

Response 7:  We completely agree with your assessment regarding the methodological boundaries of our study. Claiming that GenAI acts as an 'instrumental equalizer' or drives 'educational equity' based exclusively on self-reported perception data constituted a conclusion overreach. Consequently, we have thoroughly audited the manuscript to excise these unsupported assertions. In both the Introduction and the Discussion (Section 5.4), we have replaced terminology implying objective educational equity with phrasing that aligns strictly with our empirical findings. We now frame GenAI as a 'motivational catalyst' that democratizes willingness to engage in complex tasks, rather than as a means of democratizing objective skills. We sincerely appreciate this critical feedback, as it has ensured that our conclusions remain firmly anchored within the actual boundaries of our perception data.

 

Reviewer 2 Report

Comments and Suggestions for Authors

This manuscript examines the adoption of Generative AI (Google Gemini) versus instructional videos among non-STEM students (n=117) completing Java programming tasks. The study reports initial “cognitive friction” (lower pre-test Perceived Ease of Use for GenAI) followed by significant gains in enjoyment, ease of use, and behavioral intention after a 5‑week intervention. While the topic is timely and the focus on an under‑represented population is welcome, the manuscript suffers from fundamental problems listed below: 

• The study claims to use the Technology Acceptance Model (TAM) but does not measure Perceived Usefulness (PU),  a central TAM construct! Instead, it measures Perceived Enjoyment (from IMI). Without PU, the study cannot test or extend TAM, and conclusions about “reconceptualising ease of use” are unsupported!
  Either add PU measurement and re‑analyse, or explicitly reframe the study as using an enjoyment‑extended model without claiming to test TAM.
• The manuscript states that the experimental group used “Google Gemini 3.5” during the second semester of 2024 and first semester of 2025. All verifiable sources (e.g., Google release notes, industry reports) indicate that Gemini 3.5 was not publicly available at that time!!! This factual error is fatal to credibility.
  Correct the tool version to one that was actually available and used! 
• The manuscript repeatedly claims that GenAI acts as an “instrumental equalizer”... However, the study measured only perceptions (enjoyment, ease, intention), not actual learning outcomes or skill acquisition. These strong claims are not supported by the data.
  Remove or substantially soften such claims. Limit conclusions to what the perception data actually show (e.g., changes in attitudes and intentions).
• The RCI analysis uses a ±1.96 threshold without the standard RCI formula (which requires test‑retest reliability). The finding that almost no student showed reliable change contradicts the significant group‑level gains, and this discrepancy is not addressed!

Author Response

Comment1: The study claims to use the Technology Acceptance Model (TAM) but does not measure Perceived Usefulness (PU),  a central TAM construct! Instead, it measures Perceived Enjoyment (from IMI). Without PU, the study cannot test or extend TAM, and conclusions about “reconceptualising ease of use” are unsupported!
Either add PU measurement and re‑analyse, or explicitly reframe the study as using an enjoyment‑extended model without claiming to test TAM.

Response 1: We are truly grateful for your feedback, as it highlighted a fundamental theoretical inconsistency in our original framing of the Technology Acceptance Model (TAM). We completely agree that attempting to validate TAM without including the Perceived Usefulness construct lacked methodological rigor. Consequently, we have repositioned the manuscript as an extended hybrid motivational model, removing all claims to have tested the classical framework. This shift is now reflected throughout the Abstract, Introduction, and Discussion. To support this new approach, we have integrated the statistical analysis from the IMI 'Value/Usefulness' subscale into the Results (Sections 4.1–4.3), which demonstrates how GenAI influences perceived pedagogical value independently of operational utility. We have also addressed this methodological decision in the Limitations section (5.5), and we are confident that these adjustments have significantly strengthened the validity and academic rigor of our work.

Comment 2: The manuscript states that the experimental group used “Google Gemini 3.5” during the second semester of 2024 and first semester of 2025. All verifiable sources (e.g., Google release notes, industry reports) indicate that Gemini 3.5 was not publicly available at that time!!! This factual error is fatal to credibility.
Correct the tool version to one that was actually available and used! 

Response 2: Thank you sincerely for catching that error; you are absolutely right that Gemini 3.5 was not available during our experimental period. This was a clerical oversight during the drafting process, and we really appreciate your diligence in identifying it. We have updated Section 3 (Method) to specify that participants utilized the standard Google Gemini interface via their institutional Google Workspace for Education accounts. The text now correctly states that the service operated under the Gemini 1.5 architecture (specifically the Gemini 1.5 Flash model) throughout the second semester of 2024 and the first semester of 2025. Thank you for helping us ensure the technical accuracy of our study.

