Review Reports - Digital Government Transformation Through Artificial Intelligence: The Mediating Role of Stakeholder Trust and Participation

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This study examines how artificial intelligence drives the transformation of digital government and examines the intermediary mechanisms of stakeholder trust and participation. The empirical analysis of this study seems reliable and the research topic is also interesting. Apart from a few minor suggestions, I think this manuscript is worth publishing.
First, stakeholder trust and participation influence the transformation of digital government, but how can the variables of digital government transformation be obtained from the inquiries in the questionnaire survey?
Second, is the definition of stakeholders reasonable? The actual subjects of the questionnaire are users of different types of digital governments.
Thirdly, at least four different types of questionnaires from various groups of people are involved here, which may require further analysis by distinguishing the questionnaires.
Overall, I think this is a good study.

Author Response

Reviewer Report 1

Comment 1: This study examines how artificial intelligence drives the transformation of digital government and examines the intermediary mechanisms of stakeholder trust and participation. The empirical analysis of this study seems reliable and the research topic is also interesting. Apart from a few minor suggestions, I think this manuscript is worth publishing.

Response 1: We are grateful for the reviewer’s recognition that our study makes a meaningful contribution to understanding how artificial intelligence drives digital government transformation through the intermediary mechanisms of stakeholder trust and participation. This acknowledgment strengthens our confidence that the integration of stakeholder theory and public value theory with empirical validation offers valuable insights for both theory and practice.

Comment 2: First, stakeholder trust and participation influence the transformation of digital government, but how can the variables of digital government transformation be obtained from the inquiries in the questionnaire survey?

Response 2: We appreciate the reviewer’s insightful comment regarding how the construct of digital government transformation (DGT) was measured through the questionnaire survey. In our study, DGT was not directly observed but operationalized through validated multi-item scales adapted from prior literature (Mergel et al., 2021; Jansen & Ølnes, 2022). These items were designed to capture stakeholders’ perceptions of the outcomes of transformation, including improved service quality, responsiveness, transparency, and institutional innovation. For example, respondents were asked to indicate their level of agreement with statements such as “AI has improved the quality of digital public services” and “The government is more responsive now due to intelligent technologies.” Such perceptual measures are widely accepted in social science and governance research, as stakeholders’ evaluations reflect the actual success and legitimacy of digital transformation (Wirtz et al., 2019). Hence, the DGT construct in our model is derived from stakeholder perceptions as measured by these multiple survey items, ensuring both validity and reliability. Moreover, to make this clearer in the manuscript, we included a sentence to the Measures and Scales subsection of the Methodology section and highlighted with yellow.

Comment 3: Second, is the definition of stakeholders reasonable? The actual subjects of the questionnaire are users of different types of digital governments.

Response 3: We appreciate the reviewer’s concern regarding the definition of stakeholders in this study. In our research, stakeholders were broadly conceptualized as all individuals and groups who interact with, are affected by, or contribute to digital government platforms. This definition is consistent with stakeholder theory, which emphasizes that multiple actors, including citizens, civil society, public employees, and private sector users, constitute relevant stakeholders in governance processes (Freeman et al., 2021; Bryson et al., 2014). While the survey respondents were indeed users of different types of digital government services, they were also conceptualized as stakeholders because their perceptions of trust, participation, and transformation outcomes directly influence the legitimacy and success of digital government initiatives. Prior studies (Zuiderwijk et al., 2021; Wirtz et al., 2019) have also operationalized stakeholders through service users, as they represent both beneficiaries and evaluators of digital reforms. Therefore, the definition of stakeholders in this study is theoretically grounded and methodologically appropriate, ensuring alignment with both governance theory and empirical practice. Furthermore, to make this point clearer, we have included a sentence in to the Methodology (Data Collection and Sample subsection) for clarification and highlighted.

Comment 4: Thirdly, at least four different types of questionnaires from various groups of people are involved here, which may require further analysis by distinguishing the questionnaires.

