Exploring the Impact of Affective Pedagogical Agents: Enhancing Emotional Engagement in Higher Education

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The manuscript investigates whether an affect-aware pedagogical agent (APT) can outperform human feedback in a project-based course by comparing emotional engagement. The authors report higher scores on several positive-affect scales for the APT group and suggest that cognitively framed agent feedback may improve learning.

Comments:

1. The course context (name) is described inconsistently across sections. “Web and multi- 12 media application design with user-centered methodology course” or “Database Systems and Design” ? Please revise.
2. Participant counts differ between the methods narrative and the results tables. Clarify missing/excluded cases, and the final analysis sample.
3. The Results section is too brief and relies solely on questionnaires.
4. In addition, the questionnaire is described as having two parts, yet the results section reports only Part B in a single table. Why was Part A not analyzed/reported? Please include the analyses for Part A.
5. Please also consider analyze student performance or conduct interviews to obtain richer, triangulated evidence.
6. Figures/tables contain template remnants, incomplete captions, and numbering mismatches. Please revise.

 

Comments on the Quality of English Language

The manuscript contains noticeable grammatical errors. A thorough copy-edit is recommended.

Author Response

Reviewer #1: All corrections and changes that concern Reviewer #1 are explicitly addressed and shown in the text of the manuscript, highlighted in green color .

The manuscript investigates whether an affect-aware pedagogical agent (APT) can outperform human feedback in a project-based course by comparing emotional engagement. The authors report higher scores on several positive-affect scales for the APT group and suggest that cognitively framed agent feedback may improve learning.

Comments:

1. The course context (name) is described inconsistently across sections. “Web and multimedia application design with user-centered methodology course” or “Database Systems and Design”? Please revise.

Response: Thank you for your observation. We acknowledge the inconsistency in the course name mentioned throughout the manuscript. This was an oversight on our part. The correct course title is "Database Systems and Design", and we have carefully revised the entire manuscript to ensure that this title is used consistently across all sections. We appreciate your attention to detail, which helped us improve the clarity and coherence of the paper.

 

2. Participant counts differ between the methods narrative and the results tables. Clarify missing/excluded cases, and the final analysis sample.

Response: Thank you for your suggestion. In response, we have included a Consort-style flow diagram (as shown in Figure 3) to clearly illustrate the participant allocation and progression through the experimental phases. Additionally, we have added a corresponding explanatory paragraph in Section 2.2.1 to provide further context and clarity regarding the flow of participants across groups and study completion. We believe this addition improves the transparency and rigor of methodological description.

 

3. The Results section is too brief and relies solely on questionnaires. In addition, the questionnaire is described as having two parts, yet the results section reports only Part B in a single table. Why was Part A not analyzed/reported? Please include the analyses for Part A.

Response: We appreciate the reviewer’s observation regarding the limited scope of the Results section and the omission of analyses related to Part A of the questionnaire. In response, we have significantly expanded this section to include a detailed analysis of Part A, which captures students’ perceptions of the cognitive and affective feedback received. These results are now presented in Tables 5a and 5b, supported by descriptive statistics, independent samples t-tests, and effect size estimations. Furthermore, to ensure a more comprehensive examination of the data, we have introduced a new research question (RQ2) that specifically investigates the relationship between the feedback types and students’ emotional states. This addition allowed us to conduct both bivariate and multivariate analyses, including correlation matrices, principal component analysis (PCA), and cluster analysis, thereby enriching the interpretive depth of the study. These revisions directly address the reviewer’s concern and reinforce the empirical robustness and pedagogical relevance of our findings.

 

4. Please also consider analyze student performance or conduct interviews to obtain richer, triangulated evidence.

Response: We thank the reviewer for the valuable suggestion to enrich the study through triangulated evidence. In response, we would like to clarify that, in addition to the quantitative questionnaire items, we included two open-ended questions at the end of each section of the instrument—one in Part A and one in Part B. These questions invited students to reflect on how the APT’s feedback influenced their understanding, organization, confidence, and emotional experience during the activity. Their qualitative responses were analyzed and integrated into the Results section to complement the statistical findings and provide deeper insight into students’ perceptions. These open-ended items were not included in the initial version of the manuscript to avoid excessive length, but have now been incorporated to strengthen the interpretive depth and align with the reviewer’s recommendation for triangulation.

 

5. Figures/tables contain template remnants, incomplete captions, and numbering mismatches. Please revise.

Response: We thank the reviewer for highlighting the importance of clarity and consistency in the presentation of tables and figures. In the revised manuscript, we carefully reviewed all original visual elements and made the necessary adjustments to improve their accuracy and readability. Additionally, we have incorporated new tables and figures to reflect the expanded analyses, including those related to Part A of the questionnaire and the newly added research question. All tables and figures have been renumbered and referenced appropriately throughout the text to ensure coherence and facilitate interpretation.

Reviewer 2 Report

Comments and Suggestions for Authors

1. This article mentioned "Enhancing", but it did not specify which "delta" that indicates an enhancement.

2. The references <= 5 years were only three but almost of all were too old.

3. The numerical data in Table 2 use comma for decimal. It indicates the authors' lack of readiness when composing the article.

4. The problem identification/definition was not clear, only relied on the authors' curiosity.

5. This article only mentioned 128 people as participants in a course but the Authors did not specify their profile more detail, such as their cognitive ability or digital literacy.

6. The article did not expose the research scenario clearly and systematically.

Author Response

Reviewer #2: All corrections and changes that concern Reviewer #2 are explicitly addressed and shown in the text of the manuscript, highlighted in blue color .

1. This article mentioned "Enhancing", but it did not specify which "delta" that indicates an enhancement.

Response:  We appreciate you for your observation regarding the use of the term “Enhancing” in the title and the need to specify the corresponding “delta” that indicates measurable improvement. In the revised version of the manuscript, we have addressed this concern by explicitly reporting the differences in emotional engagement between the control and experimental groups. These differences are quantified through independent samples t-tests and effect size estimations (Cohen’s d and Hedges’ g), as presented in Tables 5a–5c. We have highlighted in blue the exact phrases from the manuscript where these improvements (“deltas”) are explained in detail.

2. The references <= 5 years were only three but almost of all were too old.

Response: Thank you for your observation. In the revised version of the manuscript, we have addressed this point by adding 10 new up-to-date references at the end of the reference list. These additions strengthen the theoretical and empirical foundation of the study and reflect the latest research in the field. Furthermore, these new references have been integrated into the Discussion, Conclusions, and Limitations sections to support key arguments, provide broader context, and highlight future research directions. This ensures that the manuscript aligns with current scholarly discourse and enhances its contribution to the topic.

3. The numerical data in Table 2 use comma for decimal. It indicates the authors' lack of readiness when composing the article.

Response: Thank you for your observation. We acknowledge that the use of commas instead of periods as decimal separators in Table 2 did not conform to international formatting standards. This issue was unintentional and reflects regional conventions in numerical notation. In response to your comment, we have corrected Table 2 by replacing commas with periods to align with the expected format. Furthermore, we have carefully reviewed all existing and newly added tables throughout the manuscript to ensure that decimal formatting is consistent and appropriate across the entire document. We appreciate your attention to detail and your valuable suggestion to improve the quality of the submission.

4. The problem identification/definition was not clear, only relied on the authors' curiosity.

Response: We appreciate the reviewer’s observation regarding the clarity of the problem identification. In response, we have revised the introduction to better articulate the research problem, grounding it not only in our academic interest but also in well-documented gaps in the literature. Specifically, we now emphasize the increasing need to explore emotionally intelligent educational technologies in higher education, and the lack of empirical evidence regarding their impact on students' emotional engagement and learning outcomes. These modifications aim to strengthen the rationale of the study and clarify its relevance within the current research context.

