PLOS ONE

Referee Report on the Manuscript ID PONE-D-25-20092

“Reliability and validity analysis of AI voice emotion recognition app and correlation between its results and the background variables and physical symptoms or diseases of nurses”

We are grateful for your invaluable assistance in the revision of this paper. For this revision, we have closely followed the guidance provided by the editor and reviewers.

Detailed responses to each of the editor’s and reviewers’ comments have been provided below, with reviewers’ comments, followed by our comments and revisions.

In accordance with Editor’s suggestions:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Response:

We thank the Editor for the reminder. We have ensured that our manuscript meets PLOS ONE’s style requirements, including those for file naming. We have referred to the official templates provided for the style examples. (see the Title page and Text)

2. Please note that PLOS ONE has specific guidelines on code sharing for submissions in which author-generated code underpins the findings in the manuscript. In these cases, we expect all author-generated code to be made available without restrictions upon publication of the work. Please review our guidelines at https://journals.plos.org/plosone/s/materials-and-software-sharing#loc-sharing-code and ensure that your code is shared in a way that follows best practice and facilitates reproducibility and reuse.

Response:

We understand PLOS ONE’s guidelines on code sharing for submissions involving author-generated code. However, the artificial intelligence (AI) voice emotion recognition (VER) algorithm used and developed in this study is legally under trade secret protection. Due to intellectual property and trade secret considerations, we are unable to disclose any details regarding the algorithm content, source code, or internal computational mechanisms. However, we confirm that the data analysis, results, and conclusions of this study are all based on the output of the validated VER system and do not involve the disclosure of or reliance on algorithm details. We are committed to adhering to relevant regulations and properly protecting proprietary technologies and trade secrets while ensuring the scientific rigor and reproducibility of the research.

3. Thank you for stating the following financial disclosure: [This research was funded through an industry-academia collaboration agreement between Shuang Ho Hospital, Taipei Medical University and Bamboo Technology Co., Ltd. in Taiwan, with the project code A-112-006-S.].

Please state what role the funders took in the study. If the funders had no role, please state: ""The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.""

If this statement is not correct you must amend it as needed.

Please include this amended Role of Funder statement in your cover letter; we will change the online submission form on your behalf.

Response:

We thank the Editor for the reminder. Please update our information in the online submission form. We have also added the following paragraph to the cover letter: "This research was funded through an industry-academia collaboration agreement between Shuang Ho Hospital, Taipei Medical University, and Bamboo Technology Co., Ltd. in Taiwan, with the project code A-112-006-S. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The first author, Chu-Ying Huang, is an employee of Bamboo Technology Co., Ltd. and contributed only to the conceptualization of the study. All other aspects of the research design, data collection and analysis, and manuscript preparation were solely carried out by the corresponding author, Wen-Pei Chang." (see the Cover letter)

4. Thank you for stating the following in the Acknowledgments Section of your manuscript: [This research was funded through an industry-academia collaboration agreement between Shuang Ho Hospital, Taipei Medical University and Bamboo Technology Co., Ltd. in Taiwan, with the project code A-112-006-S.]

We note that you have provided funding information that is not currently declared in your Funding Statement. However, funding information should not appear in the Acknowledgments section or other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows: [This research was funded through an industry-academia collaboration agreement between Shuang Ho Hospital, Taipei Medical University and Bamboo Technology Co., Ltd. in Taiwan, with the project code A-112-006-S.].

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

Response:

We thank the Editor for this reminder. We understand that funding information should not appear in the Acknowledgments section and should only be placed in the Funding Statement section of the online submission form. We have removed all funding-related text from the manuscript. Please update our information in the online submission form. We have also added the following paragraph to the cover letter: "This research was funded through an industry-academia collaboration agreement between Shuang Ho Hospital, Taipei Medical University, and Bamboo Technology Co., Ltd. in Taiwan, with the project code A-112-006-S. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript. The first author, Chu-Ying Huang, is an employee of Bamboo Technology Co., Ltd. and contributed only to the conceptualization of the study. All other aspects of the research design, data collection and analysis, and manuscript preparation were solely carried out by the corresponding author, Wen-Pei Chang." (see the Cover letter)

5. In the online submission form, you indicated that [The data that support the findings of this study are available on request from the corresponding author. The data are not publicly available due to privacy or ethical restrictions.].

