Review Reports - Social Influence and Prospective Adoption of ORA and REDCIA in Amazonian Cooperation

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

Reviewer Report – Sustainability Journal (MDPI)

Manuscript Titlle: Determinants of technology adoption in Amazonian cooperation: a prospective analysis of intentions to use ORA and REDCIA using an extended UTAUT model with a KAP approach
Manscript ID: sustainability-4194526
Recommendation: Major Revision

Comments to the Authors

This manuscript examines prospective technology adoption within the OTCA context using a UTAUT sub-model operationalized through a KAP survey. The focus on ORA and REDCIA during their pre-implementation phase is timely and relevant, and the empirical setting is original.

However, while the technical execution is competent, the manuscript presents notable conceptual, measurement, and methodological weaknesses. The theoretical contribution is overstated relative to the empirical validation, and several key statistical diagnostics require clearer reporting and interpretation. Substantial revision is necessary to strengthen rigor and analytical precision..

Revision Suggestions to the Authors

Recalibrate the Theoretical Contribution”

Problem

The manuscript presents the integration of KAP and UTAUT as a methodological innovation. However, the empirical implementation relies on ex post mapping of survey items onto UTAUT constructs without formal validation.

Required Revision

Clarify whether this is:

A conceptual integration, or
A pragmatic operational alignment.

If claiming theoretical integration:

Report construct reliability (e.g., Cronbach’s alpha).
Provide factor analysis (EFA or CFA) demonstrating that KAP items load onto the intended UTAUT constructs.
Justify aggregation procedures more rigorously.

If not providing validation:

Tone down the claims regarding “novel hybridization” and present the approach as an exploratory operational adaptation.

Right now, the theoretical ambition exceeds the empirical demonstration.

Address the Ceiling Effect More Rigorously

Problem

Descriptive statistics show very high means and strong negative skewness for Behavioral Intention and Performance Expectancy (Tables 6–7). This indicates substantial range restriction.

This has major implications:

Reduced variance
Suppressed regression effects
Inflated significance of variables that retain minimal variation

Required Revision

Explicitly discuss range restriction bias.
Clarify how limited variance may have affected the non-significance of PE.
Avoid interpreting non-significance as theoretical irrelevance.
Consider running robustness checks:

Dichotomized BI (High vs Moderate/Low)
Alternative model specifications

Without this, the conclusion that “Social Influence dominates” is overstated.

Clarify Ordinal Model Interpretation

Problem

The regression coefficients for Social Influence are negative (Tables 9–10), yet the interpretation suggests increasing SI raises probability of higher BI.

This may be correct under cumulative logit parameterization — but it is not clearly explained.

Required Revision

Explicitly state how categories were ordered.
Clarify interpretation of coefficient signs.
Report threshold parameters.
Ideally, provide marginal effects or predicted probabilities to improve transparency.

At present, statistically literate readers may question the interpretation.

Test and Report Proportional Odds Assumption

Problem

The manuscript uses ordinal logistic regression but does not report diagnostics for the proportional odds assumption.

Given:

Skewed distributions
Quasi-separation
Penalized estimation

Testing this assumption is essential.

Required Revision

Report Brant test (or equivalent).
If violated, consider:

Partial proportional odds model
Generalized ordered logit

This is a methodological requirement, not optional refinement.

Strengthen the Discussion of Sampling Limitations

Problem

The sample is purposive and non-probabilistic. Complete cases reduce observations further.

The manuscript draws policy-level implications that exceed the inferential scope of the data.

Required Revision

Explicitly acknowledge limits to generalizability.
Reframe conclusions as exploratory institutional diagnostics.
Avoid language implying structural determinants of adoption across OTCA.

This will increase credibility rather than weaken the paper.

Reconsider the Interpretation of Non-Significant PE and FC

Problem

The discussion interprets non-significant PE and FC as indicating social primacy in early adoption stages.

However:

High PE with low variance limits discrimination.
FC measured as “number of barriers” may not capture perceived capability.
The non-significance may be measurement-driven rather than theoretically meaningful.

Required Revision

Clarify that lack of significance may reflect:

Measurement compression
Operationalization choices
Pre-implementation cognitive idealization

Avoid strong theoretical claims based solely on null effects.

Improve Measurement Transparency

Specifically:

Explain why PE was divided into three sub-dimensions but SI was not.
Justify aggregation method (simple mean vs weighted).
Explain why FC is measured as a count of barriers (which reverses conceptual direction).
Clarify recoding logic in greater detail.