Comment 3:  The manuscript repeatedly claims that GenAI acts as an “instrumental equalizer”... However, the study measured only perceptions (enjoyment, ease, intention), not actual learning outcomes or skill acquisition. These strong claims are not supported by the data.
Remove or substantially soften such claims. Limit conclusions to what the perception data actually show (e.g., changes in attitudes and intentions).

Response 3: We sincerely thank the reviewer for highlighting this critical distinction regarding conclusion overreach. We fully agree that our original terminology overstepped the boundaries of our study, which relied on self-reported perceptions via the TAM and IMI frameworks rather than direct measurement of objective learning outcomes or cognitive skill acquisition. Following your guidance, we have thoroughly audited the manuscript to align our claims with the empirical reality of our findings. We replaced assertions about students "grasping" logic or GenAI serving as an "instrumental equalizer" with more accurate language, such as reframing the tool as a "motivational catalyst" that encourages students to willingly engage with challenging tasks. Similarly, in the Discussion and Conclusions, we have softened our claims about skill democratization to focus specifically on students' increased persistence and willingness to engage with abstract concepts, ensuring our conclusions remain firmly anchored in our perception data.

Comment 4:  The RCI analysis uses a ±1.96 threshold without the standard RCI formula (which requires test‑retest reliability). The finding that almost no student showed reliable change contradicts the significant group‑level gains, and this discrepancy is not addressed!

Response 4:  We are truly grateful for your guidance regarding the methodological omission of our RCI calculation; you are absolutely correct that the +-1.96 threshold requires both the underlying mathematical formulation and a clear reliability justification. To address this, we have incorporated a new Data Analysis section (3.5) that explicitly details the Jacobson and Truax equations and explains our reliance on Cronbach’s alpha as the reliability coefficient in the absence of a distinct test-retest pilot. Furthermore, we have added a dedicated paragraph at the end of Section 5.2 to address the apparent discrepancy between our significant group-level gains and the limited number of students who crossed the individual RCI threshold. We clarify that this is not a contradiction but a reflection of metric sensitivity: while the group-level statistics reflect a consistent, systemic shift—resulting in the observed significance—the conservative nature of the RCI aims to isolate extreme individual changes from measurement error, which is naturally less common within a five-week intervention. We appreciate you pushing us to clarify this statistical nuance, as it has significantly deepened the methodological rigor of our manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript addresses a timely and relevant topic by exploring the adoption of Generative AI tools in programming education among non-STEM students. The integration of the Technology Acceptance Model (TAM) and Intrinsic Motivation Inventory (IMI) provides an interesting perspective on behavioral and motivational dimensions associated with AI-supported learning. The study is well structured, includes recent literature, and presents a pedagogically relevant context.

However, several issues require substantial revision before the manuscript can be considered for publication.

The main concern relates to the central theoretical construct of "cognitive friction". Although this concept constitutes the foundation of the article's argument, it is not directly operationalized or empirically measured. The study infers cognitive friction through TAM and motivational variables, but no direct cognitive load measures (e.g., NASA-TLX, cognitive load scales, performance indicators, response time, error rates) were included. Consequently, the empirical evidence does not fully support the proposed conceptual claims.

Additionally, there appears to be methodological ambiguity regarding the study design. The manuscript repeatedly describes the research as quasi-experimental, while also reporting automated random assignment through Moodle. The distinction between a quasi-experimental and randomized design should be clarified, including a more explicit discussion of attrition effects and internal validity implications.

Another limitation concerns the absence of learning-performance measures. While perceptions, enjoyment, and behavioral intention are valuable indicators, no evidence is provided regarding whether students achieved superior learning outcomes or programming competence. Therefore, conclusions regarding instructional effectiveness should be interpreted cautiously.

The literature review is comprehensive but excessively descriptive in some sections. Greater synthesis and stronger integration between previous studies and the proposed theoretical framework would strengthen the manuscript.

Furthermore, more information regarding psychometric properties should be provided, including reliability indicators and validation procedures for the adapted TAM and IMI constructs.

Finally, several expressions throughout the manuscript appear somewhat promotional and should be reformulated in a more neutral academic style.

Overall, the study has potential and addresses an important educational issue; however, substantial methodological and theoretical clarification is necessary.