All PLOS journals now require all data underlying the findings described in their manuscript to be freely available to other researchers, either 1. In a public repository, 2. Within the manuscript itself, or 3. Uploaded as supplementary information.

This policy applies to all data except where public deposition would breach compliance with the protocol approved by your research ethics board. If your data cannot be made publicly available for ethical or legal reasons (e.g., public availability would compromise patient privacy), please explain your reasons on resubmission and your exemption request will be escalated for approval.

Response:

We thank the Editor for this reminder. We understand PLOS ONE’s open data policy. In accordance with this policy, we have uploaded an Excel file of the raw data as supplementary information for other researchers to review and reproduce the study results. (see the S1 Dataset)

6. We note that Figure 1 in your submission contain copyrighted images. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

1. You may seek permission from the original copyright holder of Figure 1 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an ""Other"" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

2. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

Response:

We thank the Editor for this reminder. We understand that all PLOS ONE content must be permitted for publication under the Creative Commons Attribution License (CC BY 4.0), and we have confirmed that the images in Fig 1 meet this requirement. We have sought permission from the original copyright holder of Fig 1 to publish the images in PLOS ONE under the CC BY 4.0 license, completed and uploaded the Content Permission Form, and added the following text to the caption of Fig 1: "(Image provided by Bamboo Technology Co., Ltd., published under a CC BY license, with permission from Bamboo Technology Co., Ltd., original copyright 2025.)" (see the Figure 1)

In accordance with Reviewer #1’s suggestions:

1. Ethical Considerations:

The manuscript mentions IRB approval and informed consent but lacks details on data anonymization and voice data protection. Given the sensitivity of biometric data, clarify how voice recordings were stored, processed, and anonymized to ensure participant confidentiality.

Response:

We thank the Reviewer for this reminder. We have added the following explanation to the Ethical considerations section: "To further protect the security of voice data and participant privacy, several data protection measures were implemented, including storing voice recordings on encrypted servers accessible only by authorized research team members using password protection. During data processing, all voice recordings were subjected to feature extraction using dedicated software, retaining only the acoustic parameters relevant to the analysis; the original voice files were not used or disclosed further. Any names or personally identifiable information contained in the recordings were deleted or anonymized prior to analysis, and the data were labeled with random codes; this completely removed any link to personal identities. All data were retained in accordance with the IRB-approved storage period and were securely destroyed after the retention period expired. The VER model used in the study was developed by Bamboo Technology Co., Ltd., which also funded the research. Although the company initiated the collaboration, it was not involved in data analysis, research processes, or decisions regarding study conclusions." (see the Ethical considerations Section on lines 126-138 on Page 8 and 9)

2. Sample Representation:

While the sample is predominantly female (85.4%), analyses involving menstrual cycles apply only to female participants. The manuscript should explicitly state whether male nurses were excluded from these specific analyses and discuss the implications for generalizability.

Response:

We thank the reviewer for the guidance. We have supplemented the following content in the Statistical analyses section: "As menstrual cycles are a physiological phenomenon unique to females, the menstrual cycle analysis was performed with only the data from female nurses. Male nurses were excluded from related analyses, and the results are only applicable to female nurses." (see the Statistical analyses Section on lines 280-283 on Page 16)

We have also added a note to the Menstrual cycles item in Table 2: "*Analysis only included the 298 female nurses." (see Table 2 on Page 26)

We have also added the following content to the Discussion section: "The analyses of specific physical symptoms (e.g., migraines and peptic ulcers) in this study included all of the participants (N = 349), that is, both male and female nurses. However, the analysis regarding menstrual cycles was limited to female nurses (n = 298); male participants were naturally excluded. As menstrual cycles are a female-specific physiological phenomenon, the findings cannot be directly generalized to male nurses or nurses of other genders. This analytical design limits the generalizability of the results across genders, particularly when exploring how physiological variables are related to emotions and physiological states. To enhance the external validity and applicability of future research, we recommend that future studies include participants of diverse genders or gender identities to examine the impact of their respective physiological and psychological variables on work-related emotions and health statuses, thus filling gaps in the existing body of research [46]." (see the Discussion Section on lines 440-450 on Page 36)

In response to this addition, we have cited the following study:

46. Polit FD, Beck CT. Nursing research: generating and assessing evidence for nursing practice. 11th Edition. Lippincott Williams & Wilkins; 2021.