The measurement section needs more precision to match the sophistication of the theoretical framing.

Tighten the Discussion Section

The discussion currently reads as highly confident in theoretical implications.

To strengthen it:

Distinguish clearly between:

Empirical findings
Interpretive speculation
Theoretical extension

Reduce rhetorical intensity around “institutional isomorphism” unless directly tested.
Emphasize exploratory nature of ex ante diagnostics.

Subtlety will improve scholarly credibility.

Optional but Strongly Recommended Enhancements

To substantially improve the manuscript:

Provide marginal effects plots.
Report model fit indices beyond AIC (pseudo R²).
Include robustness check using binary logistic model.
Address potential common-method bias.

These additions would elevate the work from competent applied modeling to rigorous empirical contribution.

Final Evaluation Summary for Authors

This manuscript addresses an important and original institutional context and demonstrates solid technical competence. However, it currently:

Overstates theoretical innovation,
Underreports statistical diagnostics,
And overinterprets null findings.

With substantial revision and methodological clarification, it has strong potential to make a valuable contribution to sustainability governance and technology adoption research.

Author Response

I sincerely appreciate your careful reading of the manuscript and the comments provided. Your observations have been highly valuable in strengthening the article’s conceptual clarity, methodological precision, and argumentative consistency. In the revised version, I have addressed each comment in detail by incorporating changes to the title, abstract, theoretical foundation, methodological explanation, interpretation of results, and delimitation of the study’s inferential scope. In particular, I have clarified that the manuscript should be understood as an ex ante relational model, rather than an individual predictive model, which has allowed me to reformulate several analytical decisions and their interpretation more accurately. Below, I respond point by point to each observation and indicate the corresponding changes made in the revised manuscript.

1. Theoretical contribution.- Thank you for this important observation. We have explicitly recalibrated the manuscript’s theoretical contribution. In the revised version, the articulation between KAP and UTAUT is no longer presented as a validated theoretical integration, but rather as an operational and exploratory alignment in an ex ante setting. Accordingly, we replaced wording that could suggest a validated “hybridization” with a more cautious formulation, clarifying that KAP items function as analytical proxies and that the findings should be interpreted as exploratory evidence of conceptual fit and preliminary analytical usefulness, rather than as psychometric validation of a hybrid KAP–UTAUT model. This directly addresses your concern regarding the imbalance between theoretical ambition and empirical demonstration.

2. Ceiling effect.- Thank you for highlighting this issue. We have now incorporated an explicit discussion of the range restriction / ceiling effect observed in BI and PE. In the revised version, we explain that the high means and strong negative skewness compress the available variance, reduce the discriminatory power of PE, and may cause variables with greater relative dispersion—such as SI—to appear comparatively more influential. Accordingly, PE’s non-significance is no longer interpreted as theoretical irrelevance, but also as a likely consequence of measurement compression in an ex ante phase characterized by broadly shared high expectations. This allowed us to moderate the original conclusion that social influence “dominates” for purely substantive reasons.

3. Ordinal model interpretation.- We appreciate this observation. We improved the explanation of the penalized cumulative ordinal model. In the revised version, we explicitly state the ordering of the dependent variable categories, clarify that negative SI coefficients imply a lower probability of belonging to lower categories and therefore a higher relative probability of belonging to higher categories, and we report odds ratios and 95% confidence intervals. In addition, because the manuscript is framed as a relational rather than predictive model, we clarify that thresholds or cut-points are not the main interpretive focus of the analysis. They are part of the ordinal specification, but the substantive interest of the study lies in the direction, magnitude, and significance of the associations, not in the individual prediction of probabilities.

4. Proportional odds assumption.- Thank you for this methodological observation. In the revised version, we explain more precisely why the Brant test is not appropriate in the present case. The models were estimated using penalized ordinal logistic regression to handle quasi-separation; therefore, it would not be methodologically consistent to apply a test designed for standard ordinal models estimated by non-penalized maximum likelihood. Instead, we assessed parallelism by comparing a parallel-effects specification with an alternative without that restriction within the same penalized framework. Although the non-parallel model failed to converge, this is consistent with the structure of the data and reinforces that the parallel specification is the most stable and parsimonious option in this study.

5. Sampling limitations.- We appreciate this point. We have strengthened the discussion of the study’s inferential scope. The revised version now explicitly states that the sample is purposive and non-probabilistic, composed of institutional actors and experts linked to the Amazonian ecosystem, and that the findings cannot be generalized to all OTCA stakeholders or to the Amazonian population as a whole. Likewise, the conclusions were reframed as an exploratory institutional diagnostic of this specific analytical sample. We believe this adjustment improves the manuscript’s credibility and aligns it with the actual research design.