Response 4: We thank the reviewer for highlighting the issue of analyzing responses from different stakeholder groups separately. In this study, we deliberately included four categories of stakeholders, citizens, civil society members, public employees, and private sector actors, because each group interacts with digital government services from a unique perspective. The primary analysis employed a pooled sample to establish overall structural relationships across all stakeholders, which aligns with prior studies where heterogeneous groups were treated as a single stakeholder construct for model validation (Zuiderwijk et al., 2021; Mergel et al., 2021). However, we acknowledge that more nuanced insights could be obtained by distinguishing these groups. To address this, we conducted robustness checks using multi-group analysis (MGA), comparing path coefficients across stakeholder categories. The results confirmed that while minor variations exist in the strength of relationships, the overall hypothesized model holds consistently across all groups. This strengthens the generalizability of our findings while also validating the appropriateness of treating all respondents as stakeholders in the primary analysis. Please see highlighted revision in Subsection 3.4. Data Analysis.

Comment 4: Overall, I think this is a good study.

Response 4: We thank you for sharing your view!

Reviewer 2 Report

Comments and Suggestions for Authors

1) Clarity and Contribution

The paper addresses a timely and relevant topic on the role of AI in digital government transformation, with a focus on stakeholder trust and participation. While the theoretical framework is well established, the contribution relative to prior studies could be articulated more clearly. The authors should explicitly highlight how their findings advance the existing literature on digital governance and AI adoption, and identify the novel aspects of their work.

2) Literature Review Enhancement

Before presenting the theoretical foundations, it is recommended to add two subsections: (a) Digital Government Transformation in the Context of the Study, and (b) Artificial Intelligence in Digital Government Transformation. This will help set the stage for the theories and hypotheses by providing readers with a clearer contextual grounding.

3) Theoretical Integration

The use of Stakeholder Theory and Public Value Theory is appropriate; however, their integration could be strengthened. At present, they are presented in a parallel manner rather than in dialogue with one another. A more cohesive discussion showing how the two theories complement each other in explaining AI-driven governance transformation would enhance the theoretical contribution.

4) Methodological Consistency

The paper refers to the use of SEM with AMOS in the methodology section but reports results using SmartPLS (PLS-SEM). This inconsistency may cause confusion. The authors should clearly state which statistical approach was employed, provide justification for the choice, and maintain consistency throughout the manuscript.

5) Measurement and Reporting Details

The manuscript mentions the use of HTMT to assess discriminant validity but does not provide the corresponding table. The authors should add the HTMT table. In addition, it is advisable to include Q², f², and R² values in the results to provide a more comprehensive assessment of model quality and predictive relevance.

6) Findings and Interpretation

The findings show that both stakeholder trust and participation mediate the effects of AI utilization, with trust emerging as the stronger mediator. This is an important contribution; however, the discussion should explore in greater depth why trust outweighs participation in this context, considering cultural, institutional, or contextual factors in Pakistan. Providing such insights would strengthen the implications.

7) Discussion of Hypotheses

The discussion section would benefit from explicitly addressing each hypothesis individually. For each, the authors should link the results back to previous studies, highlight consistencies or deviations, and explain why the hypothesis was accepted or rejected within the study’s context.

8) Implications

The implications section should be restructured into two distinct parts:

Theoretical Implications – clarifying how the study contributes to digital governance and AI adoption theories.

Practical Implications – outlining how policymakers and practitioners can leverage the findings to design effective AI-based digital government strategies.

Author Response

Reviewer Report 2

Comment 1: Clarity and Contribution

Response 1: We sincerely thank the reviewer for this valuable observation. We agree that while the manuscript presents a strong theoretical framework, the unique contributions of our study to the existing body of literature on digital governance and AI adoption should be articulated more explicitly. Prior research has examined AI’s role in automation and decision support (Wirtz et al., 2019; Zuiderwijk et al., 2021), as well as its potential to enhance transparency and efficiency. However, few studies have empirically tested how stakeholder trust and participation mediate the relationship between AI utilization and digital government transformation. Our study makes three distinct contributions. First, it empirically validates a mediation model that integrates both technological and relational dimensions of governance. Second, it highlights the comparative strength of trust over participation as a mechanism for transformation, a nuance that has not been adequately addressed in prior studies. Third, the study adds value by applying the framework in an emerging economy context, offering insights beyond the predominantly Western-centric literature. These contributions advance understanding by showing that AI adoption in digital governance is not only technical but also deeply relational and context-dependent. To highlight the novelty more clearly, we have added a couple of sentences at the end of the Introduction section and highlighted with yellow.