5. This article only mentioned 128 people as participants in a course but the Authors did not specify their profile more detail, such as their cognitive ability or digital literacy.

Response: Thank you for your suggestion. In response, we have included a Consort-style flow diagram (as shown in Figure 3) to clearly illustrate the participant allocation and progression through the experimental phases. Additionally, we have added a corresponding explanatory paragraph in Section 2.2.1 to provide further context and clarity regarding the flow of participants across groups and study completion. We believe this addition improves the transparency and rigor of the methodological description.

1. The article did not expose the research scenario clearly and systematically.

Response: We thank the reviewer for this valuable comment. In response, we have revised Section 2.1 ("Learning scenario and course context") to provide a clearer and more systematic description of the research scenario. This includes details about the academic setting, the structure and duration of the learning activity, the digital tools used, and the conditions under which the experimental and control groups interacted with the learning environment. These improvements aim to enhance transparency and allow for better replication and understanding of the study's context.

Reviewer 3 Report

Comments and Suggestions for Authors

The manuscript needs to be well framed all through; the abstract stated that the study was undertaken as part of a course on "Web and multimedia application design with user-centred methodology" while in methods section subsection 2.1.1, it jolts the reader into a "Database Systems and Design" environment. The authors should ensure that the choice of a course format was made prior to the start of the course, in order to observe uniformity; or, in the event of an intentional change, clearly indicate the reason (i.e., change in student population, institution, or pedagogical purpose). 

The difference in sample size should be clarified urgently. The text suggests a total of N = 128 participants, with equal randomization into the experimental and control arms; however, based on the analyses provided in Table 2, it is shown that there are only 57 and 58 participants in the experimental and control arms, respectively. The author/s should provide a Consort-style flow diagram or equivalent accounting that lists attrition, exclusions, and missing data information and give an open justification for why each participant was lost to follow-up or excluded from the study. 

In terms of the statistical design, the manuscript states that nine independent-samples t-tests were carried out on the affect variables after checking for homogeneity of variance using Levene's test and normality using skewness/kurtosis limits. The authors are encouraged to either provide a substantive rationale for not adjusting for multiplicity - for instance, by arguing that each outcome under consideration is a separate theoretical construct - or to use a correction procedure such as Holm or Benjamini-Hochberg. A more principled approach would be to do a multivariate analysis of variance (MANOVA) or a multivariate model using regression, accounting collectively for family wise error rate. All Levene and t-statistics (with sample degrees of freedom, p-values and effect sizes such as Cohen's d) should be presented unavoidably in the main text or in the supplementary material. 

The mechanics of the Affect-Aware Teaching (APT) are not yet specified in a form that will allow it to be reproduced. The framework lists activity theory and analytical model based on joint discourse and sentiment, but the decision rules which locate detected emotional states and temporal thresholds to certain types of cognitive or affective feedback are left as open. Simpler and smarter: define a new subsection in the other authors' fellowship that specify: (a) trigger conditions (e.g sustained negative sentiment for over 30s) (b) structure of the content libraries in terms of didactics taxonomy (c) any personalization logic (e.g. adaptive scaling of difficulty based on previous performance) and (d) pseudocode / examples of the rule engine. Furthermore, accuracy, inter-rater agreement, or benchmark diffusion weighted imaging performance data must be reported on the emotion detection effort along with the associated confidence intervals (preferably). 

With regard to the questionnaire, its reliability coefficient (a > 0.97), suggests the existence of possible redundancy between items. Further, if item-total correlations, exploratory factor analysis, and item reduction and composting score construction were completed based on factor loadings, please describe each of those results. Adding a table of factor loadings and Cronbach alpha for each of the subscales would strengthen the construct validity of the instrument. 

The referencing inside the manuscript is not completely correct; the results section points to "Table 2" for the main comparisons while in the Discussion section the same data is cited to "Table 6". Authors are expected to ensure that all tabular references are to the correct tables, by updating legends and cross references, etc. 

There is no IRB or statement of ethics. The Methods should specify clearly the Institutional Review Board (or other appropriate ethics committee) that has approved the study, with the reference number of its approval, and should also include how informed-consent was obtained (including the nature of the consent document used, and the fact that the subject's participation was voluntary). Exemption: If the research was determined to be exempt, this must be recorded including the basis for the exemption. 

A number of figures mapped to figures that are referred to but not attached to the article in the current submission: figures 1 and 2: the project timeline and a diagram of the emotion model respectively. Both figures need to be indicated with self-contained captions that briefly explain what they're about. Place holders relating to template text (for example, "should include a description...") should be removed. 

A good copy-editing will improve readability of the manuscript. The sentence "The aim of this study is to explore..." has to be corrected to "The aim of this study is to study...". A few examples of these errors are the words "ani-mated" should be corrected to "animated" and the keyword "Emotional Intelligent" should be corrected to "Emotionally Intelligent". While such changes are minimal they are sufficient to make sure the language is up to the expectations of an academic audience. 

Finally, there should be reconciliation on the funding and acknowledgement sections. While the paper states that "no external funding" there was, on the same page it indicates that "partial support" has come from a range of projects. The authors should maintain this information in an appropriate manner: the Funding section should contain the names of titles of all grants and their unique identifiers with the funding agencies while any acknowledgement paragraph should be employed to acknowledge those individuals or institutions for non-financial contributions. 

A de-identified dataset and the questionnaire instrument should be deposited in an open repository (Zenodo, OSF etc.) in order to maximize data availability. The link to the repository along with DOI and a short explanation of data content should be put in the accompanying Data Availability Statement; this will increase the transparency and allow future attempts for replication. 

In addition, the authors will be encouraged to detail whether they operationalized positive and negative affect states as means of individual items, composite score threshold scores, or alternative scoring algorithms. Where they found higher anxiety or insecurity in the experimental group, that would be supported by qualitative coding excerpts from the focus-group transcripts with a brief description of the coding method used. 

These methodological clarifications and alignment changes will make an impression of much stronger contribution of the paper to the existing literature on affect-aware instructional agents.

Author Response

Reviewer #3: All corrections and changes that concern Reviewer #2 are explicitly addressed and shown in the text of the manuscript, highlighted in violet color .

1. The manuscript needs to be well framed all through; the abstract stated that the study was undertaken as part of a course on "Web and multimedia application design with user-centred methodology" while in methods section subsection 2.1.1, it jolts the reader into a "Database Systems and Design" environment. The authors should ensure that the choice of a course format was made prior to the start of the course, in order to observe uniformity; or, in the event of an intentional change, clearly indicate the reason (i.e., change in student population, institution, or pedagogical purpose).

Response: Thank you for your observation. We acknowledge the inconsistency in the course name mentioned throughout the manuscript. This was an oversight on our part. The correct course title is "Database Systems and Design", and we have carefully revised the entire manuscript to ensure that this title is used consistently across all sections. We appreciate your attention to detail, which helped us improve the clarity and coherence of the paper.

2. The difference in sample size should be clarified urgently. The text suggests a total of N = 128 participants, with equal randomization into the experimental and control arms; however, based on the analyses provided in Table 2, it is shown that there are only 57 and 58 participants in the experimental and control arms, respectively. The author/s should provide a Consort-style flow diagram or equivalent accounting that lists attrition, exclusions, and missing data information and give an open justification for why each participant was lost to follow-up or excluded from the study.