6. Non-significant PE and FC.- Thank you for this observation. We have reformulated the interpretation of PE and FC in order to avoid excessive theoretical conclusions. In the revised version, we clarify that PE’s non-significance may be related to measurement compression derived from the ceiling effect, whereas FC’s operationalization as a count of barriers may not capture the full conceptual scope of facilitating conditions in UTAUT. Thus, the absence of statistical significance is no longer presented as evidence of substantive absence of effect, but rather as a result conditioned by the ex ante design, by operationalization choices, and by the structure of variation observed in the sample.

7. Measurement transparency.- We appreciate this comment. We expanded the measurement section to explain more clearly why PE is divided into three subdimensions while SI is not, why a simple mean was used for PE, and why FC was operationalized as barriers. We also incorporated an additional methodological clarification: PE, SI, and FC are not assumed to be reflective constructs, but rather formative/compositional composites built from conceptually relevant KAP items. Under this logic, Cronbach’s alpha, EFA, or CFA are not appropriate central criteria. Instead, the validity of the approach rests on the conceptual justification of the mapping, the transparency of recoding, and the inferential consistency of the results.

8. Discussion tone and optional enhancements.- Thank you for this observation. We have moderated the tone of the discussion by distinguishing more clearly between empirical findings, plausible interpretations, and theoretical extensions. In particular, we reduced overly conclusive expressions and reinforced the exploratory nature of the ex ante diagnostic. We also added additional fit indicators—such as pseudo R² alongside AIC—and expanded the robustness discussion with contextual controls. However, given the relational and exploratory nature of the study, we did not incorporate marginal effects, individual predictions, or alternative binary models as the core of the analysis, since those would correspond to a predictive logic different from the one adopted here.

Reviewer 2 Report

Comments and Suggestions for Authors

Thanks for inviting me to review the manuscript entitled “Determinants of technology adoption in Amazonian cooperation: a prospective analysis of intentions to use ORA and REDCIA using an extended UTAUT model with a KAP approach”. This manuscript investigates the prospective adoption intentions of two Amazonian regional cooperation technologies by adapting UTAUT model with a KAP approach, using survey data from 162 stakeholders across OTCA member countries. The topic is highly relevant to intergovernmental cooperation and digital governance in sustainability contexts, addressing a critical gap in pre-implementation technology adoption research for complex regional networks. The methodological integration of UTAUT and KAP is innovative, and the penalized ordinal logistic regression analysis addresses statistical challenges like quasi-separation. However, substantial gaps in theoretical contextualization, empirical analysis depth, and practical implication specificity require more rigorous revision.

The title concludes the main information, but it is too long and not insightful, maybe it can be changed as “Amazonian Tech Adoption: Social Norms Outweigh Utility in ORA & REDCIA” or “Social Influence Dominates: Adoption Intent for Amazon’s ORA & REDCIA”. Besides, the abstract can be restructured as “research signficance-data and method-findings-implications”.

The introduction outlines the research context but lacks three critical elements: first, explicit differentiation between ORA and REDCIA’s functionalities and user groups beyond brief table descriptions; second, a comprehensive justification for excluding Effort Expectancy (EE) that engages with counterarguments from UTAUT literature; third, insufficient articulation of how Amazonian cooperation’s unique challenges shape technology adoption. Moreover, the practical contribution section is broad, with no clear link between findings and OTCA policies. Therefore, please add a subsection detailing ORA and REDCIA’s goals, user segments, and operational models to justify separate analyses; Expand EE exclusion rationale by engaging with counterevidence to test its power; Integrate Amazonian-specific contextual factors to frame the research as addressing context-specific adoption challenges; Refine practical contributions to link to OTCA’s strategic priorities.

The theoretical framework synthesizes UTAUT and KAP but suffers from three limitations: first, the integration of KAP with UTAUT constructs is underdeveloped; second, the UTAUT literature review is outdated; third, the framework fails to address potential moderators that UTAUT literature identifies as critical in interorganizational settings. Hence, the authors should strengthen KAP-UTAUT integration by citing studies that validate each mapping and include a visual model illustrating how KAP dimensions operationalize UTAUT constructs; Update literature review to incorporate recent UTAUT extensions and meta-analyses to discuss why additional constructs are irrelevant or should be included; Justify redefining FC as “barriers” by referencing UTAUT adaptations in pre-implementation contexts and clarify how this aligns with the original construct’s intent; Add a subsection discussing potential moderators and why they were excluded or included as controls.