3. Measurement Validity:

The AI VER app’s accuracy (70%) is moderate. Discuss limitations arising from this accuracy level (e.g., potential misclassification bias) and its impact on results.

Justify the choice of a decision tree classifier over more advanced models (e.g., neural networks) for emotion classification, addressing trade-offs between interpretability and performance.

Response:

We thank the reviewer for this reminder. We have supplemented the following content in the Research instruments section: "The preliminary model was trained using the RAVDESS data, which include 7,356 audio files of the voices of 24 professional actors (12 females and 12 males) between the ages of 21 and 40. All of these actors were native English speakers with North American accents. Their lines were delivered in various emotional states, which were then fine-tuned using a convolutional neural network classifier, achieving a preliminary accuracy of 79.2%." and "To enhance the appropriateness of the model for the expressions and culture of Mandarin speakers, the app collected voice data containing natural Mandarin spoken with a Taiwanese accent. The dataset contains 13,075 audio recordings submitted by 6,457 independent users and is 28,814 minutes in total. In the recordings, the participants talked freely and described their personal experiences and emotions. Three certified psychological counselors annotated all of the recordings using the same emotions used with RAVDESS. To process the subjectivity of the annotations, triangulation was adopted, and recordings with consistent annotations from at least two of the psychological counselors were retained. Ultimately, 1,961 training samples and 491 pen-and-paper test s

Attachments

Attachment

Submitted filename: 20250613_Response Letter.docx

PLOS ONE

Editor-in-Chief

Aug 3, 2025

Dear Editor-in-Chief,

Thank you very much for giving us the opportunity to again revise our paper entitled “Reliability and validity analysis of AI voice emotion recognition app and correlation between its results with the background variables and physical symptoms or diseases of nurses” (Manuscript ID PONE-D-25-20092R1). We have completed the required revisions, closely following the suggestions of the editor and reviewers. You may find the details of these revisions in the attached files. We believe that we have fulfilled all of the requirements for submission of the paper to PLOS ONE. If you find that any details have been overlooked, or if you need any further information, please do not hesitate to contact me.

Yours sincerely,

Wen-Pei Chang, RN, PhD

PLOS ONE

Referee Report on the Manuscript ID PONE-D-25-20092R1

“Reliability and validity analysis of AI voice emotion recognition app and correlation between its results with the background variables and physical symptoms or diseases of nurses”

We are grateful for your invaluable assistance in the revision of this paper. For this revision, we have closely followed the guidance provided by the reviewers.

Detailed responses to each of the reviewers’ comments have been provided below, with reviewers’ comments, followed by our comments and revisions.

In accordance with Reviewer #1’s suggestions:

The study addresses an important topic with potential practical value. However, the critical lack of detail regarding the AI app and its validation, coupled with limitations in study design and analysis, currently undermine the conclusions. Addressing these concerns, particularly providing full transparency about the AI tool and its validation, and controlling for confounders, is essential for the manuscript to make a significant scientific contribution. The ethical handling of sensitive voice data must also be clearly demonstrated.

Response:

We thank the reviewer for their affirmation of the importance of the research topic and its potential practical value as well as for their concrete and valuable suggestions. As shown below, we have made revisions and supplemented content as suggested:

