Question and contribution
Is the question clear, relevant and supported by an accurate account of prior work? Is the claimed contribution demonstrated?
Apply the confidentiality and human-assessment requirements, and handle integrity concerns with evidence.
AI and Decision Sciences · Reviewer guidance
A review should help the editor judge the research and help the author understand the evidence, weaknesses and possible remedies.
Adopted guidance, version 1.1 · 11 November 2026. This shared page is a preparation aid. The six journal policy libraries and scientific guides provide the complete adopted requirements for each discipline. Policy adoption does not appoint editors or establish completed review. See the prospective author agreement and records operating rules.
Before accepting a review
Accept only when the subject or methodological scope fits your expertise and you can meet the agreed timing. Tell the handling editor which parts fall outside your expertise.
Declare relevant collaborations, institutional relationships, financial interests, prior paid assistance and personal or professional conflicts. Anyone who helped develop the submitted work must not review or decide that work.
AI and Decision Sciences uses double-anonymous review and seeks at least two independent external reports for research articles. Author and reviewer identities are not routinely shared during review. Tell the editor if an identity becomes apparent; do not try to identify authors or contact them. This guide does not create a reviewer appointment or invitation.
What to examine
Is the question clear, relevant and supported by an accurate account of prior work? Is the claimed contribution demonstrated?
Do the methods fit the question? Are comparisons fair, metrics defined, assumptions justified and uncertainty examined?
Inspect the actual availability, rights, documentation and version references. State precisely whether anything was accessed or rerun and under which conditions.
Do the conclusions follow from the evidence? Are limitations, unsuccessful results and alternative explanations addressed?
Check the declared approvals, consent, contributions, conflicts, assistance and AI use. Bring integrity concerns to the editor privately and with evidence.
Separate necessary corrections from optional suggestions. Explain each major concern and what information could resolve it. Avoid coercive citations.
Three distinct questions
Preparation completeness, computational execution and scientific assessment answer different questions.
Can the relevant files and documentation be accessed, subject to legitimate restrictions?
Was a stated calculation rerun for a particular code, data and environment version, and what happened?
Are the design, evidence, interpretation and contribution sound within the reviewer’s stated expertise?
A lab approval, author checklist or successful software run does not substitute for the journal’s independent peer review.
Confidentiality and AI
Do not upload confidential manuscripts, reports, participant information or unpublished materials to public AI tools or other third-party services without the permissions and safeguards required by the journal.
The AI and Decision Sciences policy does not permit generative AI to create the reviewer’s scientific assessment. Any permitted routine automated assistance must be disclosed and checked by the responsible person.
Do not delegate a review or contact authors directly without the editor’s agreement. Review reports, reviewer identities and author responses will not be published automatically. Any future transparent-review option requires a clear policy and the necessary consent.
Read the AI and confidentiality frameworkA useful reviewer report
Recognition should describe the review actually completed. No review certificate or publication decision is issued by reading this guide.
Your next step
Authors can use the evidence workspace to make the research design, material availability and disclosures easier to follow.