The methodology is rigorous but has four flaws: first, the sampling strategy is poorly justified, yet the manuscript provides no details; second, construct operationalization lacks transparency, it does not report factor analysis results; third, the handling of missing data is problematic—11.7% missing cases may introduce bias; fourth, the model specification does not account for potential multicollinearity suggest moderate collinearity. The authors should: Enhance sampling transparency by detailing recruitment criteria, OTCA’s role in participant outreach, and a comparison of sample characteristics to the target population; Add CFA results to validate the factor structure of PE subdimensions and report composite reliability, AVE, and HTMT values to confirm construct validity; Address missing data by conducting multiple imputation and comparing results to complete case analysis; Report VIF values for all predictor variables to rule out multicollinearity, and if collinearity exists, use orthogonalization or combine highly correlated PE subdimensions into a single construct.

The results section presents key findings but lacks depth in three areas: first, the descriptive analysis does not explore differences in BI, PE, SI, or FC across subgroups; second, the interpretation of penalized regression coefficients is misleading; third, the robustness analysis with control variables is underreported—Table 11 includes coefficient estimates but no discussion. The manuscript fails to address the ceiling effect in PE beyond descriptive observations, without testing if this effect varies by subgroup. Theauthors should: Add subgroup descriptive analyses comparing BI, PE, SI, and FC across institutional types and countries; Clarify the SI coefficient by explaining regression’s threshold logic; Expand robustness analysis by reporting model fit statistics for models and discuss how contextual factors influence results; Test for differential ceiling effects in PE by subgroup.

The discussion interprets key findings but is limited by three gaps: first, “Performance Expectancy Paradox” is under-theorized; seond, practical implications for ORA and REDCIA are generic; third, discussion fails to address “intention–behavior gap” in depth. The link between findings and Amazonian cooperation’s challenges is superficial, with no discussion of how SI can be leveraged to address cross-border trust issues. Theauthors should: Expand PE paradox discussion by testing alternative explanations and discussing how PE may become a significant predictor; Refine practical implications to be mechanism-specific; Address the intention–behavior gap by proposing pre-implementation actions to resolve FC barriers identified in the survey; Link findings to Amazonian-specific challenges by discussing how SI can foster cross-border trust to support data sharing in ORA/REDCIA.

The conclusions summarize findings but have two limitations: first, the methodological contribution of integrating KAP and UTAUT is overstated; second, the future research directions are vague but lacks specificity. The limitations section understates sampling and generalizability concerns, failing to acknowledge how purposive sampling may bias results. Therefore, the authors should: Strengthen methodological contribution by comparing the proposed KAP-UTAUT model to a standard UTAUT model using model fit statistics to demonstrate superior predictive power; Refine future research directions to be specific; Expand limitations to discuss generalizability constraints and propose solutions for future research; Link conclusions to OTCA’s long-term goals by emphasizing how SI-focused implementation can support regional sustainability and intergovernmental collaboration.

The structureis basiclly correct, but some sections should be moved, especially the implications. The implications should be moved to the subsection after the conclusions. Moreover, there is only one subsection 6.1 in the conclusion section, which is strange. The section 6 should be consist of the three subsections, such as conclusion, implication, limitation and research direction.

The references in this manuscript is relatively comprehensive, but some of them are out of date and in low quality, which should be improved. The authors need to make sure 70% of the references are published in recent five years and most of the references are from SSCI/SCI journal papers.

Comments on the Quality of English Language

the quality of English language should be improved.

Author Response

1. Title and abstract.- Thank you for this suggestion. In response, we shortened the title in order to improve its clarity and analytical focus, highlighting the study’s central finding while reducing the excessive length of the original title. We also restructured the abstract following a clearer logic of research significance – data and method – findings – implications, emphasizing that the manuscript examines the prospective intention to adopt ORA and REDCIA through an ex ante relational model, and that social influence emerges as the only statistically significant predictor across both mechanisms. This adjustment aims to make the main finding and the manuscript’s methodological specificity more visible from the outset.