Response: Thank you for your suggestion. In response, we have included a Consort-style flow diagram (as shown in Figure 3) to clearly illustrate the participant allocation and progression through the experimental phases. Additionally, we have added a corresponding explanatory paragraph in Section 2.2.1 to provide further context and clarity regarding the flow of participants across groups and study completion. We believe this addition improves the transparency and rigor of methodological description.

3. In terms of the statistical design, the manuscript states that nine independent-samples t-tests were carried out on the affect variables after checking for homogeneity of variance using Levene's test and normality using skewness/kurtosis limits. The authors are encouraged to either provide a substantive rationale for not adjusting for multiplicity - for instance, by arguing that each outcome under consideration is a separate theoretical construct - or to use a correction procedure such as Holm or Benjamini-Hochberg. A more principled approach would be to do a multivariate analysis of variance (MANOVA) or a multivariate model using regression, accounting collectively for family wise error rate. All Levene and t-statistics (with sample degrees of freedom, p-values and effect sizes such as Cohen's d) should be presented unavoidably in the main text or in the supplementary material.

Response: Thank you for your thoughtful and detailed suggestion. In response to your comment, we have taken the following actions:

• First, we clarified in the manuscript that each of the affective outcome variables analyzed through the independent-samples t-tests represents a distinct theoretical construct, based on the structure of the affective feedback dimension in our model. Therefore, we justified not applying a multiplicity correction, as these constructs are theoretically independent rather than multiple measures of the same latent variable.
• Nevertheless, in recognition of the concern regarding family-wise error rate, we also re-analyzed the data using a more robust multivariate approach. Specifically, we conducted a Principal Component Analysis (PCA) followed by cluster analysis to examine how affective feedback items group together and relate to students’ emotional states in a multivariate context. These results are now integrated into the revised Results section under RQ2, complementing the bivariate t-test analyses.
• Additionally, we included all relevant Levene's F-values, t-values, degrees of freedom, p-values, and Cohen's d effect sizes for each test in Table 4, as recommended. This ensures full transparency and allows readers to evaluate the statistical robustness of each result.

We believe these revisions address your concerns and enhance the statistical integrity of the manuscript while maintaining clarity and accessibility for readers.

4. The mechanics of the Affect-Aware Teaching (APT) are not yet specified in a form that will allow it to be reproduced. The framework lists activity theory and analytical model based on joint discourse and sentiment, but the decision rules which locate detected emotional states and temporal thresholds to certain types of cognitive or affective feedback are left as open. Simpler and smarter: define a new subsection in the other authors' fellowship that specify: (a) trigger conditions (e.g sustained negative sentiment for over 30s) (b) structure of the content libraries in terms of didactics taxonomy (c) any personalization logic (e.g. adaptive scaling of difficulty based on previous performance) and (d) pseudocode / examples of the rule engine. Furthermore, accuracy, inter-rater agreement, or benchmark diffusion weighted imaging performance data must be reported on the emotion detection effort along with the associated confidence intervals (preferably).

Response: Thank you for your suggestion. As this manuscript is part of a broader research program, its focus here is exploratory: the present study investigates the effects of cognitive and affective feedback in an authentic learning scenario with the Affective Pedagogical Tutor (APT), building on our prior work in which a fuzzy‑logic model was developed for emotion awareness and feedback in online learning (Arguedas et al., 2018). In the current paper, we maintained a detailed item‑level analysis using independent‑samples t‑tests to provide fine‑grained insights into distinct theoretical constructs rather than repeated measures on the same latent variable. We believe this approach is appropriate for an exploratory study, with future work planned to apply more comprehensive multivariate modelling techniques.

Reference: Arguedas, M., Xhafa, F., Casillas, L. A., Daradoumis, T., & Caballé, S. (2018). A model for providing emotion awareness and feedback using fuzzy logic in online learning. Soft Computing.

5. With regard to the questionnaire, its reliability coefficient (a > 0.97), suggests the existence of possible redundancy between items. Further, if item-total correlations, exploratory factor analysis, and item reduction and composting score construction were completed based on factor loadings, please describe each of those results. Adding a table of factor loadings and Cronbach alpha for each of the subscales would strengthen the construct validity of the instrument.

Response: Thank you for your insightful comment. We acknowledge that the high reliability coefficient (α > 0.97) could suggest potential redundancy among items. However, as stated in the manuscript, the present study is part of an exploratory research project aimed at evaluating the feasibility of using an Affective Pedagogical Agent (APA) and its influence on students’ emotional engagement in an authentic learning environment.

At this exploratory stage, we chose to retain all questionnaire items in order to ensure comprehensive coverage of the theoretical constructs under investigation (cognitive feedback, affective feedback, and emotional states). While Cronbach’s alpha values are indeed high, we have also included additional analyses to preliminarily support the instrument's validity:

• Item-total correlations were computed and reported,
• Normality assessments (skewness and kurtosis) were conducted,
• Multivariate techniques such as Principal Component Analysis (PCA) and cluster analysis were used to explore inter-item relationships and potential groupings.

We agree that further psychometric validation—including exploratory factor analysis, item reduction, and the construction of composite scores based on factor loadings—is essential for future work. These steps are already planned for a subsequent study involving a larger and more diverse sample.

To strengthen construct validity in the current manuscript, we have added:

• Reliability coefficients by subscale (all exceeding α = 0.85),
• Item-total correlations,
• And a justification for retaining all items at this phase of the research.

This process is built upon previous work by the authors, where the design and emotional modeling of pedagogical agents were studied in depth (e.g., Arguedas et al., 2018).

Reference: Arguedas, M., Daradoumis, T., & Xhafa, F. (2018). Analyzing how emotion awareness influences students’ motivation, engagement, self-regulation and learning outcome. Educational Technology & Society, 21(2), 1–17.

6. The referencing inside the manuscript is not completely correct; the results section points to "Table 2" for the main comparisons while in the Discussion section the same data is cited to "Table 6". Authors are expected to ensure that all tabular references are to the correct tables, by updating legends and cross references, etc.

Response: Thank you for pointing out the inconsistency in table referencing between the Results and Discussion sections. We have carefully reviewed all in-text citations and cross-references to tables throughout the manuscript. As a result, we have corrected the mismatched references to ensure consistency and accuracy. Specifically, we have:

• Verified and updated all references to Table 2 and Table 6 to ensure they point to the correct data.
• Revised the table legends for clarity and alignment with the corresponding sections.
• Ensured that each table is only referenced in relation to the appropriate content (e.g., learning outcomes, emotional states, or feedback type).

We appreciate your attention to detail, which helped us improve the precision and coherence of the manuscript.

7. There is no IRB or statement of ethics. The Methods should specify clearly the Institutional Review Board (or other appropriate ethics committee) that has approved the study, with the reference number of its approval, and should also include how informed-consent was obtained (including the nature of the consent document used, and the fact that the subject's participation was voluntary). Exemption: If the research was determined to be exempt, this must be recorded including the basis for the exemption.

Response: Thank you for your observation. In response, we have added a paragraph under the “Materials and Methods” section to clarify the ethical procedures followed. The study was conducted in a private university, where prior authorization was formally obtained from the institution’s management. In accordance with university policies and national ethical standards, all participants provided informed consent before participating. The consent form was reviewed by both a psychologist and a legal advisor to ensure compliance with ethical guidelines. Due to internal institutional regulations, the name of the university and approval reference number have not been disclosed in the manuscript.

8. A number of figures mapped to figures that are referred to but not attached to the article in the current submission: figures 1 and 2: the project timeline and a diagram of the emotion model respectively. Both figures need to be indicated with self-contained captions that briefly explain what they're about. Place holders relating to template text (for example, "should include a description...") should be removed.