Comments on the Quality of English Language

The manuscript is generally well written and understandable. However, several sections would benefit from language refinement to improve clarity and academic precision. Some sentences are excessively long and occasionally difficult to follow, while certain expressions appear overly emphatic or promotional for scientific writing (e.g., strong claims and rhetorical wording). Minor stylistic editing and language polishing would improve readability and strengthen the overall presentation of the research.

Author Response

Comment 1:   The main concern relates to the central theoretical construct of "cognitive friction". Although this concept constitutes the foundation of the article's argument, it is not directly operationalized or empirically measured. The study infers cognitive friction through TAM and motivational variables, but no direct cognitive load measures (e.g., NASA-TLX, cognitive load scales, performance indicators, response time, error rates) were included. Consequently, the empirical evidence does not fully support the proposed conceptual claims.

Response 1:  We are grateful for your sharp methodological critique, as we fully acknowledge that our study did not employ direct psychometric instruments, such as the NASA-TLX, or objective performance indicators to measure cognitive load. We agree that operationalizing 'cognitive friction' strictly as a theoretical proxy—inferred from the observed deficit in the 'Perceived Ease of Use' construct—does not constitute an empirical validation of cognitive load dynamics. To rectify this overreach, we have thoroughly audited the manuscript to ensure our claims remain strictly within the boundaries of our perceptual data. We added an explicit paragraph to the Limitations section (5.5) stating that there are no direct cognitive load scales or objective performance metrics, clarifying that our findings reflect perceived pedagogical value and behavioral intentions rather than quantified mental effort or objective learning effectiveness. Furthermore, we have refined the Discussion and Conclusions to ensure all terminology is appropriately qualified to reflect these perceptual boundaries. We thank you for pushing us to tighten the alignment between our theoretical framework and our empirical instruments, which has significantly improved the rigor of our conceptual claims.

Comment 2:  Additionally, there appears to be methodological ambiguity regarding the study design. The manuscript repeatedly describes the research as quasi-experimental, while also reporting automated random assignment through Moodle. The distinction between a quasi-experimental and randomized design should be clarified, including a more explicit discussion of attrition effects and internal validity implications.

Response 2: We thank the reviewer for highlighting this taxonomic contradiction, as you are entirely correct that the execution of automated individual random assignment classifies this study as a true experiment. Our use of the term 'quasi-experimental' was an overly conservative misnomer stemming from the naturalistic classroom setting, which we have now corrected throughout the Abstract and Section 3 to accurately reflect a 'randomized field experiment.' Furthermore, in accordance with your recommendation, we have expanded Section 3.1 to explicitly address the experimental attrition that resulted in the minor group imbalance (57 vs. 60). We have clarified that this dropout was due to non-completion of the five-week cycle within the control group and included evidence from our baseline equivalence testing to confirm that this attrition did not compromise the internal validity or the homogeneous motivational baseline of our final sample.

Comment 3:   Another limitation concerns the absence of learning-performance measures. While perceptions, enjoyment, and behavioral intention are valuable indicators, no evidence is provided regarding whether students achieved superior learning outcomes or programming competence. Therefore, conclusions regarding instructional effectiveness should be interpreted cautiously.

Response 3:  We entirely agree with your assessment regarding the absence of objective learning-performance measures, as we recognize this is a critical methodological boundary that we must respect. As this limitation was also raised by another reviewer, we have addressed it comprehensively throughout the manuscript. We have added a clear declaration to the Limitations (Section 5.5) stating that the study relies exclusively on self-reported perception metrics rather than objective measures of academic performance or cognitive skill acquisition. Furthermore, we have performed a rigorous audit of the Discussion and Conclusions to remove any terminology that might imply objective instructional efficacy. By strictly limiting our claims to changes in attitudes, intrinsic motivation, and behavioral intentions, we have ensured that our findings are interpreted with the precise caution your feedback rightly demands.

Comment 4:  The literature review is comprehensive but excessively descriptive in some sections. Greater synthesis and stronger integration between previous studies and the proposed theoretical framework would strengthen the manuscript.

Response 4: We sincerely appreciate your constructive stylistic critique regarding the structure of our literature review, as we agree that earlier drafts leaned too heavily on a sequential catalog of studies rather than providing a cohesive theoretical anchor. In response, we have condensed the more descriptive passages and introduced a comprehensive synthesis paragraph at the conclusion of Section 2.3.4. This new section integrates our reviewed streams—specifically the programming challenges faced by non-STEM students, classical TAM components, and GenAI complexities—into a unified theoretical framework. By explicitly articulating how this literature justifies our conceptualization of 'cognitive friction,' we have successfully repositioned 'Perceived Ease of Use' from a static baseline to an acquired proficiency driven by intrinsic motivation. This revision creates a much stronger, logically integrated bridge between the prior research and our specific methodology, which we believe significantly improves the narrative flow and conceptual depth of the manuscript.