2. Introduction, ORA/REDCIA differentiation, EE exclusion, Amazonian context.- Thank you for this observation. We strengthened the introduction and Section 1.1 to more clearly differentiate the two mechanisms under analysis. The revised version explicitly distinguishes ORA’s intergovernmental level from REDCIA’s collaborative/scientific level, as well as their different institutional logics. We also expanded the rationale for excluding Effort Expectancy, incorporating more developed methodological and contextual arguments. Finally, we reinforced the Amazonian contextualization and linked the practical contribution more clearly to regional governance challenges and to implementation decisions within OTCA/OCTA.

3. KAP–UTAUT integration, updated literature, moderators, FC as barriers.- We appreciate this comment. We updated the theoretical review with more recent literature and clarified that the KAP–UTAUT linkage should be understood as a heuristic and exploratory approximation, rather than as a closed equivalence. We also strengthened the justification for treating FC as “barriers” by clarifying that it functions as an operational inverse of facilitating conditions, suitable for an ex ante study based on perceived obstacles. Regarding moderators, we chose not to incorporate them as formal constructs due to parsimony and sample-size considerations; instead, we explain that country and institution type were included as controls to assess the contextual stability of the estimated pattern.

4. Methodology: sampling, factor analysis, missing data, multicollinearity.- Thank you for this observation. In the revised version, we improved sampling transparency by detailing the secondary-data nature of the dataset, the profile of included stakeholders, and the purposive/non-probability nature of the sample. We also explicitly acknowledge the limitation of the complete-case analysis and the possible bias if non-response were associated with contextual or attitudinal variables. We did not include CFA or alpha/AVE/HTMT, because the measurement strategy was explicitly redefined in formative/compositional rather than reflective terms. Regarding multicollinearity, we added Spearman correlations among PE dimensions and clarified that the observed associations do not suggest severe collinearity. We nevertheless acknowledge that multiple imputation and VIF were not performed, and we state this as a methodological limitation.

5. Results: subgroup analyses, SI coefficient, robustness with controls.- We appreciate this observation. We clarified the interpretation of SI coefficients within the framework of a penalized cumulative ordinal regression and expanded the robustness discussion by incorporating odds ratios, confidence intervals, AIC, pseudo R², and a more explicit discussion of models with country and institution-type controls. We acknowledge that we did not add subgroup descriptive analyses or subgroup-specific ceiling-effect tests, because the study’s primary focus is relational rather than subgroup-comparative. However, the inclusion of contextual controls shows that the main pattern—the centrality of SI—remains stable across both specifications.

6. Discussion and practical implications.- Thank you for this observation. We expanded the discussion of the PE “paradox”, emphasizing that its lack of significance does not imply lack of substantive relevance, but rather reflects a combination of ceiling effects, measurement compression, and the pre-implementation stage. We also made the practical implications more specific for ORA and REDCIA, highlighting that future implementation should rely less on purely technical assumptions and more on processes of institutional legitimation, champions, mandates, and collaborative networks. Finally, we more clearly acknowledged the intention–behavior gap and proposed that post-implementation longitudinal research will be necessary to assess whether PE and FC gain explanatory relevance at the stage of actual use.

7. Conclusions, structure, and references.- We appreciate this suggestion. We moderated the claim regarding the methodological contribution and strengthened the limitations and future research agenda, avoiding any presentation of the KAP–UTAUT approach as a superior predictive validation. Regarding structure, the final section was reorganized to more clearly connect conclusions with practical implications and future directions. In addition, the references were updated with more recent and directly relevant studies concerning UTAUT, technology adoption in public settings, and collaborative platform integration. We acknowledge that the share of recent literature could still be further improved, but the revised version already represents a substantial improvement over the original manuscript.

