Adoption of Artificial Intelligence-Driven Fraud Detection in Banking: The Role of Trust, Transparency, and Fairness Perception in Financial Institutions in the United Arab Emirates and Qatar

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

The focus of the study is on the use of AI in fraud detection, but the introduction and literature review seem to focus more on AI in general. Significant work is needed to ensure the appropriateness of the references, and the literature review need to be expanded. It is difficult to see the representativeness of the sample used in the analysis, specifically with reference to the fraud function at banks. More information is needed around the methodology used and the tables presented. The results do not seem to indicate strong relationships, which makes it difficult to support the conclusions drawn.

Comments for author File: Comments.pdf

Author Response

 

Dear Reviewer,

We thank you for your thorough and constructive comments on our manuscript titled “Trust and Transparency in Machine Learning for Accounting Fraud Detection: Evidence from UAE and Qatar.” Your suggestions were invaluable in improving the depth, clarity, and empirical rigor of the paper.

As per your suggestions, we have restructured and revised the Introduction, Literature Review, Methodology, and Results sections to align more closely with the study’s aims and to address concerns of relevance, validity, and clarity. Detailed responses to each of your comments are provided below.

 

? Abstract

Comment:
Language editing is needed.

Response:
We have carefully revised the abstract to enhance clarity, grammar, and conciseness. The revised version better reflects the restructured research framework and highlights key findings and practical implications.

 

? Introduction

Comment:

• Define fraud and clarify which types are addressed.
• Provide clear examples distinguishing AI-generated alerts from traditional rule-based systems.
• Ensure citations directly support fraud detection.

Response:
To address these points, we have revised the Introduction to include:

• A concise definition of financial fraud, particularly within the context of compliance-based banking.
• Clarification of fraud types considered (e.g., anomalous transactions, synthetic identity fraud).
• An example contrasting AI-generated fraud alerts (e.g., anomaly-based triggers) with rule-based detection, highlighting differences in interpretability and justification.
• Additionally, all references have been reviewed, and only sources directly relevant to fraud detection in banking have been retained or updated.

These changes contribute to framing a more focused problem statement and set the stage for the research gap.

 

? Literature Review

Comment Summary:

• Some references are not relevant to fraud detection (e.g., Morshed & Khrais, Tahir et al.).
• Expand on Explainable AI (XAI), fairness-aware systems, and clarify how these relate to trust in fraud detection.
• Provide more discussion on regional regulatory differences (UAE vs. Qatar) and global AI governance frameworks.
• Explain SHAP and LIME, and include real-world examples (e.g., FAB, Deutsche Bank).
• Define terms like interpretability and justify fairness as a concern in fraud detection.

Response:
We have completely rebuilt the Literature Review to reflect your feedback:

• All citations have been audited and updated. Irrelevant studies have been removed or recontextualized with clear justification of their relevance to AI in fraud detection.
• A dedicated section now defines interpretability, discusses SHAP vs. LIME, and illustrates how they enhance fraud alert justification.
• We added real-world cases, such as First Abu Dhabi Bank’s use of compliant AI systems, and Deutsche Bank’s deployment of fairness-aware AI in fraud analytics.
• A comparative discussion now links global AI governance frameworks (e.g., EU AI Act, OECD AI Principles) with UAE and Qatari regulatory trends, reinforcing the significance of regional compliance.
• We have restructured the narrative on fairness perception, distinguishing its role in fraud detection versus credit risk, and clarifying how algorithmic bias can manifest in false positives and discriminatory alerts.

These additions respond directly to your concerns and ensure the literature review is both conceptually sound and fraud-specific.

 

? Methodology

Comment:

• Clarify participant expertise in fraud detection.
• Explain sampling and recruitment.
• Justify using PLS-SEM over CB-SEM.
• Detail construct validation.