Comment 5: Furthermore, more information regarding psychometric properties should be provided, including reliability indicators and validation procedures for the adapted TAM and IMI constructs.

Response 5:  We appreciate you highlighting this essential methodological requirement, as we agree that reporting psychometric properties is critical to establishing the structural integrity of the adapted TAM and IMI constructs. To address this, we have expanded Section 3.4 (Instrument) to detail both our validation procedures and the reliability indicators. Beyond outlining our content validation process—which involved rigorous translation, contextualization for GenAI-assisted programming, and expert review—we have now provided the internal consistency reliability coefficients (Cronbach’s alpha) for all measurements.

Comment 6:  Finally, several expressions throughout the manuscript appear somewhat promotional and should be reformulated in a more neutral academic style.

Response 6:  We thank the reviewer for this observation, as we agree it is essential to maintain the objective tone required for scholarly communication. In response, we have revised the manuscript throughout to eliminate promotional or overly assertive language, replacing terms such as ‘unprecedented,’ ‘transformative,’ ‘democratize,’ and ‘paradigm shift’ with more cautious alternatives like ‘suggests,’ ‘indicates,’ ‘was associated with,’ and ‘may support.’ Furthermore, we have ensured that our conclusions clearly specify that these findings relate to perceived usefulness, enjoyment, ease of use, and behavioral intention within our specific sample, rather than implying broad causal or promotional claims about GenAI technology in general. And we revised the manuscript to improve clarity, reduce overly emphatic wording, and adopt a more neutral academic tone. In particular, strong formulations were replaced with more cautious language, and several long sentences were shortened to improve readability and precision.

Comment 7:  Overall, the study has potential and addresses an important educational issue; however, substantial methodological and theoretical clarification is necessary.

Response 7:  We appreciate your positive assessment of the study's relevance and have focused our revision primarily on addressing the methodological and theoretical clarifications you raised. We have refined our methodology by clarifying the experimental design, explicitly accounting for attrition, and reporting the psychometric reliability of our instruments. To tighten our theoretical framework, we have now operationalized 'cognitive friction' as a theoretical proxy rather than an empirical measure, ensuring internal consistency across our claims. Furthermore, we have audited the entire manuscript to standardize its tone, adopting a neutral, objective academic voice throughout. We are confident that these rigorous adjustments effectively address your core concerns and significantly enhance the scientific validity of our work.

Round 2

Reviewer 3 Report

Comments and Suggestions for Authors

The revised manuscript has been substantially improved and the authors have addressed many of the concerns raised in the previous review. The theoretical positioning is clearer, the methodological limitations are now more explicitly acknowledged, and the presentation of the statistical results has been strengthened. Nevertheless, several minor but important issues should be addressed before publication.

First, the abstract states that the control group’s perceptions “remained stagnant”. However, the results indicate a statistically significant pre–post improvement in Behavioral Intention in the control group. This statement should therefore be revised to specify that no significant changes were observed in Enjoyment, Ease of Use, and Value/Usefulness, while Behavioral Intention increased significantly.

Second, Section 3.4 states that the instrument comprises “three fundamental dimensions”, although four constructs are subsequently presented and analysed: Perceived Ease of Use, Behavioral Intention, Perceived Enjoyment, and Value/Usefulness. This should be corrected.

Third, the manuscript should ensure consistent terminology regarding the experimental design. The abstract refers to a “randomized field experiment”, whereas other descriptions of the study have used quasi-experimental terminology. The authors should clearly explain the unit and procedure of random assignment and use the same designation throughout the manuscript.

Fourth, expressions suggesting formal moderation or mediation should be used cautiously. Cognitive load was not directly measured through a validated instrument, and no formal moderation or mediation analysis appears to have been conducted. Statements that task complexity “moderates” adoption or that intrinsic motivation acts as a “mediator” should therefore be reformulated as interpretations or associations unless supported by an appropriate statistical analysis.

Fifth, the entire reference list should be carefully verified. Several works concerning GenAI, GPT-3.5, and contemporary technology acceptance are dated 2001, which appears chronologically implausible. Publication years, author names, titles, DOI information, duplication, and formatting should be checked against the original sources.