Reviewer 3 Report

Comments and Suggestions for Authors

The paper examines determinants of prospective intention to use two Amazonian regional knowledge mechanisms, ORA and REDCIA, by adapting UTAUT to a pre-implementation setting and operationalizing constructs through a KAP-based survey. Using a purposive, non-probability sample of 162 respondents and complete-case samples of 144/143, the authors estimate penalized ordinal logistic regression models for behavioral intention and conclude that social influence is the only robust predictor across both mechanisms, while performance expectancy and facilitating conditions have limited explanatory value, likely due to ceiling effects. The topic is relevant, and the attempt to study technology adoption in a multilateral Amazonian governance setting is interesting and potentially valuable. The manuscript’s main contribution is the effort to extend UTAUT to a context where the systems were not yet operational and to compare an intergovernmental mechanism with a collaborative research network.
Drawbacks:
1. The paper claims to integrate KAP with UTAUT, but the operational mapping is weakly justified and, in some places, conceptually mismatched. In particular, practices are mapped to social influence, knowledge to facilitating conditions, and attitudes to performance expectancy and behavioral intention. This is not self-evident and needs much stronger theoretical and empirical support. Practices such as information-searching or sharing behavior may reflect prior behavior, professional role, or engagement. However, they are not equivalent to normative pressure from important others, which is the core meaning of UTAUT social influence. Similarly, counting perceived barriers as a proxy for facilitating conditions compresses a multidimensional construct into a crude count and reverses its substantive meaning. Without evidence from reliability analysis, factor analysis, or validation against the original survey structure, it is difficult to conclude that the estimated models actually test UTAUT constructs rather than serve as loosely related proxies.
2. The manuscript does not report internal consistency, inter-item correlations beyond a limited collinearity check among PE dimensions, exploratory or confirmatory factor analysis, item loadings, ordinal alpha, or any other evidence that the grouped KAP items form valid latent dimensions. This is especially problematic because the manuscript’s main methodological claim is that KAP can be used to operationalize UTAUT in ex ante settings. That claim cannot be supported without psychometric evidence. At present, the study is closer to an ad hoc recoding exercise than a validated hybrid framework.
3. The study uses an intentional non-probability sample drawn from heterogeneous institutions across countries and languages. That may be acceptable for exploratory work, but the paper sometimes presents its findings beyond the design’s scope. Because the sample is purposive and likely elite- or network-biased, the very high means on behavioral intention and performance expectancy may partly reflect recruitment strategy rather than true regional sentiment. The authors should be much more explicit that the analysis supports inference only for this analytical sample, not for OTCA stakeholders in general.
4. In both ORA and REDCIA models, the estimated coefficients for SI are negative (−1.40 and −2.69). However, the text interprets them as indicating that a higher SI increases the probability of being in higher BI categories. That may be true depending on outcome coding, threshold parameterization, or reverse coding of SI, but the manuscript does not explain this clearly. As written, the results are confusing and appear internally inconsistent. The same problem affects the discussion, where SI is presented as a positive driver despite negative coefficients. This needs a careful clarification of variable coding, the cumulative-logit direction, and threshold interpretation. Otherwise, readers cannot verify the estimates’ substantive meaning.
5. The authors appropriately note imbalance and quasi-separation and therefore use penalized ordinal logistic regression, which is reasonable. However, there is no report of the proportional odds assumption, no sensitivity analysis using alternative ordinal formulations, and no explanation of which penalization method and package was used. Because the dependent variables are highly skewed toward the top category, model estimates may be unstable and substantively fragile. Reporting only coefficients and p-values is insufficient; confidence intervals, threshold estimates, and fit diagnostics should be included.
6. The descriptive tables indeed show means close to the upper limit and strong negative skew for PE and BI, suggesting limited variance. However, once ceiling effects are this pronounced, the correct conclusion is not simply that PE is high, but that it is non-discriminatory. The survey instrument may have poor discriminatory capacity for this setting. This is a measurement limitation, not merely a substantive insight. The study’s main conclusion about the primacy of SI, therefore, needs to be tempered, because SI may appear dominant partly because it retained more variance than the other predictors.
7. The paper uses complete-case analysis after removing about 11.7% of cases and asserts that the sample size remains adequate because there are more than 10 observations per predictor. That rule of thumb is not enough, especially for skewed ordinal outcomes, clustered multilingual responses, and quasi-separated data. There is no assessment of whether missingness was random, no comparison of retained vs. excluded cases, and no robustness analysis using multiple imputation or inverse probability methods. This omission may bias estimates if missingness is associated with country, institution type, or technology attitudes.
8. Table 1 presents ORA as more mandatory and institutional and REDCIA as more voluntary and collaborative, but the analysis does not really exploit this contrast. The two models are estimated separately, yet there is no formal comparison of coefficients, no interaction terms, and no test of whether SI is significantly stronger in one mechanism than the other. Since the paper claims the two mechanisms represent distinct levels of governance, a stronger comparative strategy would substantially improve the contribution.
9. The manuscript refers to respondents from eight countries, including French Guiana, while the acknowledgment mentions 170 opinion leaders from the nine Amazon basin countries. French Guiana is not an OTCA member state, and the paper reports a sample size of 162 rather than 170. These discrepancies may reflect differences between the original project and the analytic subset, but this must be clarified precisely. There are also several language and editorial issues throughout the manuscript, including grammar, repetition, and some awkward phrasing.