Response: Thank you for your observation. We have carefully reviewed the manuscript and confirm that all placeholders or template instructions (e.g., “should include a description...”) have been removed from the current version. Additionally, we have ensured that Figures 1 and 2 are now correctly included in the submission, each accompanied by a self-contained caption that clearly explains its content and relevance.

9. A good copy-editing will improve readability of the manuscript. The sentence "The aim of this study is to explore..." has to be corrected to "The aim of this study is to study...". A few examples of these errors are the words "ani-mated" should be corrected to "animated" and the keyword "Emotional Intelligent" should be corrected to "Emotionally Intelligent". While such changes are minimal they are sufficient to make sure the language is up to the expectations of an academic audience.

Response We sincerely thank the reviewer for highlighting these important language and formatting issues. In response, we have carefully reviewed and thoroughly copy-edited the manuscript to ensure clarity, consistency, and alignment with academic standards. These and other minor language corrections have been implemented throughout the document to improve overall readability and meet the expectations of an academic audience.

10. Finally, there should be reconciliation on the funding and acknowledgement sections. While the paper states that "no external funding" there was, on the same page it indicates that "partial support" has come from a range of projects. The authors should maintain this information in an appropriate manner: the Funding section should contain the names of titles of all grants and their unique identifiers with the funding agencies while any acknowledgement paragraph should be employed to acknowledge those individuals or institutions for non-financial contributions.

Response: We thank the reviewer for pointing out the inconsistency between the funding and acknowledgment sections. In response, we have revised both sections to clearly differentiate between financial and non-financial support. The Funding section has been updated to list the titles of the supporting projects, along with their corresponding grant identifiers and funding agencies. Moreover, the Acknowledgments section now only includes references to individuals or institutions that contributed to the research without providing financial support.

11. A de-identified dataset and the questionnaire instrument should be deposited in an open repository (Zenodo, OSF etc.) in order to maximize data availability. The link to the repository along with DOI and a short explanation of data content should be put in the accompanying Data Availability Statement; this will increase the transparency and allow future attempts for replication.

Response: We appreciate the reviewer’s suggestion regarding data availability and transparency. However, we regret to inform you that the dataset and questionnaire cannot be publicly shared due to the data protection policies of the private university where the study was conducted. The institution’s internal regulations do not authorize the public dissemination of student data, even in de-identified form, nor of instruments developed or adapted under institutional supervision. Participants were informed of these limitations during the informed consent process, and the study was approved on the condition that all data would remain confidential and used solely for academic research purposes within the approved scope. Nonetheless, we are open to considering controlled access under specific academic requests and with prior written authorization from the institution's academic board and ethics advisor.

12. In addition, the authors will be encouraged to detail whether they operationalized positive and negative affect states as means of individual items, composite score threshold scores, or alternative scoring algorithms. Where they found higher anxiety or insecurity in the experimental group, that would be supported by qualitative coding excerpts from the focus-group transcripts with a brief description of the coding method used.

Response:  We thank the editor for this valuable suggestion. However, we would like to clarify that the present study is exploratory in nature and does not seek to provide a definitive modeling of individual affective states. Instead, our primary goal was to analyze general emotional trends associated with the cognitive and affective feedback provided by APT. For a more detailed operationalization of emotional states, including the use of fuzzy logic models and scoring algorithms, we refer to our previous work in Soft Computing (Arguedas et al., 2016), where such modeling is addressed in depth. Future developments derived from this study will consider a more rigorous treatment of affective state quantification and coding strategies.

Arguedas, M., Xhafa, F., Casillas, L., Daradoumis, T., Peña, A., & Caballé, S. (2016). A model for providing emotion awareness and feedback using fuzzy logic in online learning. Soft Computing, 21(20), 5859–5870. https://doi.org/10.1007/s00500-016-2399-0

Reviewer 4 Report

Comments and Suggestions for Authors

The article presents an interesting exploration of how pedagogical agents (those agents that can acts as colearners or tutors) can enhance emotional engagement in higher education contexts. By combining theoretical insights with empirical evidence, the authors contribute to the growing body of research that emphasizes the importance of emotional factors in digital and online learning. Here are a few remarks that can improve the paper:

1. The section two starts with a one sentence and then the subsections are presented. I would like generaly each section to have a few words of introduction rather than just saying: "Below, the materials and methods will be described with details." Obviously this is evident from the title since the title is stated as materials and methods. You also have detailed subsection titles such as "research aims, learning scenario and course context, research questions, definition of variables, instruments used and analyses." At least provide a three sentence intro under section 2 detailing that you will be mentioning the subsections.
2. It is hard to find the subsections of section 2 (which is a very important section) as there are no gaps before the subsection title and there is no bold or italic. I hope the final typesetting will correct these. However I have a concern too on section 2.2.3. it is stated "a series of statistical analyses were conducted" now this seems general it is true that later you state what the analyses are it is somewhat vague to start the section by saying a series of analysis. Fİrst of all statistical procedures are divided into various types such as descriptive stats, regression , SEM etc. You can at least state the broader category of the methods rather than saying a series of. 
3. line 365 "accepted threshold of 0.70, confirming" according to what source or sources
4. line 369 you can also use a common test for normality. Again who states that the levels of 2 and 7 for skewness and kurtosis are required for normality. What is the source ? Also kurtosis is taken as two types one in which negative values are not allowed and the other type which centers at 0 as the kurtosis of the normal distribution. Please specify. 
5. in line 373 there is a two part sentence first part can be rewritten as "The subsequent section provides a comprehensive presentation of the analysis results" The second part goes into discussion and should go into either discussion or implications section maybe the latter is more appropriate. 
6. Again the first sentence of results section is rather terse. Sİmilar to the comment above maybe you can expand on it maybe 3 4 sentences summarizing which results are presented. 
7. Also there is a big problem here from a fundamental perspective. The results are provided in one section half a page is this all the analyses ? The section 2 detailing the types of analyses I think talked about much more. So here the paper should live up to the expectations of a rigorous journal article. I think the authors should think here more on boosting this section highly and adding some subsections maybe as in the previous section 2. 
8. line 485 limitations and future research can be a separate section. 

Thanks for the work!

 

Author Response

Reviewer #4: All corrections and changes that concern Reviewer #1 are explicitly addressed and shown in the text of the manuscript, highlighted in orange color .

The article presents an interesting exploration of how pedagogical agents (those agents that can act as colearners or tutors) can enhance emotional engagement in higher education contexts. By combining theoretical insights with empirical evidence, the authors contribute to the growing body of research that emphasizes the importance of emotional factors in digital and online learning. Here are a few remarks that can improve the paper:

1. The section two starts with a one sentence and then the subsections are presented. I would like generaly each section to have a few words of introduction rather than just saying: "Below, the materials and methods will be described with details." Obviously this is evident from the title since the title is stated as materials and methods. You also have detailed subsection titles such as "research aims, learning scenario and course context, research questions, definition of variables, instruments used and analyses." At least provide a three sentence intro under section 2 detailing that you will be mentioning the subsections.

Response: Thank you very much for your insightful suggestion. We fully agree that providing a more substantial introductory paragraph enhances the clarity and structure of the section. In response, we have revised the beginning of Section 2 to include a brief introductory text outlining the contents of the subsections. Specifically, we have added the following introductory paragraph:

“This section outlines the design and implementation of an experimental study con-ducted within the “Database Systems and Design” course, aimed at evaluating the impact of affective and cognitive feedback delivered by an Affective Pedagogical Tutor (APT) compared to a human teacher. The study involved 130 undergraduate students, randomly assigned to control and experimental groups, and structured around a three-week collaborative project. Emotional engagement was assessed through a combination of forum-based discourse analysis and post-activity surveys. The APT operated within an Activity Theory framework, integrating real-time emotion analysis and time-sensitive feedback mechanisms. Independent variables included affective and cognitive feedback, while the dependent variable was students’ emotional states. Data collection instruments were designed to capture perceptions of feedback and emotional responses using Likert-type scales. Statistical procedures included reliability testing (Cronbach’s alpha), normality verification (K-S test and Box-Cox transformation), and descriptive analyses, ensuring methodological rigor and validity for subsequent inferential analyses.”

