Impartiality in review: on double-blind peer review

The debate over peer-review models is one of scholarly publishing’s perennials. Advocates of open review argue that transparency will raise quality; in the single-blind model the reviewer knows the author but not vice versa; in the double-blind model, neither side sees the other’s identity.

On this question, researchers themselves are fairly unambiguous. In IOP Publishing’s 2024 state-of-peer-review study, 52% of respondents said they prefer double-anonymous (double-blind) review — well ahead of open review. Another finding from the same study hints at why: in 2024, 16% of researchers reported experiencing bias in the review process. That is down from 24% in 2020 — an improvement, but hardly a solved problem.

Bias is not always bad faith, either. Reading a famous name’s manuscript more generously, keeping an unknown institution’s work at arm’s length from the outset — these are usually tendencies people are not even aware of. The double-blind model exists precisely for this: it forces the text to be judged independently of its owner.

Double-blind in Türkiye: less a preference than the de facto standard

For academic journals in Türkiye, double-blind review is mostly not even a matter of preference. The TR Dizin evaluation criteria — TR Dizin being Türkiye’s national journal index, run by TÜBİTAK ULAKBİM — require at least two reviewers per article, ideally from different institutions, and expect reviewer reports to contain genuine scholarly assessment and process records to be retained. The great majority of journals declare a double-blind model in their editorial policy. So the question is not “should we be double-blind?” but “can we actually deliver the double-blindness we declare?”

That second question is far harder than the first.

Where anonymity leaks

On paper, double-blind review has a simple definition; in practice, anonymity leaks through dozens of small holes:

• The file itself. The name sits in the Word file’s “author” field; the institution’s name rides along in the PDF metadata. The title page has been removed, but the footnote saying “this study continues our project X” is still there.
• Email traffic. If the process runs on email, one “reply all” accident or a misdirected CC ends months of carefully guarded anonymity in a second.
• Indirect clues. The author’s name appearing in reviewer correspondence instead of the manuscript code; the revision file being named “smithjones_article_v3.docx”…

Note that none of these leaks involves bad faith. All of them are the natural consequence of entrusting anonymity to human attentiveness. And attentiveness is the scarcest resource of an editorial office dealing with forty submissions a day.

The system’s job: taking anonymity out of the attention economy

Our position on this is plain: double-blindness should not be a courtesy convention but a rule the system takes ownership of.

In Nasirus, the reviewer’s screen shows the work, not the author; on the author’s side, reviewers live as “Reviewer 1, Reviewer 2.” Because correspondence flows inside the system, the misdirected-CC class of accident simply does not exist. At no point is anonymity entrusted to anyone’s attentiveness — the system treats guarding it as its own job.

There is a side benefit as well: the process documentation TR Dizin asks for — who was assigned, when the report arrived, how the decision was reached — accumulates by itself while anonymity is preserved. Record-keeping and identity-hiding are two conflicting workloads in a manual process; in a system, they are two outputs of the same mechanism.

Being honest about double-blind’s limits

We advocate for the model, but its limits deserve stating too: double-blind anonymity is not absolute anonymity.

In narrow fields, specialists recognize each other’s lines of work; there will always be a reviewer who guesses the author from “this style, this dataset, this citation pattern.” Self-citations (“as we showed in our earlier study…”) are among the classic giveaways — which is why many journals ask that self-citations be neutralized in the review copy.

These limits do not devalue the model; they sharpen its aim. The goal of double-blind is not a guarantee of unguessability but a guarantee of unverifiability: the reviewer may guess but cannot know, and the decision has to be written about the text, not the name. The system’s job, moreover, is to leave no room for the leaks that make guessing easy — so that the reviewer is left with nothing more than speculation.

There is also the far end of the process: once review concludes and the work is accepted, when and how anonymity lifts must be defined as well. The production team working with the author naturally knows who they are; but reviewers stay outside author-production correspondence even after the process ends. Anonymity is not a phase — it is the discipline of a wall.

The model debate will go on; the discipline is not up for debate

Open-review experiments continue and are producing valuable results in some fields; which model is “right” is probably a question whose answer varies by discipline. But whichever model is chosen, one thing does not change: a journal must apply its declared model in fact, without exception, and in a documentable way. A journal whose policy page says “double-blind” while its reviewer-invitation emails carry the author’s name has a trust problem — regardless of where anyone stands in the model debate.

To see where anonymity might be leaking in your own process, we suggest a simple drill: open every file and every message from the review process of the last article you accepted, and read them through an author’s eyes, asking “would they have recognized me here?” If the result surprises you — it surprises most journals — write to us to talk about a workflow that protects anonymity by system.