Author Response

1. Weak conceptual mapping between KAP and UTAUT.- We appreciate this observation, which we consider central. In the revised version, we explicitly acknowledge that the mapping between KAP dimensions and UTAUT constructs should be understood as a heuristic and pragmatic approximation, not as a strict equivalence. In particular, we clarify that observed practices may reflect collaborative habits and environmental expectations, although they do not perfectly coincide with the normative pressure implied by UTAUT’s original SI construct; likewise, knowledge and perceived barriers only partially approximate the notion of FC. For this reason, we abandoned any claim of closed equivalence and redefined these variables as analytical proxies suitable for an ex ante relational diagnostic.

2. Lack of psychometric validation.- Thank you for this comment. We addressed this concern through a fundamental methodological clarification. The revised version states that the composites used in the study were not treated as reflective constructs, but rather as formative/compositional composites built from conceptually relevant KAP items. Consequently, we do not consider Cronbach’s alpha, EFA, or CFA to be appropriate central validation criteria. Rather than claiming a psychometric validation that the study cannot provide, we chose to be methodologically explicit and honest in stating that the approach relies on the conceptual justification of the mapping, the transparency of recoding, and the consistency of the inferential results.

3. Inferential scope and generalizability.- We appreciate this observation. The revised version explicitly strengthens the inferential delimitation of the study. It now states that the sample is purposive, non-probabilistic, and restricted to a specific analytical sample of institutional actors and experts linked to the Amazonian ecosystem. Therefore, the findings are presented as an exploratory institutional diagnostic, rather than as evidence generalizable to all OTCA/ACTO stakeholders or to the Amazonian population as a whole. We believe this adjustment improves the argumentative validity of the manuscript.

4. Negative SI coefficients.- Thank you for highlighting this interpretive inconsistency. In the revised version, we corrected and expanded the explanation of SI’s negative coefficients. We specify that, in a cumulative ordinal logistic model, a negative coefficient implies a reduction in the probability of belonging to lower categories of the dependent variable and, consequently, a greater relative probability of belonging to higher categories. This clarification is accompanied by the reporting of odds ratios and confidence intervals, so that the substantive interpretation is fully aligned with the parameterization of the estimated model.

5. Penalized model, Brant, thresholds.- We appreciate this methodological observation. We strengthened the explanation of the use of penalized ordinal logistic regression via bracl/brglm2, as well as the justification of why the Brant test is not appropriate in this context. We also added a specific clarification regarding cut-points/thresholds, explaining that, although they are part of the ordinal specification, they are not the main interpretive focus of the study because the objective is not individual prediction, but rather the examination of statistical associations in an ex ante relational model. Accordingly, the analysis focuses on the estimators, their significance, their ORs, and their confidence intervals.

6. PE and FC interpretation.- Thank you for this observation. We revised the discussion in order to avoid overinterpreting the non-significance of PE and FC. The revised version emphasizes that PE shows high means and low variability, which limits its discriminatory power, while FC—being operationalized as a count of barriers—represents an operational inverse of the construct rather than a full capture of the classical notion of facilitating conditions. Accordingly, the non-significance of these variables is presented as a result conditioned by measurement, by the ex ante design, and by the structure of the sample, rather than as proof of theoretical irrelevance.

7. Missing data and complete-case analysis.- We appreciate this comment. In the revised version, we explicitly acknowledge that the use of complete-case analysis reduced the sample to 144 and 143 valid observations and that no formal assessment of the missingness mechanism or systematic comparison between retained and excluded cases was performed. Rather than presenting this decision as methodologically neutral, we now describe it as a relevant limitation, since bias cannot be ruled out if non-response was associated with contextual or attitudinal variables. Although we did not incorporate multiple imputation, we considered it important to make this issue fully transparent.

8. ORA–REDCIA comparison.- Thank you for this observation. We improved the comparison between ORA and REDCIA by more clearly highlighting their institutional differences and the distinct magnitude of SI’s effect across mechanisms. However, we also explicitly clarify that this comparison remains primarily descriptive and analytical, since no mechanism × predictor interactions or formal coefficient-difference tests were estimated. In this way, we avoid presenting as inferential what is in fact an interpretive comparison between two parallel specifications.