Response:
As part of restructuring the paper, we revised the Methodology section to address all points:

• We clarify that participants were professionals (auditors, compliance officers, risk managers) with demonstrated involvement in fraud detection, AI model review, or regulatory compliance.
• The sampling strategy now includes detailed eligibility criteria, recruitment channels (e.g., LinkedIn, banking networks), and strategies to mitigate response bias (e.g., randomization, anonymity).
• We now justify PLS-SEM based on the study’s focus on prediction and theory development, and its suitability for moderate sample sizes and non-normal data distributions.
• The measurement section explicitly states that constructs (e.g., fairness perception, trust) are adapted from validated instruments used in prior studies and were tested through pilot surveys to ensure contextual relevance.

 

? Results and Discussion

Comment:

• R² values seem low—clarify statistical relevance.
• Strengthen connection to prior research.
• Emphasize practical implications for banks and regulators.

Response:
The Results section has been revised to include:

• Justification for R² values in the 0.40–0.48 range, which are moderate and acceptable in behavioral research using PLS-SEM (cited: Hair et al., 2021).
• Revised tables and labels to better align with the final hypothesis model (now 8 hypotheses, not 12), reducing confusion and improving clarity.
• We have enhanced the Discussion by linking findings more directly to past studies (e.g., trust as a predictor of adoption, fairness perception as a moderating influence).
• A new subsection now translates the empirical findings into practical guidance for:
• Banks (e.g., investing in explainable and auditable AI tools),
• Regulators (e.g., formalizing fairness reviews or transparency disclosures),
• Developers (e.g., incorporating human-in-the-loop mechanisms).

 

We sincerely appreciate your feedback. Your comments helped us refine not only the structure but also the substance of our study. The resulting manuscript presents a more compelling, relevant, and practice-informed contribution to the fields of AI compliance and financial governance.

Sincerely,

 

Reviewer 2 Report

Comments and Suggestions for Authors

The study investigates the adoption of AI-driven fraud detection in financial institutions in the UAE and Qatar, focusing on how trust, transparency, and fairness perception influence adoption among auditors, compliance officers, and risk managers.

The authors use a quantitative approach with Partial Least Squares Structural Equation Modeling (PLS-SEM) and Multi-Group Analysis (MGA) to analyze responses from banking professionals. The results show that transparency significantly enhances trust, which in turn drives AI adoption, while fairness perception plays a moderating role, indicating that addressing algorithmic bias improves AI credibility.

While the paper provides valuable insights, it requires revisions to strengthen the theoretical framework, clarify the methodological justification, and improve the discussion of findings and policy implications.

Suggestions for Improvements

1. Introduction

• The introduction provides a solid background, but the research gap should be more clearly defined.
• Consider explicitly stating how your study extends previous research on AI transparency, trust, and fairness in fraud detection.
• Adding more recent references on AI ethics and regulatory compliance would strengthen the theoretical foundation.

2. Literature Review

• While the literature review covers key topics, it should include more discussions on global AI governance frameworks and how they compare to UAE and Qatari regulations.
• The role of Explainable AI (XAI) is mentioned but not deeply analyzed. Consider expanding this discussion.

3. Methodology

• The use of PLS-SEM and Multi-Group Analysis (MGA) is appropriate, but the justification for choosing these methods over alternatives (e.g., CB-SEM) should be elaborated.
• More details are needed on how variables like fairness perception and trust were measured. Were these constructs validated in prior studies?
• The sample selection process and potential response biases should be discussed in more detail. How were participants recruited, and how was bias mitigated?

4. Results and Discussion

• The results are well-structured, but clearer comparisons with previous studies would improve the discussion.
• Some tables could be simplified for better readability. Consider summarizing key findings in a more concise manner.
• The discussion of findings should better connect the empirical results with practical implications. What do these findings mean for banking institutions and policymakers?

5. Conclusion and Practical Implications

• The conclusion should more explicitly connect the study’s findings to concrete recommendations for financial institutions and regulators.
• Consider suggesting specific AI governance strategies or regulatory frameworks that could improve trust and fairness in AI adoption.

