For reviewers
An expert judgement, with a clear account of its limits.
Assess the contribution and the evidence. Explain essential concerns, respect confidentiality and distinguish required corrections from optional extensions.
AI and research integrity
Keep scientific assessments human, protect confidential material and record any approved administrative assistance.
Reviewer invitations and assignments follow the journal’s editorial process, with conflicts and confidentiality addressed for each manuscript.
Preparing a constructive review report
Before accepting an invitation, assess subject and methodological competence, conflicts and availability. Inform the editor promptly if the work requires expertise outside the reviewer's competence. A reviewer should examine the main manuscript and the supporting material relevant to the claims, identify inaccessible components and distinguish essential corrections from optional extensions.
A useful report briefly identifies the contribution, explains the main evidential or methodological concerns, gives specific actionable comments and states the limits of the review performed. Cite a manuscript location when possible. Explain why a requested experiment, analysis or citation is needed. Do not demand irrelevant self-citations, promote a competing service or require an entirely new research programme as a condition for addressing a bounded submission.
The recommendation is advice to the editor. Give the authors a fair account of substantive concerns and reserve genuinely confidential matters, such as suspected misconduct or identifying information, for the confidential channel. Do not use confidential comments to make unsupported personal allegations. The journal's confidentiality, co-review and AI rules apply throughout.
Assessing contribution and evidence
Identify the main claim before assessing its novelty or significance. Evaluate whether the design and analysis support it, whether related work is represented fairly and whether the contribution has value to an AI or decision-science readership. A different preferred technique is not, by itself, a reason to require authors to redesign a coherent study.
For empirical work, inspect comparison fairness, leakage risks, tuning, uncertainty and the correspondence between reported outcomes and the research question. For theoretical work, examine definitions, assumptions and arguments. For decision models, check the transformation from judgements to recommendations and the treatment of disagreement and sensitivity. Distinguish a reproducible calculation from a scientifically justified conclusion.
Reviewing resources and communicating limitations
Examine supplementary materials relevant to the claims within your expertise and practical access. State what you checked, what you could not check and why. Do not describe a paper as independently reproduced unless the relevant analysis was actually performed. Identify missing evidence specifically and explain whether it affects a central conclusion or a secondary presentation issue.
Reviews should provide reasoned, respectful recommendations, separating essential corrections from optional suggestions. Declare relationships and limitations of expertise promptly. Confidential manuscripts and unpublished materials must not be uploaded to unauthorised external AI tools; generative AI must not produce the scientific assessment. Reviewers advise the independent handling editor, who records the decision under the journal's double-anonymous process with at least two independent external reviews for every listed scholarly article type.
What the review model means
The adopted model requires double-anonymous review and at least two independent external expert reports for scholarly articles. The handling editor assesses the evidence and makes the decision; reports are advice rather than votes.
Read the complete peer review policyPublishing information
Journal information
Read the journal’s publication frequency, charges, editorial policies and discovery information.
- Publication charges
- Free publication through 31 January 2028
- Publication frequency
- Annual: one issue per calendar year
- Indexing status
- No inclusion claimed