Finally, the manuscript would benefit from a final language and formatting revision. The numbering of RQ3 should also be corrected. Subject to these minor revisions, the manuscript will make a relevant contribution to research on GenAI adoption in non-STEM programming education.

Comments on the Quality of English Language

The manuscript is generally well written and understandable. However, several sections would benefit from language refinement to improve clarity and academic precision. Some sentences are excessively long and occasionally difficult to follow, while certain expressions appear overly emphatic or promotional for scientific writing (e.g., strong claims and rhetorical wording). Minor stylistic editing and language polishing would improve readability and strengthen the overall presentation of the research.

Author Response

Comment 1: First, the abstract states that the control group’s perceptions “remained stagnant”. However, the results indicate a statistically significant pre–post improvement in Behavioral Intention in the control group. This statement should therefore be revised to specify that no significant changes were observed in Enjoyment, Ease of Use, and Value/Usefulness, while Behavioral Intention increased significantly.	Response 1: The entire abstract has been reviewed to align with the study's actual scope. In particular, the mention to stagnant perceptions in the control group in lines 25-26 (previous manuscript) has been replaced by to acknowledge that Behavioral Intention increased significantly in control group (lines 24-25 highlighted)
Comment 2: Second, Section 3.4 states that the instrument comprises “three fundamental dimensions”, although four constructs are subsequently presented and analyzed: Perceived Ease of Use, Behavioral Intention, Perceived Enjoyment, and Value/Usefulness. This should be corrected.	Response 2: Such mention was in line 634 of the previous manuscript. It was corrected to accurately state that the instrument comprises four fundamental dimensions (highlighted line 637 of actual manuscript). Also, the 4 constructs’ descriptions were edited to better align with the methodological description and the constructs analyzed in the results.
Comment 3: Third, the manuscript should ensure consistent terminology regarding the experimental design. The abstract refers to a “randomized field experiment”, whereas other descriptions of the study have used quasi-experimental terminology. The authors should clearly explain the unit and procedure of random assignment and use the same designation throughout the manuscript.	Response 3: The terminology has been standardized throughout the manuscript to define the design exclusively as a "randomized field experiment.", both in abstract (line 15 highlighted) and in the method sections (line 469 highlighted). Also, Section 3.1 has been revised to explicitly define the individual student as the unit of randomization, and the text now details the randomization procedure conducted via the Moodle Learning Management System, including an explanation of experimental attrition and the platform configurations used to prevent cross-contamination between the experimental and control conditions.(highlighted Lines 496-501)
Comment 4: Fourth, expressions suggesting formal moderation or mediation should be used cautiously. Cognitive load was not directly measured through a validated instrument, and no formal moderation or mediation analysis appears to have been conducted. Statements that task complexity “moderates” adoption or that intrinsic motivation acts as a “mediator” should therefore be reformulated as interpretations or associations unless supported by an appropriate statistical analysis.	Response 4: The observation is correct, the original terminology implied formal mediation and moderation analyses, which were not performed. The terms "moderates," "mediating mechanism," and "mediators" have been deleted from the Introduction, Research Question 3, and Section 5.3. In particular, changes in lines 74, 84 (RQ3),  110-113, 824 (section 5.3 title), and 847 have been corrected and highlighted.
Comment 5: Fifth, the entire reference list should be carefully verified. Several works concerning GenAI, GPT-3.5, and contemporary technology acceptance are dated 2001, which appears chronologically implausible. Publication years, author names, titles, DOI information, duplication, and formatting should be checked against the original sources.	Response 5: The dates of the incorrect references have been reviewed and corrected. This change was made in both the text and the references section. In particular, corrections were made in references: ·       Chabani, Z., & Askri, S. (2026). Line 1021 ·       Izquierdo-Álvarez, V., & Jimeno-Postigo, C. (2025) line 1107 ·       Radhwan, M. G., & Radhwan, M. G. (2025). Line 1183 ·       Tariq, M. U., & Tariq, M. U. (2024). Line 1234
Comment 6: Finally, the manuscript would benefit from a final language and formatting revision. The numbering of RQ3 should also be corrected. Subject to these minor revisions, the manuscript will make a relevant contribution to research on GenAI adoption in non-STEM programming education.	Response 6: The whole manuscript was edited for proofreading in round 2, reason why in that previuos versión were highlighted in yellow the entire abstract, introduction, and conclusion sections, as well as various sections of the literature review and discussion, as those were edited to improve English spelling, punctuation, and clarity.