How to give good critique

Open research only works if the critique is as open as the publishing. A good review makes the work — and the field — better. A bad one just makes noise. This is how we try to review here.

Start from the strongest version

Before you object, restate the author's claim in its strongest form. If they'd agree with your summary, you've earned the right to push on it. Most weak reviews attack a claim the author never made.

Separate three things

When you respond to a piece of research, keep these apart:

• The claim — is it clearly stated and falsifiable?
• The evidence — does it actually support the claim?
• The presentation — is it clear, honest about limitations, reproducible?

Say which one you're critiquing. "I don't believe the result" and "I don't understand the method" need very different fixes.

Make critique actionable

Every objection should point at a next step:

• Vague: "This doesn't seem rigorous."
• Useful: "The baseline is missing — without comparing to X, we can't tell if the gain is real. Can you add it?"

If you can name the experiment, the citation, or the control that would change your mind, name it.

Be generous with intent, strict with claims

Assume the author is acting in good faith and is smart. Then hold the work to a high bar. Warmth toward the person and rigor toward the claim are not in tension — they're the whole point.

What good critique looks like

• Ask before asserting. "Did you control for X?" beats "You ignored X."
• Distinguish fatal from fixable. Flag which concerns block publication and which are nice-to-haves.
• Credit what's right. Naming the strong parts isn't flattery; it tells the author what to keep.
• Reproduce when you can. "I re-ran it and got Y" is the most valuable comment on this whole site.

Reviewing here

Comments on any published piece are open to members. If you're reviewing a submission as a maintainer, remember: we publish sound work we disagree with. Review for rigor and clarity, not for whether you'd have reached the same conclusion.

Disagreement, done well, is a feature of an open lab — not a bug.