1. It is hard to find the subsections of section 2 (which is a very important section) as there are no gaps before the subsection title and there is no bold or italic. I hope the final typesetting will correct these.

Response: Thank you for your valuable observation. We agree that clear formatting is essential, especially in a key section such as Section 2. In response, we have thoroughly revised the formatting of the manuscript in accordance with the journal’s template guidelines to improve the visibility and readability of all sections and subsections. This includes adding appropriate spacing and ensuring that all subsection titles are properly distinguished using bold font styling, as required. We trust that this will enhance the structure and clarity of the manuscript in its final version.

1. However, I have a concern too on section 2.2.3. it is stated "a series of statistical analyses were conducted" now this seems general it is true that later you state what the analyses are it is somewhat vague to start the section by saying a series of analysis. Fİrst of all statistical procedures are divided into various types such as descriptive stats, regression , SEM etc. You can at least state the broader category of the methods rather than saying a series of. 

Response: Thank you very much for your helpful comment. We agree that the expression “a series of statistical analyses” was too general and could lead to ambiguity. In response, we have revised the beginning of Section 2.2.3 to specify in greater detail the statistical procedures applied to analyze the data collected. The updated version now distinguishes between descriptive statistics, correlation analysis, t-tests, multivariate analyses (such as PCA and cluster analysis), and other relevant techniques used in the study. We believe this clarification improves the transparency and scientific rigour of the methodology section.

1. line 365 "accepted threshold of 0.70, confirming" according to what source or sources

Response: Thank you for your observation. We acknowledge the omission of the reference supporting the commonly accepted threshold of 0.70. In response, we have now included the appropriate citation in line 365 to support this statement. Specifically, we refer to Nunnally and Bernstein (1994), who suggest that Cronbach’s alpha value of 0.70 or higher is generally considered acceptable to establish internal consistency in social science research.

1. line 369 you can also use a common test for normality. Again who states that the levels of 2 and 7 for skewness and kurtosis are required for normality. What is the source ? Also kurtosis is taken as two types one in which negative values are not allowed and the other type which centers at 0 as the kurtosis of the normal distribution. Please specify. 

Response: Thank you for your detailed observation. In response, we have clarified the statistical rationale and added specific references to support the interpretation of skewness and kurtosis values. In the revised manuscript, we now cite George and Mallery (2010), who consider values between -1 and +1 as excellent indicators of normality, and Gravetter and Wallnau (2014), who accept values between -2 and +2 as adequate for most statistical analyses. Additionally, to justify the criteria used for multivariate normality, we have included the reference to West, Finch, and Curran (1995), who recommend absolute skewness values below 2 and kurtosis values below 7 as acceptable thresholds when assessing multivariate normality in the context of SEM and parametric tests. We have also clarified that we are using excess kurtosis, which centers around zero for a normal distribution. These modifications are now reflected in the paragraph preceding Table 4.,

1. in line 373 there is a two part sentence first part can be rewritten as "The subsequent section provides a comprehensive presentation of the analysis results" The second part goes into discussion and should go into either discussion or implications section maybe the latter is more appropriate. 

Response: Thank you for this suggestion. We have implemented your recommendation by rewriting the first part of the sentence at line 373 as: "The subsequent section provides a comprehensive presentation of the analysis results." Additionally, the second part of the sentence, which introduced elements more appropriate for the discussion or implications section, has been relocated accordingly to ensure a clearer and more coherent structure of the manuscript. These changes improve the organization and clarity of the text, and we appreciate your valuable feedback.

1. Again the first sentence of results section is rather terse. Sİmilar to the comment above maybe you can expand on it maybe 3 4 sentences summarizing which results are presented. 

Response: Thank you for your suggestion. We have revised the opening of the Results section to provide a clearer and more informative overview. The new introductory paragraph outlines the structure and focusses of the results, summarizing the main analyses conducted, including descriptive statistics, tests of normality and reliability, comparisons between groups using t-tests, and both bivariate and multivariate analyses (such as correlation analysis, PCA, and cluster analysis). This addition helps the reader anticipate the content and enhances the clarity and coherence of the section.

1. Also there is a big problem here from a fundamental perspective. The results are provided in one section half a page is this all the analyses ? The section 2 detailing the types of analyses I think talked about much more. So here the paper should live up to the expectations of a rigorous journal article. I think the authors should think here more on boosting this section highly and adding some subsections maybe as in the previous section 2. 

Response: Thank you very much for your valuable observation. We fully agree with your suggestion and have substantially expanded the Results section to reflect the depth and rigor of the analyses described in Section 2. Specifically, we have:

• Included two additional research questions, expanding the scope and depth of our analysis.
• We have structured the Results section into two clear subsections, each corresponding to one of the research questions (RQ1 and RQ2), to improve clarity and alignment with the study objectives.
• For RQ1, we now present a comprehensive analysis of students' emotional engagement based on cognitive and affective feedback, including descriptive statistics, Levene’s tests, t-test results with degrees of freedom, p-values, and Cohen’s d effect sizes.
• For RQ2, we include both bivariate and multivariate analyses, such as Pearson correlation coefficients, Principal Component Analysis (PCA) and cluster analysis, to explore associations between students’ emotional states and the types of feedback received (cognitive and affective).
• We have also incorporated student quotes and qualitative insights to reinforce and contextualize the quantitative results, enhancing the interpretative depth of the findings.

These enhancements significantly strengthen the Results section and ensure consistency with the statistical procedures described in Section 2.2.3, improving the manuscript’s alignment with academic standards and expectations.

1. line 485 limitations and future research can be a separate section. 

Response: Thank you for your valuable suggestion. In response, we have restructured the manuscript and moved the original subsection 5.1 "Limitations and Future Research" into a standalone Section 6, as recommended. We agree that presenting this content as a separate section improves the clarity and organization of t

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

After carefully reviewing the revised manuscript, I appreciate that the authors have made substantial changes. However, several fundamental issues remain unresolved.

First, there are serious inconsistencies in the description of the study context. In the previous version, the course name was inconsistent. In the current revision, the context is now internally conflicting: in some sections, the study is described as being conducted in a master-level course, while in others, it is presented as an undergraduate course. The sample size has also been revised by the authors this time without any explanations. These raise concerns about the reliability and rigor of the reported study.

Second, although the authors have added further statistical analyses, many of these are not reported or justified in a methodologically sound way, and the figures and tables remain cluttered and poorly organized. In their current form, the analyses and reported results are not sufficiently robust or clearly presented to address the stated research questions convincingly.

Given these concerns regarding the rigor of the study, the adequacy of the analyses, and the overall clarity and credibility of the evidence, I do not believe this is suitable for publication. 

Comments on the Quality of English Language

The manuscript contains noticeable grammatical errors. A thorough copy-edit is recommended.