1.Regarding the data processing details of the emotion recognition model

We have supplemented the following explanation in the Research instruments section: "To enhance the performance of the model and cultural appropriateness for Mandarin Chinese speakers, the app initially collected natural Mandarin Chinese voice data from individuals with a Taiwanese accent. The dataset contains Mandarin Chinese voice data from users in Taiwan. A total of 6,457 participants took part since data collection began, producing 13,075 items of voice data. The recordings reached 28,814 minutes in total length, averaging approximately 2.2 minutes each. Furthermore, over 100,000 voice files have been accumulated in the historical data of the platform, showing scalability potential for future model training. All voice data underwent standard preprocessing, including noise reduction and automatic segmentation, to ensure audio quality and analysis consistency. The emotions in each item of voice data were labeled by three nationally certified counseling psychologists. The emotion label classifications included eight basic emotions: worried, happy, sad, angry, surprised, fearful, calm, and neutral. Multi-label emotion classifications were permitted to reflect the emotional complexity of actual voice expressions. To increase label consistency and reliability, we employed the triangulation method and only included data in which at least two of the psychologists assigned the same labels. Ultimately, 13,075 recordings were included. Using an 8:1:1 ratio, the recordings were divided into a training set (n = 10,460), a validation set (n = 1,307), and a test set (n = 1,308). A total of 1,961 training samples and 491 test samples were retained. The optimized Mandarin Chinese model had a 79.8% accuracy rate. The open participation mechanism of the app allowed for a wide range of natural voice variations, thereby enhancing the ecological validity of the model and reflecting real-world usage scenarios. Note that we did not systematically obtain structured demographic statistics during the data collection process, which prevented us from analyzing model performance from an ethnic aspect. Future studies could include such variables for more comprehensive assessments of model versatility." (see the Research instruments Section on lines 219-242 on Page 13-14)

2.Regarding the training and validation of the emotion recognition model

We have supplemented the following explanation in the Research instruments section: "The emotion recognition model in this study was based on the wav2vec2-Large pre-trained audio representation framework and fine-tuned to adapt to Mandarin Chinese voice emotion recognition tasks. The model was trained using the triangulation-validated voice data. After each epoch, the performance of the model was validated using the test set, which was not part of the training, to observe the learning progress and generalization ability of the model. The model displayed a steady learning trend during the initial phases of training, with accuracy steadily rising from 48.9% after the first epoch to 85.9% after the fifteenth epoch. The relatively stable performance of the model demonstrated that voice emotion characteristics have been effectively learned with good classification performance and consistency." (see the Research instruments Section on lines 243-252 on Page 14)

3.Regarding the control of confounders

We have added a section with the heading “Explanatory model of emotion recognition app after controlling confounding factors” and Table 4 to the Results, with the following content: "To clarify the factors that impact the emotion recognition results from the app and control potential confounding factors, we referred to the background variables and psychological scale results presented in Tables 1 through 3 and employed multivariate statistical methods to control potential confounding effects. Confounders were selected based on the p values of the univariate analysis results to prevent over-tuning the model. The regression analysis results in Table 4 show that the model possesses significant explanatory power with regard to the recognition results of the various emotions (R2 ranging from .212 to .352, p < . 001). The VIFs of the variables were also all less than 2.5, thereby indicating the absence of multicollinearity. This analysis provides a basis for inference after controlling confounders and enhances the reliability and explanatory power of this study in terms of the validity of the emotion recognition model.

In the anger model, almost never exercising (β = .11, p = .025), the PANAS positive affect score (β = -.15, p = .007), the PSS score (β = - .13, p = .006), and the NAI score (β = .37, p < .001) were significant factors. In the fear model, peptic ulcers (β = .19, p < .001) and the PSS score (β = .36, p < .001) were significant factors. In the happiness model for the female participants, one or more cups of coffee a day (β = .11, p = .041), irregular menstrual cycles (β = -.13, p = .014), the PANAS positive affect score (β = .17, p = .006), the PANAS negative affect score (β = -.14, p = .034), and the BDI score (β = -.14, p = .034) were significantly correlated. In the sadness model for the female participants, irregular menstrual cycles (β = .30, p < .001), the PSS score (β = -.11, p = .031), the BDI score (β = .40, p < .001), and the NAI score (β = -.14, p = .003) were significantly correlated. Thus, these regression results indicate that after personal and psychological variables were controlled, the app still demonstrated significant validity and explanatory power in recognizing different emotions." (see the Results Section on lines 356-379 on Page 29-30 and Table 4 on Page 31-32)

Furthermore, to maintain the consistency of the entire paper, we have added the following content to the abstract: "To control for potential confounding variables, demographic and health-related factors including age, gender, work experience, exercise habits, and history of physical symptoms were collected and statistically adjusted in correlation analyses." and "Multiple linear regression analysis revealed the following: (1) nurses without regular exercise exhibited more anger, while those with peptic ulcers reported greater fear; (2) drinking over one cup of coffee daily was linked to higher happiness; and (3) female nurses with irregular menstrual cycles had lower happiness and greater sadness." (see the Abstract on lines 22-36 on Page 2-3)