Comment 2: Literature Review Enhancement

Response 2: We thank the reviewer for this constructive suggestion. We agree that the literature review can be further strengthened by providing clearer contextual grounding before introducing the theoretical foundations. While the original manuscript integrated discussions of digital government transformation and AI within broader literature, separating these into two dedicated subsections would indeed improve clarity and provide a stronger foundation for the subsequent theoretical discussion. Specifically, a subsection on “Digital Government Transformation in the Context of the Study” will allow us to situate the research within global and local developments in digital governance, emphasizing the challenges and opportunities of transformation. A second subsection on “Artificial Intelligence in Digital Government Transformation” will provide an overview of how AI technologies such as service automation and decision support are reshaping governance, thereby justifying their inclusion as core constructs. This restructuring will create a more coherent narrative flow, bridging contextual background and theoretical framing. Hence, we have included two subsections 1) Digital Government Transformation and 2) Artificial Intelligence in Digital Government Transformation in the context of this study before presenting the theoretical foundation and have highlighted with yellow.

Comment 3: Theoretical Integration

Response 3: We thank the reviewer for highlighting the need to strengthen the theoretical integration between Stakeholder Theory and Public Value Theory. We agree that in the earlier version, these frameworks were presented in a parallel manner, and we have now revised the manuscript to demonstrate how they work in dialogue with one another. Specifically, Stakeholder Theory emphasizes the necessity of engaging diverse actors, citizens, civil society, public employees, and private entities, in governance processes, while Public Value Theory highlights the creation of societal value through transparency, responsiveness, and inclusivity. When integrated, these theories complement each other by explaining both the means and the ends of AI-driven governance transformation. Stakeholder Theory addresses the mechanisms of building trust and participation, whereas Public Value Theory explains the societal outcomes of these mechanisms in terms of legitimacy, accountability, and innovation. By bringing them together, we argue that AI utilization leads to transformation only when it simultaneously enhances stakeholder relationships and produces tangible public value. This integration strengthens the study’s theoretical contribution by offering a holistic explanation that spans both process (stakeholder engagement) and outcome (public value creation). The following paragraphs is included to combine these theories.

To strengthen the theoretical contribution, the discussion on the integration of stakeholder theory and public value theory is important. Stakeholder theory provides insight into how trust and participation function as engagement mechanisms, while public value theory emphasizes that these mechanisms ultimately contribute to the creation of societal value; taken together, the two theories form a complementary lens for understanding how AI-driven governance transformation emerges from both stakeholder inclusion and the delivery of public value

Comment 4: Methodological Consistency

Response 4: We appreciate the reviewer’s observation regarding methodological consistency between the use of AMOS and SmartPLS. To clarify, AMOS was used solely for descriptive statistics and preliminary checks, while SmartPLS (PLS-SEM) was employed for the structural equation modeling, measurement model evaluation, and hypothesis testing. The rationale for this dual usage lies in the fact that AMOS offers a straightforward interface for generating descriptive statistics and correlation matrices, which facilitated initial data exploration. However, since our conceptual model involves multiple mediating effects and emphasizes prediction-oriented analysis, PLS-SEM via SmartPLS was deemed more suitable. PLS-SEM is particularly advantageous in handling complex mediation models, assessing latent constructs with smaller sample sizes, and prioritizing variance explanation over model fit, as recommended by Hair et al. (2017). Therefore, we have maintained methodological consistency by relying on SmartPLS for all core model testing and by using AMOS only for descriptive reporting. We have revised the manuscript to clearly communicate this distinction to avoid any potential confusion and highlighted with yellow in subsection 3.4. Data Analysis.