Author Response

Reviewer #1: All corrections and changes that concern Reviewer #1 are explicitly addressed and shown in the text of the manuscript, highlighted in green color.
After carefully reviewing the revised manuscript, I appreciate that the authors have made substantial changes. However, several fundamental issues remain unresolved.
1. First, there are serious inconsistencies in the description of the study context. In the previous version, the course name was inconsistent. In the current revision, the context is now internally conflicting: in some sections, the study is described as being conducted in a master-level course, while in others, it is presented as an undergraduate course.
Response: Thank you for your observation. You are correct in pointing out the inconsistency regarding the description of the course context. We confirm that the study was conducted as part of an undergraduate program in Computer Science, specifically within the “Database Systems and Design” course. We have now reviewed and corrected all references throughout the manuscript to ensure consistency with this information. We appreciate your attention to this detail, which has helped us improve the clarity and accuracy of the manuscript.
2. The sample size has also been revised by the authors this time without any explanations. These raise concerns about the reliability and rigor of the reported study.
Response: We thank the reviewer for the comment regarding the revision of the reported sample size. We would like to clarify that this adjustment was made in response to a specific request from another reviewer, who highlighted a discrepancy between the total randomized sample (N = 128) and the number of participants included in the final analysis. That reviewer explicitly asked for a CONSORT-style flow diagram detailing attrition, exclusions, missing data, and a transparent justification for participant loss. As shown in Figure 3, the flow diagram provides a complete overview of participant progression throughout the study. Of the 130 students initially assessed, 2 withdrew from the course prior to randomization. The remaining 128 students were allocated equally to the control group (CG) and the experimental group (EG). The difference between the randomized sample and the final analytic sample is due exclusively to non-response to the post-experience questionnaire: 7 participants in the CG and 6 participants in the EG did not complete the final measures and therefore could not be included in the analysis. Accordingly, the revision of the sample size does not reflect any change in the study data; rather, it represents a clarification introduced to improve transparency and methodological rigor in line with the reviewer’s request.
3.Second, although the authors have added further statistical analyses, many of these are not reported or justified in a methodologically sound way, In their current form, the analyses and reported results are not sufficiently robust or clearly presented to address the stated research questions convincingly.
Response: We appreciate the reviewer’s comment and have thoroughly revised Section 2.2.3 to strengthen the methodological clarity and reporting of all statistical procedures. The revised section now provides a complete description of the assumption checks, including full Kolmogorov–Smirnov results (Table 3), skewness and kurtosis values (Table 4), and the justification for applying the Box–Cox transformation. We also explicitly justify the use of each statistical technique in relation to the research questions, detailing the rationale for employing t-tests, correlation analyses, PCA and hierarchical clustering, as well as the specific criteria used for component retention, rotation and clustering methods. Additionally, we address the issue of Type I error inflation due to multiple comparisons and clarify the inclusion and interpretation of effect sizes following APA recommendations. Finally, the revised section provides a clearer justification for the use of parametric tests based on sample size, distributional diagnostics and established methodological guidelines. Together, these revisions
substantially improve the robustness, coherence and transparency of the statistical analyses.
4. the figures and tables remain cluttered and poorly organized.
Response: We appreciate the reviewer’s comment regarding the organization and clarity of the figures and tables. In response, we have conducted a comprehensive revision of all visual and tabular elements throughout the manuscript. All figures and tables have been reformatted in accordance with APA 7 style guidelines, including standardized titles, captions, numbering, spacing, and layout. We have also verified the order of appearance to ensure that each figure and table is introduced in the correct sequence within the text and that all of them are explicitly referenced in the corresponding sections. Furthermore, we have improved the visual clarity of the figures and streamlined the tables to enhance readability and interpretability. These revisions ensure that all graphical and tabular components are now clear, well structured, stylistically consistent, and fully aligned with APA 7 requirements.
Given these concerns regarding the rigor of the study, the adequacy of the analyses, and the overall clarity and credibility of the evidence, I do not believe this is suitable for publication.

Reviewer 2 Report

Comments and Suggestions for Authors

Following the comments in previous review, the reviewer found:

1. The references <5 years (issue no.2) was still too few. Although the Authors only add about 10 references, the number of old articles was too many so that the proportion was still not quite qualified.

2.  Figure 3 (issue no.5) did not answer the comment adequately since the content had no relevance with the issue.

3. Issue no.1 and 3 can be accepted.

4. The Authors still could not define the problems by contrasting the expectation vs. reality (issue no.4). 

Author Response

Reviewer #2: All corrections and changes that concern Reviewer #2 are explicitly addressed and shown in the text of the manuscript, highlighted in blue color.

 

1. The references <5 years (issue no.2) was still too few. Although the Authors only add about 10 references, the number of old articles was too many so that the proportion was still not quite qualified.

Response: We appreciate the reviewer’s helpful observation regarding the temporal distribution of the references. In this revision, we performed an additional screening of the bibliography and removed several older studies for which more recent and methodologically updated research is now available—particularly in the field of pedagogical agents and affective learning.

To make these changes completely transparent, we have highlighted in blue the references that were removed from the list because they were no longer necessary for the theoretical justification of the manuscript. At the same time, we preserved several foundational sources that continue to be essential in current research (e.g., Pekrun’s academic emotions framework; Hattie & Timperley’s model of feedback; D’Mello & Graesser’s affect-sensitive tutoring foundations). These works remain widely cited in recent literature and are required to maintain the conceptual coherence of the study.

With these adjustments, the revised manuscript now incorporates a significantly higher proportion of references published within the last five years, while retaining those older works that are indispensable for theoretical grounding. We believe this revision responds fully to the reviewer’s concern and strengthens the timeliness and relevance of the paper’s scholarly context.

2. Figure 3 (issue no.5: This article only mentioned 128 people as participants in a course but the Authors did not specify their profile more detail, such as their cognitive ability or digital literacy) did not answer the comment adequately since the content had no relevance with the issue.

Response: Thank you for pointing out that our previous reply did not sufficiently address the concern regarding the detailed profile of the 128 participants. We appreciate the opportunity to clarify this aspect. In the revised manuscript, we have now added a detailed paragraph providing the requested information about participants’ cognitive readiness, prior academic preparation, and digital literacy. This paragraph has been inserted in Section 2.2.1 and reads as follows:

“Participants consisted of second-year undergraduate students enrolled in a Computer Science program. The group presented a relatively homogeneous academic profile in terms of cognitive abilities, prior academic performance, and familiarity with digital tools. As part of their curriculum, students had completed foundational courses in programming, information systems, and introductory database concepts, which ensured adequate cognitive readiness for the project-based tasks included in the study. Owing to their continuous engagement with digital learning environments and the institutional Learning Management System (LMS), participants demonstrated sufficient digital literacy to effectively interact with the technological tools and the Affective Pedagogical Tutor (APT). Furthermore, participants possessed the necessary competencies to complete open-ended tasks, respond meaningfully to survey questions, and provide reflective and critical feedback, ensuring the relevance and validity of the data collected.”

We trust that this revision adequately addresses the reviewer’s concern by providing a clearer and more comprehensive description of the participants’ profile.

 

3. The Authors still could not define the problems by contrasting the expectation vs. reality (issue no.4.- The problem identification/definition was not clear, only relied on the authors' curiosity.)

Response: We thank the reviewer for this observation. In response, we have added a new paragraph at the end of the Introduction that explicitly clarifies the definition and identification of the problem addressed in this study. The paragraph now highlights the theoretical–empirical discrepancy regarding the expected impact of cognitive and affective feedback on emotional engagement and explains why analyzing feedback types together with students’ emotional states is necessary. This addition strengthens the conceptual justification of the study and directly addresses the reviewer’s request.

A new paragraph has been inserted at the end of the Introduction, clarifying the problem definition and explicitly identifying the expectation–reality gap that motivates the study.