In addition, we have supplemented the following explanation in the Discussion: "Multiple linear regression further confirmed that several background and psychological factors were significantly associated with the emotions detected by the app, supporting the preliminary psychological validity and explanatory power of the model." (see the Discussion Section on lines 406-408 on Page 36)

4.Regarding the ethical handling of sensitive voice data

We have added the following explanation to the Ethical considerations section: "Users themselves could remove their voice data on the platform at any time. Furthermore, all data were stored on a server encrypted by Microsoft Azure and managed in accordance with IRB-approved security regulations to ensure the security and legality of data access and storage processes." (see the Ethical considerations Section on lines 130-133 on Page 8)

In accordance with Reviewer #2’s suggestions:

Thank you for your revision that significantly enhanced the manuscript quality.

I recommend to accept the paper after my last comment: Regarding point 10 that I raised (Reviewer 2). I do not see any questionnaire that the nurses filled out even though you answered that you will provide it. Is this an error ? It should be in Supplementary file.

King regards.

Response:

We thank the reviewer for their affirmation of our revisions. We have supplemented and uploaded the questionnaire as a supplementary file. We apologize for this oversight and thank you for the reminder and correction.

Attachments

Attachment

Submitted filename: 20250803_Response Letter.docx

PLOS ONE

Editor-in-Chief

Nov 2, 2025

Dear Editor-in-Chief,

Thank you very much for giving us the opportunity to again revise our paper entitled “Reliability, validity, and correlates of an AI voice emotion recognition app among nurses” (Manuscript ID PONE-D-25-20092R2). We have completed the required revisions, closely following the suggestions of the editor and reviewers. You may find the details of these revisions in the attached files. We believe that we have fulfilled all of the requirements for submission of the paper to PLOS ONE. If you find that any details have been overlooked, or if you need any further information, please do not hesitate to contact me.

Yours sincerely,

Wen-Pei Chang, RN, PhD

PLOS ONE

Referee Report on the Manuscript ID PONE-D-25-20092R2

“Reliability, validity, and correlates of an AI voice emotion recognition app among nurses”

In accordance with Editor’s suggestions:

1.The title is overly long; consider shortening to emphasize the primary focus.

Response:

We have shortened the title to make it more concise and focused on the main aim of the study. The revised title is as follows: “Reliability, validity, and correlates of an AI voice emotion recognition app among nurses”.

This version highlights the psychometric evaluation as the primary focus while retaining the key contextual information regarding its use among nurses. The change has been applied to both the title page and the manuscript.

2.Abstract lacks detail on the psychometric methods (e.g., type of reliability coefficients, what “moderate and positive correlation” means numerically).

Response:

We have revised the Methods and Results sections of the Abstract to include detailed psychometric information, specifying the type of reliability coefficient (intraclass correlation coefficient, ICC) and the numerical values of the convergent validity correlations (Page 2-3, Abstract, highlighted in red). Specifically, Pearson’s correlations between app-derived emotions and corresponding standardized scales were reported (anger–Novaco Anger Inventory-Short Form, r = .42; fear–Perceived Stress Scale, r = .41; happiness–Oxford Happiness Questionnaire, r = .45; sadness–Beck Depression Inventory-II, r = .47; all p < .001), and test-retest reliability was evaluated using ICCs (range = 0.73-0.80).

3.The AI system’s accuracy (≈70–80%) should be contextualized — was this sufficient for clinical applications?

Response:

We have clarified in the Discussion section that although the AI VER app achieved approximately 70% accuracy, which is acceptable for research and preliminary screening purposes, this is not sufficient for direct clinical decision-making (Page 28, Discussion, highlighted in red). The system should therefore be regarded as a supplementary tool to assist in identifying emotional tendencies rather than a diagnostic system.

4.Which validity do you mean? Criterion validity, convergent validity, or concurrent validity?

Response:

We have clarified that the validity examined in this study refers to convergent validity, as the AI VER emotion scores were correlated with standardized psychological scales assessing similar emotional constructs (e.g., anger, fear, happiness, and sadness) (Page 2, Abstract; Page 13, Statistical Analyses; highlighted in red).