6. Language and Presentation

• The paper is generally well-written but contains some complex sentences and minor grammatical errors that affect readability.

Comments on the Quality of English Language

The paper is generally well-written, but there are some areas where the clarity and readability could be improved.

• Some sentences are overly complex, making it difficult to follow key arguments. Simplifying sentence structure would enhance clarity.
• Minor grammatical errors and awkward phrasing appear throughout the text, which may affect readability.
• Certain sections, particularly the methodology and discussion, could be more concise and structured for better flow.

Author Response

Dear Reviewer,

We sincerely thank you for your thoughtful and constructive feedback on our manuscript titled “Trust and Transparency in Machine Learning for Accounting Fraud Detection: Evidence from UAE and Qatar”. Your insights have significantly improved the quality and clarity of our work. Below, we provide a point-by-point response to your comments and explain the revisions made accordingly.

 

1. Introduction

Reviewer Comment:
The introduction provides a solid background, but the research gap should be more clearly defined. Consider explicitly stating how your study extends previous research on AI transparency, trust, and fairness in fraud detection. Adding more recent references on AI ethics and regulatory compliance would strengthen the theoretical foundation.

Response:
We appreciate this suggestion. The introduction has been revised to:

• Clearly articulate the research gap in examining transparency, fairness, and trust in AI adoption within highly regulated banking environments (UAE and Qatar).
• Explicitly state how this study extends previous work by integrating these factors in a comprehensive framework and applying it to a unique regulatory context.
• Incorporate recent references on AI ethics and compliance (e.g., Ghasemaghaei & Kordzadeh, 2024; McNally & Bastos, 2025) to strengthen the theoretical base.

 

2. Literature Review

Reviewer Comment:
While the literature review covers key topics, it should include more discussions on global AI governance frameworks and how they compare to UAE and Qatari regulations. The role of Explainable AI (XAI) is mentioned but not deeply analyzed. Consider expanding this discussion.

Response:
We have expanded the literature review in two key ways:

• A comparative discussion on global AI governance frameworks (e.g., EU AI Act, OECD principles) and their relation to regulatory practices in the UAE and Qatar has been added.
• The section on Explainable AI (XAI) has been revised to include deeper analysis of tools such as SHAP and LIME, their technical relevance, limitations, and implications for trust-building in fraud detection.

 

3. Methodology

Reviewer Comment:
The use of PLS-SEM and Multi-Group Analysis (MGA) is appropriate, but the justification for choosing these methods over alternatives (e.g., CB-SEM) should be elaborated. More details are needed on how variables like fairness perception and trust were measured. Were these constructs validated in prior studies? The sample selection process and potential response biases should be discussed in more detail.

Response:
The Methodology section has been updated as follows:

• Justification for using PLS-SEM over CB-SEM has been added, emphasizing its suitability for theory development and moderate sample sizes.
• Measurement constructs (e.g., fairness perception, trust) are now clearly attributed to validated prior studies and scales.
• The sampling process is now described in greater detail, including eligibility criteria, recruitment channels, and how response bias was mitigated through anonymity, pilot testing, and question randomization.

 

4. Results and Discussion

Reviewer Comment:
The results are well-structured, but clearer comparisons with previous studies would improve the discussion. Some tables could be simplified for better readability. Consider summarizing key findings in a more concise manner. The discussion of findings should better connect the empirical results with practical implications. What do these findings mean for banking institutions and policymakers?

Response:
We have revised the Results and Discussion sections to:

• Include clearer comparisons between our findings and existing literature (e.g., on the mediating role of trust and the impact of transparency).
• Simplify complex tables and highlight key statistical findings.
• Strengthen the discussion of practical implications, including what the results suggest for decision-making by financial institutions and the development of trustworthy AI systems in compliance-driven sectors.

 

5. Conclusion and Practical Implications

Reviewer Comment:
The conclusion should more explicitly connect the study’s findings to concrete recommendations for financial institutions and regulators. Consider suggesting specific AI governance strategies or regulatory frameworks that could improve trust and fairness in AI adoption.