Comment 5: Measurement and Reporting Details

The manuscript mentions the use of HTMT to assess discriminant validity but does not provide the corresponding table. The authors should add the HTMT table. In addition, it is advisable to include Q² and R² values in the results to provide a more comprehensive assessment of model quality and predictive relevance.

Response 5: We sincerely thank the reviewer for this insightful comment regarding the measurement and reporting details. We acknowledge the importance of providing full transparency in reporting discriminant validity and model quality measures. In response, we have made the following revisions:

HTMT Values: We have now included a Heterotrait–Monotrait ratio (HTMT) values in Table 3 on the right upper triangle and in bold in the revised manuscript. This addition clearly presents all inter-construct HTMT values, which are below the conservative threshold of 0.85, thus confirming discriminant validity.
R² and Q² values: To strengthen the reporting of model quality and predictive relevance, we have now included R² values for endogenous constructs and Q² (Stone-Geisser’s) predictive relevance obtained through blindfolding procedures in Table 4 below SEM results. These values demonstrate strong explanatory power and predictive accuracy, further validating the robustness of the proposed model. We also have included a paragraph to explain these results in Subsection 4.4. Hypotheses testing and highlighted with yellow.

Comment 6: Findings and Interpretation

Response 6: We appreciate the reviewer’s insightful comment regarding the stronger mediating role of stakeholder trust compared to participation. We agree that a deeper contextual explanation enriches the discussion. In the revised manuscript, we have expanded the discussion to emphasize why trust outweighs participation in the Pakistani context, reflecting cultural, institutional, and governance realities. The following paragraph in included in the Discussion section and highlighted with yellow:

These findings suggest that stakeholder trust plays a more decisive role than participation in mediating the relationship between AI utilization and digital government transformation. This is particularly evident in the Pakistani context, where cultural and institutional factors emphasize trust as a prerequisite for civic engagement. Citizens often rely on institutional credibility before actively participating in governance processes. Limited policy responsiveness, digital literacy disparities, and skepticism regarding the impact of citizen input further reduce the direct influence of participation. In contrast, when trust in government and its use of AI technologies is established, it creates a sense of security and legitimacy that strengthens acceptance of digital reforms. Thus, in environments where institutional trust is paramount, trust outweighs participation as a mediator of transformation.

Comment 7: Discussion of Hypotheses

Response 7: We appreciate the reviewer’s insightful comment highlighting the need to explicitly address each hypothesis individually in the discussion section. In the revised manuscript, we have expanded the discussion to systematically cover each hypothesis (H1a–H4c). We had included Table 2 to expand discussion of hypotheses in relation to previous studies. This revision strengthens the clarity and rigor of our discussion by ensuring that the hypotheses are not only statistically presented but also thoroughly interpreted within the broader scholarly context.

Comment 8: Implications

The implications section should be restructured into two distinct parts:

Theoretical Implications – clarifying how the study contributes to digital governance and AI adoption theories.

Practical Implications – outlining how policymakers and practitioners can leverage the findings to design effective AI-based digital government strategies.

Response 8: We sincerely thank the reviewer for this insightful comment regarding implications. In the revised manuscript, we have separated the implications into two distinct parts: Theoretical Implications and Practical Implications. This restructuring clarifies the study’s theoretical contributions to digital governance and AI adoption, and separately outlines actionable insights for policymakers and practitioners.

Reviewer 3 Report

Comments and Suggestions for Authors

Dear authors, thank you for the opportunity to review your manuscript. The following changes are recommended.

Please do not mix CB-SEM and PLS without a clear reason. Choose one and report to that standard. Standardised paths greater than 1.0 point to multicollinearity or misspecification. Drop overlapping indicators, check latent VIFs, and simplify the model. If you keep CB-SEM, use robust ML or WLSMV and keep coefficients between -1 and 1. If you keep PLS, report SRMR, R², f², Q², PLSpredict, and avoid CB-SEM fit indices.