9. 162/170 and eight/nine jurisdictions discrepancy.- We very much appreciate this observation. In the revised version, we clarify that the manuscript uses 162 usable responses corresponding to the analytical subset effectively available, coming from eight Amazonian jurisdictions, rather than necessarily the full original survey universe. This clarification seeks to resolve the tension with the broader universe mentioned in the project’s earlier documentation. We acknowledge that this point required special precision, and we therefore strengthened the wording to make explicit that the article analyzes the subset usable for analytical purposes, rather than the full raw dataset originally collected.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

Comments to Authors (Round 2)

The revised manuscript demonstrates substantial improvement. The authors have addressed key concerns raised in the previous round, particularly by clarifying the exploratory nature of the KAP–UTAUT integration, moderating theoretical claims, and improving the transparency of the methodological approach.

The use of penalized ordinal regression and the justification regarding quasi-separation are appropriate and strengthen the analytical rigor of the study. The discussion has also been revised to better reflect the limitations of the data and measurement approach.

However, a few issues still require minor clarification:

While the ceiling effect is now acknowledged, its implications are not fully addressed. The authors should more explicitly discuss how this limitation affects the interpretation of the non-significance of Performance Expectancy and the apparent dominance of Social Influence.
The treatment of constructs as formative/compositional is acceptable, but the justification appears post hoc. The authors should clarify this design choice more explicitly as a limitation rather than a methodological position.
The interpretation of coefficients in the ordinal logistic model would benefit from clearer explanation, particularly regarding the direction of effects and their substantive meaning for readers unfamiliar with cumulative logit models.

Overall, the manuscript has improved significantly and is now close to being suitable for publication. I recommend acceptance after minor revision.

Author Response

Comments 1:

We thank the reviewer for this suggestion, which has helped improve the clarity of the manuscript.

In response to this comment, we have expanded the explanation of how the coefficients in the cumulative ordinal logistic model should be interpreted. In particular, we have clarified that negative coefficients do not indicate a negative effect on the dependent variable; rather, they reflect a reduction in the probability of belonging to lower categories, which in turn implies an increase in the probability of higher levels of intention to use.

In addition, we have strengthened the substantive interpretation of the results by including odds ratios and confidence intervals, thereby facilitating understanding of the effects of the variables for readers who are not specialists in ordinal logistic models.

These improvements have been incorporated into the results section of the manuscript.

Comments 2:

Thank you for this valuable observation. We agree that the previous wording could be read as an overly strong methodological justification of the formative/compositional treatment, rather than as a pragmatic limitation of the study design. In response, we revised the manuscript in four ways. First, in the “Study Aim and Contribution” section, we toned down the claim of methodological novelty and clarified that the KAP–UTAUT linkage is an exploratory operationalization rather than a validated measurement integration. Second, in the subsection “Operationalization of UTAUT constructs from a KAP survey”, we reframed the treatment of PE, SI, and FC as a measurement limitation stemming from the use of a pre-existing KAP instrument that was not originally designed as a UTAUT scale. Third, we removed the duplicated measurement-justification paragraph from the Results section, since this discussion belongs in Methods and Limitations rather than in the presentation of findings. Fourth, we added an explicit limitation in the final Discussion/Conclusions stating that the composites were derived from a secondary and pragmatic mapping of conceptually related KAP items and therefore should be interpreted as exploratory proxy composites rather than psychometrically validated latent constructs. We believe this revision addresses the concern and clarifies that the design choice is acknowledged as a study limitation, not as a general methodological position.

Comments 3:

We thank the reviewer for this suggestion, which has helped improve the clarity of the manuscript.

These improvements have been incorporated into the results section of the manuscript.

Reviewer 2 Report

Comments and Suggestions for Authors

Thanks for the revisions.

Author Response

The new version incorporates a substantial revision of the manuscript, focused on strengthening the methodological robustness, interpretative clarity, and argumentative coherence of the study: it clarifies that the model has a relational and exploratory scope rather than a predictive one; it explicitly acknowledges as a limitation that the KAP questionnaire was not originally designed as a UTAUT instrument and that, therefore, the variables should be understood as proxy analytical composites rather than validated latent constructs; it provides a more detailed account of the operationalisation of PE, SI, and FC, as well as of the implications of using complete-case analysis; it expands the discussion of the ceiling effect and the low variability of PE in order to qualify the non-significance of some predictors; it clarifies the interpretation of negative coefficients in cumulative ordinal logistic regression, additionally incorporating odds ratios and confidence intervals to facilitate the reading of the results; it offers a more rigorous justification for the verification of assumptions under a penalised ordinal regression approach instead of applying the Brant test inappropriately; and it strengthens the practical implications for ACTO/OTCA and the avenues for future research, such that the article now presents a more precise, transparent, and robust exposition, thereby clearly improving its scientific and editorial quality.