Response:
The Conclusion section now includes:

• Specific recommendations for financial institutions to invest in XAI tools and auditability frameworks.
• Policy suggestions for regulators in the UAE and Qatar, such as developing sector-specific AI guidelines modeled after international standards.
• Emphasis on the need for collaborative AI governance involving developers, auditors, and regulators to ensure trust and fairness in AI-based fraud detection.

 

We are grateful for your valuable feedback and hope that the revised manuscript now meets the high standards expected by the journal. Please do not hesitate to let us know if any additional clarifications or modifications are needed.

Sincerely,

Reviewer 3 Report

Comments and Suggestions for Authors

Overall, the topic is very interesting for the readers in the field of compliance, and it offers new insights of the practice of AI in compliance in financial institutions while overseeing its ethical considerations. The authors have covered all the aspects. Minor adjustments noted with regard to the length of the paper. There are 12 Hypotheses discussed; some could be merged or removed. On page 4, the H3 & H4 & H5 are missing. The content could be divided for 2 difference research papers: one covering the scope of UAE and another covering the scope of Qatar. Else, the authors might think to keep the 2 countries while reducing the content. 

 

Author Response

Dear Reviewer,

Thank you for your thoughtful and encouraging feedback on our manuscript titled “Trust and Transparency in Machine Learning for Accounting Fraud Detection: Evidence from UAE and Qatar.” We are grateful for your positive comments regarding the topic's relevance, ethical contribution, and overall structure. Your suggestions were insightful and have been carefully considered in our revised manuscript.

 

1. Paper Length and Focus

Reviewer Comment:
Minor adjustments noted with regard to the length of the paper. The content could be divided for two different research papers: one covering the scope of UAE and another covering the scope of Qatar. Else, the authors might think to keep the two countries while reducing the content.

Response:
We appreciate your observation. We have chosen to retain both countries in a single study to preserve the comparative analysis central to our research question. However, we have reduced redundancy and streamlined sections throughout the manuscript to improve clarity and conciseness. The revised version is now more focused and balanced in its treatment of UAE and Qatari data while maintaining analytical depth.

 

2. Hypotheses Structure and Clarity

Reviewer Comment:
There are 12 Hypotheses discussed; some could be merged or removed. On page 4, the H3 & H4 & H5 are missing.

Response:
Thank you for this helpful observation. In response, we have restructured and reduced the number of hypotheses from 12 to 8, based on theoretical coherence and empirical overlap. This includes:

• Merging related constructs (e.g., separate trust and fairness hypotheses).
• Clarifying all hypothesis labels (H1–H8), ensuring none are skipped or duplicated.

The Conceptual Framework and Hypotheses section has been updated accordingly, and all hypotheses are now clearly presented and consistently referred to in both the methodology and results sections.

 

3. Literature, Methodology, and Results Revision

Reviewer Comment:
Overall, the topic is very interesting for the readers in the field of compliance, and it offers new insights into the practice of AI in financial institutions while overseeing its ethical considerations. The authors have covered all the aspects.

Response:
We thank you for your supportive comments on the study’s originality and relevance to the compliance and AI ethics field.

According to your comment, we have re-done the literature, methodology, and results sections. Specifically:

• The literature review was refined to integrate recent research on AI fairness, explainability, and regional regulatory comparisons.
• The methodology section was rewritten to simplify the hypothesis structure, clarify sampling procedures, and justify the use of PLS-SEM and MGA.
• The results were updated to align with the revised model and hypotheses, and now include clear tables, path coefficients, mediation/moderation analyses, and interpretations.

 

We sincerely thank you for your constructive and motivating comments, which have significantly enhanced the clarity and rigor of the paper.

Warm regards,

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

I am happy that my concerns have been addressed in the revised manuscript. 

Reviewer 2 Report

Comments and Suggestions for Authors

 I appreciate your efforts in addressing the topic and recommend the paper for acceptance in its current form. No further revisions are necessary.