Reviewer 3 Report

Comments and Suggestions for Authors

It still needs major revisions before it can be accepted.

Changes observed (positive changes):
- Methods are more detailed in ways that include the number of participants, the procedure of random assignment, attrition data and a flow diagram, also known as a CONSORT flow diagram.
This has been made more explicit and therefore the timeline of the study (nine sessions over three weeks; blocks 1-3) has been clarified, as has the Activity Theory/APT framework that is now properly introduced.
- The statistical analysis has been reinforced: correlations coefficients (Cronbach's Alpha), the normality test (Kolmogorov-Smirnov test), and t-Tests with measures of effect (Cohen's D, Hedges' G) are reported.
- Limitations have been expanded to cover the cross-section design, challenge of removing feedback effects, limitation of fixed expression and possible future directions for the Cognitive Process Analysis (CPA).

Which still needs revising 
- Context of study that is not consistent  
  The manuscript comprises an alternating design of masters levels (single 3 hour pragmatic) or undergraduate levels (3 week, 9 sessions). There is only one behaviour code, used consistently throughout the text.
- Platform inconsistency  
  References between Moodle and Blackboard can be found for the same activity. All the mentions must be unified to the actual learning management system being used in the study.
Artifacts Residual editing, English quality  
  Numerous duplicated words, spacing and punctuation glitches, and remnants of templates (e.g. "deliveriesdeliverables", "TheThis", stray headings, "Figure S1: title") are percolating throughout the manuscript. A language and formatting in-depth review is needed.
Tables: garbled lines/values.  
  Several rows have corrupted number (e.g. "≥11.416 (p<.001)") row indicated by imbalanced separators. All statistical tables need to be cleaned, checked for accuracy, and reformatted for readability.
- Normality/transformations on Likert Scales  
  The use of Box-Cox transformations in Likert data with parametric tests should be supported with relevant references and robustness analyses (e.g. non-parametric sensitivity analyses). Alternatively, analyses that are not based on the assumption of interval level normality should be presented.
- Figure cross‑references  
  Every figure mentioned (timeline, flow diagram, Emotion Analysis Model) should be there and well numbered and matched with its caption. Cross-references should be checked to make sure they are consistent.
Ethics/data statements: Polishing  
  Sections that have funding statements and acknowledgments currently have glitches in formatting. These need to be cleaned and brought to a level playing field.

Overall recommendation: Please reconsider after significant revision. Methodological description universities must achieve internal consistency; language, tables, and figures require a high quality full sweep.
- If the issues above are resolved, then the substantive additions (methods, statistics, limitations) would likely warrant an outcome of minor revision in the next round.

Author Response

Reviewer #3: All corrections and changes that concern Reviewer #2 are explicitly addressed and shown in the text of the manuscript, highlighted in violet color.

Which still needs revising

1. Context of study that is not consistent. The manuscript comprises an alternating design of master’s levels (single 3 hour pragmatic) or undergraduate levels (3 week, 9 sessions). There is only one behaviour code, used consistently throughout the text.

Response: Thank you for your observation. You are correct in pointing out the inconsistency regarding the description of the course context. We confirm that the study was conducted as part of an undergraduate program in Computer Science, specifically within the “Database Systems and Design” course. We have now reviewed and corrected all references throughout the manuscript to ensure consistency with this information. We appreciate your attention to this detail, which has helped us improve the clarity and accuracy of the manuscript.

2. Platform inconsistency. References between Moodle and Blackboard can be found for the same activity. All the mentions must be unified to the actual learning management system being used in the study.

Response: Thank you for your observation. The manuscript has been thoroughly revised to ensure consistency across all references to the learning management system. All previous mentions of Blackboard have been replaced with Moodle, which is the actual platform used in the study. The terminology has been unified throughout the document to maintain clarity and accuracy.

3. Artifacts Residual editing, English quality. Numerous duplicated words, spacing and punctuation glitches, and remnants of templates (e.g. "deliveriesdeliverables", "TheThis", stray headings, "Figure S1: title") are percolating throughout the manuscript. A language and formatting in-depth review is needed.

Response: Thank you for pointing this out. We acknowledge the presence of residual formatting artifacts, duplicated words, spacing inconsistencies, and template remnants within the manuscript. Following your recommendation, we have conducted a thorough revision to remove these issues and ensure consistency in language quality and formatting across all sections. In addition, we confirm that—upon acceptance—we will make use of the journal’s professional language-editing and formatting services to guarantee full compliance with the journal’s stylistic standards and to ensure that all remaining artifacts are fully resolved.

4. Tables: garbled lines/values. Several rows have corrupted number (e.g. "≥11.416 (p<.001)") row indicated by imbalanced separators. All statistical tables need to be cleaned, checked for accuracy, and reformatted for readability.

Response: We appreciate the reviewer’s observation regarding the presence of garbled lines, corrupted numerical values, and inconsistent separators in the statistical tables. In response, we have thoroughly revised all tables included in the manuscript, with particular attention to Tables 5(a), 5(b), and 5(c). These tables have been fully reformatted to comply with APA 7 guidelines, ensuring consistent alignment, correct decimal formatting, removal of corrupted expressions (e.g., “≥11.416 (p < .001)”), and clear presentation of Levene’s test and t-test results. We have also verified the accuracy of all numerical values against the original statistical outputs and ensured that each table is correctly referenced within the text. The revised tables are now clean, readable, and fully consistent across the manuscript.

5. Normality/transformations on Likert Scales. The use of Box-Cox transformations in Likert data with parametric tests should be supported with relevant references and robustness analyses (e.g. non-parametric sensitivity analyses). Alternatively, analyses that are not based on the assumption of interval level normality should be presented.

Response: Thank you for this valuable observation. Following your recommendation, we have revised Section 2.2.3. Statistical Assumptions and Research Analyses to more explicitly justify the use of the Box–Cox transformation with Likert-type data and to clarify why parametric procedures were retained.

In the revised version, we (a) provide a clearer conceptual rationale for using Box–Cox to stabilize variance and correct mild skewness only when necessary, (b) support this choice with established criteria for normality (e.g., thresholds for skewness and kurtosis and their methodological references), and (c) show that normality was verified through both the Kolmogorov–Smirnov test and distributional indexes prior to conducting parametric analyses. Furthermore, we included an explicit statement indicating that only two items showed significant deviations from normality and that the transformed variables exhibited improved adherence to normal distributional assumptions (as reported in Tables 3 and 4 and detailed in Appendix B).

We also strengthened the justification for the use of parametric tests by noting that, in samples larger than 50 participants per group, t-tests are considered robust to moderate violations of normality. This clarification aligns the analytical approach with best-practice recommendations in educational and psychological research.

While full non-parametric sensitivity analyses were not added—given that the distributional diagnostics supported the parametric framework—we now explicitly state the robustness of the findings and the limited impact of mild departures from normality on the conclusions.

6. Figure cross‑references. Every figure mentioned (timeline, flow diagram, Emotion Analysis Model) should be there and well numbered and matched with its caption. Cross-references should be checked to make sure they are consistent.

Response: We appreciate the reviewer’s comment regarding the organization and clarity of the figures and tables. In response, we have conducted a comprehensive revision of all visual and tabular elements throughout the manuscript. All figures and tables have been reformatted in accordance with APA 7 style guidelines, including standardized titles, captions, numbering, spacing, and layout. We have also verified the order of appearance to ensure that each figure and table is introduced in the correct sequence within the text and that all of them are explicitly referenced in the corresponding sections. Furthermore, we have improved the visual clarity of the figures and streamlined the tables to enhance readability and interpretability. These revisions ensure that all graphical and tabular components are now clear, well structured, stylistically consistent, and fully aligned with APA 7 requirements.