5.Regarding sample size: Convenience sampling in one hospital limits generalizability; this needs to be discussed as a limitation

Response:

We agree with the reviewer’s concern and have addressed this issue in the Discussion section (Page 29, Limitations, highlighted in red). Specifically, we have acknowledged that participants were recruited from a single medical institution using convenience sampling, which limits the external validity and generalizability of the findings to other settings or populations. This statement has been retained and slightly refined to ensure clarity.

6.No factor analysis was performed to assess construct validity. And no inter-rater or internal consistency analysis for the app (e.g., Cronbach’s α or split-half reliability). Justify it

Response:

The AI VER app is an algorithmic model that classifies emotions from acoustic features rather than a multi-item psychometric scale. Therefore, traditional factor analysis for construct validity and internal consistency indices (e.g., Cronbach’s α or split-half reliability) were not applicable. Instead, psychometric evaluation focused on test-retest reliability and convergent validity to assess temporal stability and consistency with standardized psychological measures. This clarification has been added to the Discussion section (Page 26, Discussion, highlighted in red).

7.Effect sizes are small (e.g., β = .11 or .13) and their practical significance is not discussed.

Response:

We have added an explanation in the Discussion section to interpret the small effect sizes in a practical context (Page 25-26, Discussion, highlighted in red). Although some standardized regression coefficients (β = .11 or .13) were small, they still indicate meaningful patterns in nurses’ emotional states, reflecting subtle but consistent influences of lifestyle and health factors (e.g., exercise habits and menstrual irregularity) on emotion recognition outcomes. In large populations, such small effects may accumulate and have practical implications for nurse well-being and workplace interventions.

8.Discussion tends to overstate the implications — the conclusion that the app can “objectively detect emotional patterns” may be premature given moderate correlations and app accuracy.

Response:

We have revised both the Discussion and Conclusion sections to adopt a more cautious and evidence-based tone (Page 25-29, Discussion; Page 30, Conclusion; highlighted in red). Specifically, we have replaced the statement that the app can “objectively detect emotional patterns” with phrasing that emphasizes its preliminary and assistive role. The revised text now indicates that the AI VER app may tentatively reflect emotional tendencies and should be regarded as a supplementary or exploratory tool, rather than a diagnostic system, given its moderate correlations and 70-80% accuracy. These revisions clarify that the current findings provide preliminary evidence of potential utility and that further validation in larger and more diverse samples is required.

9.Causal language should be avoided (e.g., “Peptic ulcers were associated with fear”→ should be “Peptic ulcers correlated with fear”).

Response:

We have carefully revised the related sentences in the Discussion section to avoid any causal wording (Page 30, Discussion, highlighted in red). Specifically, we have replaced expressions implying causation with neutral, correlational terms (e.g., “were associated with”) and explicitly stated that the cross-sectional design limits causal inference.

10.How does the app perform compared with self-reported emotion scales (e.g., accuracy, AUC)?

Response:

The present study focused on psychometric validation (reliability and validity) rather than predictive performance evaluation. We have added clarification in the Discussion stating that metrics such as accuracy or AUC were not computed but that future studies may include ROC-AUC analyses with larger datasets to further quantify predictive capability (Page 26, Discussion, highlighted in red).

11.The limitation should include Single-institution sample bias: Self-report physical health data (potential recall bias).

Response:

The single-institution sampling bias had already been addressed in the Limitations section. We have now added a statement noting that physical health data (e.g., exercise habits and medical history) were self-reported and may be subject to recall bias or social desirability effects (Page 29, Limitations, highlighted in red).

12.Was there any independent validation to prevent bias from the company developing the app?

Response:

We have clarified this point in the revised Ethical considerations section (Page 6, Ethical considerations, highlighted in red). Although the VER model was developed and funded by Bamboo Technology Co., Ltd., all data collection, statistical analyses, and result interpretations were conducted independently by our academic research team at Taipei Medical University to ensure objectivity and prevent developer bias. The company was not involved in data analysis, research procedures, or study conclusions.

Attachments

Attachment

Submitted filename: 20251101_Response Letter.docx