Review Guidelines

Purpose. These guidelines assist reviewers in assessing the scientific quality, methodological soundness, originality, and reporting transparency of manuscripts submitted to ZAHRA. A review should help the editor make an evidence-based decision and provide authors with constructive, actionable feedback. Reviewers advise; the editor makes the final decision.

1. Review model and editorial responsibilities

ZAHRA describes its evaluation system as double-blind peer review. The existing editorial policy indicates that manuscripts passing preliminary screening are handled by a section editor and assigned to at least one reviewer. Reviewer assignments must be based on subject expertise and the ability to provide an independent and timely assessment. The journal states an indicative review period of up to three weeks; specific deadlines are communicated through the editorial workflow.

• Accept a review invitation only if the subject lies within your competence and you can provide a careful assessment within the requested period.
• Inform the editor promptly about actual, potential, or perceived conflicts of interest, including recent coauthorship, institutional ties, financial interests, or competitive relationships.
• Decline a review if impartiality or confidentiality cannot be maintained; avoid attempting to identify a blinded author.
• Treat manuscripts, data, unpublished results, review correspondence, and supplementary files as confidential. Do not share or independently reuse them.
• Do not upload confidential manuscripts to third-party services, including generative-AI services, without express permission from the journal and the relevant authors.
• Support comments with specific manuscript locations, methodological reasoning, and concrete improvements; maintain respectful language.

2. Review criteria

3. Special considerations for AI and data-science submissions

• Check whether dataset leakage, duplicate observations, sampling bias, class imbalance, or non-representative test sets could compromise validity.
• Require clearly defined training/validation/test partitions, tuning strategies, baselines, and evaluation metrics appropriate to the research objective.
• For proposed deep-learning architectures, identify the specific innovation and seek component ablation, robustness checks, uncertainty or statistical analysis when essential to the claim.
• Verify that models, frameworks, hyperparameters, computational environment, random seeds, and data/code availability are reported to a reasonable reproducibility standard.
• Distinguish demonstrated results from speculation, and verify that any societal, privacy, safety, or fairness claims are proportionate to the evidence.

4. Reviewer-report structure

1. Summary: In 3–5 sentences, state the research question, claimed contribution, approach, principal evidence, and your overall understanding.
2. Major comments: Prioritize problems that could affect novelty, methodological validity, interpretation, ethics, or the reliability of conclusions. Point to exact sections, figures, or tables.
3. Minor comments: Identify fixable issues in wording, figures, table titles, citations, formatting, notation, or disclosure.
4. Questions to authors: Ask for missing experimental details, clarification, additional evidence, or explanations of unexpected results.
5. Confidential comments to the editor: Separately disclose concerns that should not be part of the public author-facing report, including conflicts of interest or suspected integrity problems.
6. Recommendation: Select the appropriate category in OJS and explain the reasoning; do not treat the recommendation as the editorial decision.

5. Suggested evaluation worksheet

For internal consistency, reviewers may use the following descriptive scale. It is not an automatic acceptance threshold and does not replace editorial judgment.

6. Editorial recommendations and follow-up

Recommendations may include acceptance, revision, resubmission for further review, submission elsewhere, or rejection, as available in the journal’s workflow. A revision must respond to each major concern and identify changes in the manuscript. Reviewers should evaluate the revision against their original concerns and any new evidence, not introduce unrelated demands without justification.