7. Ethics/data statements: Polishing. Sections that have funding statements and acknowledgments currently have glitches in formatting. These need to be cleaned and brought to a level playing field.

Response: Thank you for pointing this out. We have carefully revised all sections related to ethics, data availability, funding, and acknowledgments to ensure that they follow a consistent structure and meet standard formatting requirements. The updated version now presents these statements clearly, uniformly, and without formatting glitches.

Specifically, we have:

• Corrected inconsistencies in punctuation, spacing, and ordering.
• Standardized headings (e.g., Ethics Statement, Funding, Acknowledgments, Data Availability Statement) according to journal conventions.
• Ensured that each section includes the required information in the proper format and without duplicated or misplaced elements.

These revisions bring all statements to a coherent and polished format that meets the journal’s presentation standards.

Reviewer 4 Report

Comments and Suggestions for Authors

All revisions have been performed thanks.

Author Response

Thank you very much.

Round 3

Reviewer 1 Report

Comments and Suggestions for Authors

I thank the authors for addressing the comments. 

Author Response

Thank you very much for your thoughtful and encouraging feedback. We sincerely appreciate the time and expertise you dedicated to reviewing our work, as well as your recognition of the improvements made throughout the revision process. Your constructive comments were extremely valuable in strengthening the manuscript, and we are truly grateful for your positive assessment and supportive recommendation.

Author Response File: Author Response.pdf

Reviewer 3 Report

Comments and Suggestions for Authors

The methodological and contextual discrepancies which come up ago were the reason to attribute their work with the expression of "major revision" which have now been satisfactorily overcome.  

The statistical analyses such as evaluating normal, Box-Cox transformations and documentation of effect size are now fully elaborated and empirically proved.  

The tabular displays, graphical representations and ancillary materials have much consistency and aesthetic quality.  

However, the manuscript could benefit from further clarification of reporting and additional polishing of the manuscript before publication.  

While introduction and theoretical framing are well presented, the Methods section would benefit greatly from a clearer delineation of participant demographics, procedural protocols and analytical procedures in order to increase the study's reproducibility.  

The display of results would be enhanced considerably by the use of streamlined and highly legible tabular and diagrammatic material to accompany brief guiding narratives to highlight the main conclusions.  

The conclusions, though widely supported by the empirical data, would be enhanced by more closely linking such data to the original research questions and discussing the study's limitations in greater depth.

Author Response

1. While introduction and theoretical framing are well presented, the Methods section would benefit greatly from a clearer delineation of participant demographics, procedural protocols and analytical procedures in order to increase the study's reproducibility.

Response: Thank you for this valuable observation. In the revised version we have substantially expanded the Methods section to address it. First, in Section 2.2.1 (Participant Profile and Research Procedure) we now provide a more detailed description of the sample, including age range, mean and standard deviation, gender distribution, and academic year for both the initial cohort (N = 128) and the final sample (N = 115), disaggregated by control and experimental groups; this information is synthesized in Table 3 and visually summarized in Figure 3.

Second, in Section 2.1.1 (Learning Scenario and Course Context) we have clarified the procedural protocol of the three-week intervention by keeping the block-based description (Blocks 1–3) and Figure 1 (three-week timeline) and adding Table 1, which detail the sub-phases within each block, the concrete activities carried out in Sessions 1–9, and the specific roles of the instructor and the Affective Pedagogical Tutor (APT), together with the type of data collected at each step.

Third, in Section 2.2.3 (Statistical Assumptions and Research Analyses) we have introduced a new “Analytical Procedures” subsection that explicitly summarizes the sequential analytical workflow—assumption checking, group comparisons via independent-samples t-tests with effect sizes, correlation analyses, and the use of PCA followed by hierarchical clustering. Together, these additions provide a clearer delineation of participant demographics, procedural protocols, and analytical procedures, thereby strengthening the transparency and reproducibility of the study. 

2. The display of results would be enhanced considerably by the use of streamlined and highly legible tabular and diagrammatic material to accompany brief guiding narratives to highlight the main conclusions.

Response: We appreciate the reviewer’s recommendation to enhance the display of results through more streamlined and highly legible tabular and diagrammatic material. In response, the Results section (Section 3) has been substantially revised—highlighted in violet in the manuscript—to provide clearer, more interpretable summaries of the main findings.

First, in Section 3.2.1, Tables 7(a), 7(b), and 7(c) have been reformatted to report only the key statistics required to address RQ1: group means, standard deviations, p-values from independent-samples t-tests, and effect sizes (Cohen’s d and Hedges’ g). Each of these tables is now accompanied by line plots that jointly display the GC and GE means together with the corresponding p-values and effect sizes, offering a compact visual comparison across items.

Second, in Section 3.2.2, the multivariate results have been supported with additional diagrammatic material. The PCA and clustering analyses are now illustrated with clearer figures (e.g., variance-explained plots and combined PCA–cluster visualizations) that make the latent engagement profiles and the underlying structure of the feedback–emotion relationships easier to interpret.

Third, to synthesize the main outcomes, we have added a summary table that integrates the key empirical patterns, the associated statistical evidence, and the substantive conclusions for each domain (cognitive feedback, affective feedback, positive and negative emotional states, feedback–emotion associations, and engagement profiles). This final table provides a concise, high-level overview of the main results.

Finally, to avoid overloading the main text while preserving analytical transparency, the full detailed statistical outputs used in the analysis have been moved to Appendix C. Overall, these changes combine streamlined tables, clearer figures, and brief guiding narratives, directly addressing the reviewer’s suggestion and improving the readability and interpretability of the Results section.

3. The conclusions, though widely supported by the empirical data, would be enhanced by more closely linking such data to the original research questions and discussing the study's limitations in greater depth.

Response: We sincerely thank the reviewer for this constructive observation. In response to the comment, we revised both the Conclusions and Limitations sections to strengthen the alignment between empirical findings, the research questions, and the methodological scope of the study.

1. Enhancing the link between the conclusions and the research questions (RQ1 and RQ2): In the Conclusions section, we added explicit references to RQ1 and RQ2 to clarify how the empirical results directly address each question. Specifically, we incorporated a short paragraph explaining that the differences observed in motivation, curiosity, confidence, pleasure, and optimism—summarized in Tables 7(a–c)—provide a clear response to RQ1, showing that the Affective Pedagogical Tutor (APT) was associated with higher levels of students’ emotional engagement compared to the human teacher.

We also added a paragraph stating that the correlation analyses, PCA results, and cluster profiles answer RQ2, as they reveal meaningful relationships between the different APT feedback strategies (both cognitive and affective) and students’ emotional-state patterns. These additions make the connection between the empirical data and the research questions more explicit, as the reviewer suggested.

2. Deepening the discussion of the study’s limitations: In the Limitations section, we expanded the methodological considerations to provide a more nuanced discussion of the constraints of the study. We added a sentence highlighting that the post-test-only design and reliance on self-reported emotional measures limit the strength of causal interpretations, particularly for RQ1. Additionally, we included a clarification noting that some of the associations identified for RQ2 may be partially influenced by shared-method variance and should therefore be interpreted with caution.

We also reinforced the contextual limitations by acknowledging the single-course, single-institution setting and the potential influence of collaborative work dynamics on students’ emotional responses. These additions offer a more comprehensive and transparent account of the study’s methodological boundaries.

We believe these revisions significantly enhance the clarity, coherence, and interpretative depth of the manuscript and fully address the reviewer’s thoughtful suggestion.

Author Response File: Author Response.pdf