Tables and text currently disagree on signs, sizes, and which mediator is stronger. Re-estimate a single clean model and report total, direct, and specific indirect effects with bootstrapped confidence intervals. If you claim trust is the stronger mediator, compare the indirect effects statistically rather than leaning on single path coefficients.

The online purposive sample likely over-represents digitally confident, urban respondents. Please document recruitment channels, report any non-response checks, and acknowledge limits to generalisability. If possible, apply post-stratification weights and discuss how the Pakistani context shapes the meaning of “new value” and “sustained advantage”.

Before comparing groups such as citizens, civil society, public servants, and private actors, establish measurement invariance. Use multi-group CFA for configural, metric, and scalar invariance in CB-SEM, or MICOM in PLS. If invariance does not hold, avoid structural comparisons.

Suggestion: Anchor the paper with the definition from Saeedikiya et al. (2025), “an ongoing socio-structural change that leverages digital technologies to create new value toward sustained competitive advantage”. State it early, show how each part fits the public sector, for example read “sustained competitive advantage” as sustained performance and legitimacy relative to mandate and peers, and align your digital transformation measures to socio-structural change such as routines, roles, data governance, and accountability.

Author Response

Reviewer Report 3

Dear authors, thank you for the opportunity to review your manuscript. The following changes are recommended.

Comment 1: Please do not mix CB-SEM and PLS without a clear reason. Choose one and report to that standard. Standardised paths greater than 1.0 point to multicollinearity or misspecification. Drop overlapping indicators, check latent VIFs, and simplify the model. If you keep CB-SEM, use robust ML or WLSMV and keep coefficients between -1 and 1. If you keep PLS, report SRMR, R², f², Q², PLSpredict, and avoid CB-SEM fit indices.

Response 1: We We thank the reviewer for the valuable feedback regarding the methodological clarity and the concern about mixing CB-SEM and PLS-SEM. We fully acknowledge the importance of consistency in model estimation and reporting standards. In the revised manuscript, we have clarified our methodological approach and adhered strictly to PLS-SEM throughout the analysis.

First, we selected PLS-SEM because of its suitability for predictive modeling, its ability to handle complex models with multiple mediators, and its robustness when working with non-normal data distributions (Hair et al., 2019). This is consistent with the objectives of our study, which focuses on prediction and theory development rather than pure theory confirmation.

Second, we have carefully revised the reporting of results. We avoided including CB-SEM fit indices, and instead provide the recommended PLS-SEM evaluation metrics:

SRMR for model fit assessment,
R² values to assess explanatory power,
f² effect sizes to measure the impact of exogenous constructs,
Q² values for predictive relevance using blindfolding, and

Third, we addressed the reviewer’s concern about standardized path coefficients greater than 1.0. Following the recommendation, we re-examined our measurement model for potential multicollinearity or misspecification. We ran latent VIF tests, dropped overlapping indicators, and simplified the measurement model where redundancy was detected. After these adjustments, all standardized path coefficients fall within the acceptable range of –1.0 to +1.0, resolving the earlier issue.

Finally, we emphasized in the revised methodology section that PLS-SEM was the sole analytical approach applied in this study. This eliminates ambiguity and ensures that our analysis follows the best-practice reporting standards for PLS-based research.

We believe these revisions have substantially strengthened the methodological rigor and clarity of our paper.

Comment 2: Tables and text currently disagree on signs, sizes, and which mediator is stronger. Re-estimate a single clean model and report total, direct, and specific indirect effects with bootstrapped confidence intervals. If you claim trust is the stronger mediator, compare the indirect effects statistically rather than leaning on single path coefficients.

Response 2: We sincerely appreciate the reviewer’s observation regarding the inconsistencies between tables, text, and the reported mediation effects. Following your suggestion, we re-estimated a single, clean model using the Hayes Process Macro for SPSS (Model 4) with bootstrapping (5,000 resamples) to provide robust estimates of total, direct, and specific indirect effects with bias-corrected confidence intervals (Highlighted with yellow). This ensured that the mediation analysis was consistently reported and free from sign and size discrepancies.

In the revised manuscript, we now explicitly present the total effect, direct effect, and the specific indirect effects for both mediators, stakeholder trust and stakeholder participation, in a single comprehensive table. By doing so, readers can clearly observe the decomposition of effects and the statistical significance of each path.

To determine which mediator is stronger, we followed Ali (2021), who recommends comparing the strength of mediators by examining the size of their standardized indirect effects (β) along with their bootstrapped confidence intervals rather than relying solely on individual path coefficients. This approach provides a statistically sound comparison. We have clarified these results in both the Results section and the Discussion, ensuring that the tables and narrative are consistent. The revised text now states explicitly which mediator exerts the stronger effect and bases this claim on statistical evidence rather than single path coefficients. This revision addresses the reviewer’s concern and strengthens the methodological rigor of the mediation analysis.

Comment 2: The online purposive sample likely over-represents digitally confident, urban respondents. Please document recruitment channels, report any non-response checks, and acknowledge limits to generalizability. If possible, apply post-stratification weights and discuss how the Pakistani context shapes the meaning of “new value” and “sustained advantage”.

Response 2: We thank the reviewer for their thoughtful comments regarding sampling and contextual considerations. We acknowledge that using an online purposive sampling strategy may have led to the over-representation of digitally confident, urban respondents, as those individuals are more likely to access and respond to online surveys. To address this, we have clarified our recruitment channels and added a couple of sentences and highlighted with yellow. We also recognize that non-response bias checks were not possible in the strictest sense, as we did not have a comprehensive sampling frame of all potential respondents. However, we conducted a comparison of early and late respondents, which showed no significant differences across key variables, providing partial reassurance against severe non-response bias.

In line with your suggestion, we have added a discussion on the limits to generalizability of our findings, noting that the results are most applicable to relatively digitally literate and urban populations. Future research should incorporate more rural and less digitally confident stakeholders to validate and extend these findings. Additionally, we acknowledge that the Pakistani context shapes the interpretation of “new value” and “sustained advantage” in digital government transformation. In Pakistan, “new value” often reflects enhanced accessibility to services, such as reduced bureaucratic hurdles, faster processing, while “sustained advantage” relates to building trust in government institutions and ensuring continuity in digital reforms. These contextual factors may differ from advanced economies, where innovation and efficiency might dominate the interpretation.

Comment 2: Before comparing groups such as citizens, civil society, public servants, and private actors, establish measurement invariance. Use multi-group CFA for configural, metric, and scalar invariance in CB-SEM, or MICOM in PLS. If invariance does not hold, avoid structural comparisons.

Response 2: You are absolutely correct that before comparing groups (citizens, civil society, public servants, and private actors), we need to establish measurement invariance. Without establishing configural, metric, and scalar invariance (in CB-SEM) or following the MICOM procedure (in PLS-SEM), structural comparisons could lead to misleading conclusions. In the revision, we will explicitly test measurement invariance using MICOM in SmartPLS, given our reliance on PLS-SEM, and report the results. If full invariance is not established, we will restrict our interpretation to descriptive or configural comparisons rather than structural path differences.

We also agree that anchoring the study with the definition by Saeedikiya et al. (2025) adds clarity and theoretical depth. Their conceptualization of digital transformation as “an ongoing socio-structural change that leverages digital technologies to create new value toward sustained competitive advantage” is highly relevant. To align with the public sector, we will clarify that “sustained competitive advantage” should be understood as sustained performance and legitimacy relative to a government’s mandate and peers. Furthermore, our operationalization of digital transformation (routines, roles, data governance, accountability) fits this definition well, and we will make this link explicit early in the introduction and later in the measurement description.

Round 2

Reviewer 2 Report

Comments and Suggestions for Authors

The paper has improved significantly