Peer Review¶

Peer evaluation (review) — instantiates Competence Calibration Feedback

Brings the external judgment of domain peers to bear on someone's work, supplying an outside perspective the person cannot see from the inside — while guarding against slippage into status judgment.

Peer Review supplies the one thing self-assessment structurally cannot: an outside view from someone who actually knows the domain. Its defining move is who does the evaluating — not a rubric and not a boss, but competent peers, which yields the tacit, insider judgment that catches "smells" a checklist would miss. That same feature is its signature hazard: peer judgment carries status and identity charge, so the mechanism must actively keep the review about the work rather than about rank or likability. It generates evidence the actor is blind to, explains it in terms they can act on, and watches that the evaluation doesn't curdle into a popularity contest.

Example¶

A developer is sure their pull request is clean and ready to merge. Peer Review routes it to two colleagues who know the codebase. They surface exactly what the author couldn't see from inside their own head: an edge case where the new code silently drops errors, and a naming choice that will confuse the next maintainer. Neither issue shows up against a style linter — they're the kind of tacit "this will bite someone in six months" judgment only a domain peer makes. The feedback is specific and about the code, not the coder. And because reviews here run mutually and anchor to shared conventions, the exchange stays a technical one; the reviewers aren't ranking the author, they're improving the artifact. The developer's confidence updates on evidence they had no way to generate alone.

How it works¶

Its distinguishing feature is multiple domain-competent perspectives cross-checking a piece of work, which surfaces blind spots the actor — and even a rubric — can't reach, and averages out any single reviewer's idiosyncrasy. But peer judgment is only calibration, not status ranking, when two guards are in place: an anchor (what are we reviewing against?) that keeps it about the work, and a safety frame plus threat-monitoring that keeps it from sliding into who's up and who's down. The mechanism's discipline is aiming every comment at the artifact and reading the exchange for signs it's becoming about the person.

Tuning parameters¶

Reviewer count and independence — one peer versus several reviewing independently. More reviewers reduce idiosyncrasy and collusion but cost time.
Anonymity — open versus blind review. Blind curbs status and halo effects; open enables back-and-forth dialogue.
Benchmark anchoring — free-form opinion versus review against shared criteria. Anchoring keeps the judgment about the work rather than personal taste.
Scope discipline — reviewing the work versus the person. Narrowing strictly to the artifact is the main thing that prevents status judgment.
Reciprocity — one-directional versus mutual review. Mutual review normalizes the practice and lowers the threat of being singled out.

When it helps, and when it misleads¶

Its strength is supplying the outside view self-assessment can't generate: peers who know the domain catch blind spots the actor is structurally unable to perceive in their own work.

It misleads when the guards fail. Without a benchmark or safety frame, peer review degrades into a popularity or status judgment, where a reviewer's overall impression of the author colors the assessment of the work — the halo effect at play.^[1] Peers can collude into mutual leniency, or pile onto an easy target, and the whole apparatus can be weaponized to justify a ranking already decided. The discipline is to anchor every review to shared criteria, keep each comment about the artifact, and monitor for the identity threat and status dynamics that turn evaluation into a verdict on the person.

How it implements the components¶

Peer Review fills the external-source-and-safeguard slice:

performance_evidence_set — peers' structured observations become domain-relevant evidence about the work that the actor could not produce alone.
calibration_feedback — peers explain what they see in behavior- and artifact-specific terms the actor can act on.
identity_threat_monitor — watches that the review stays about the work and does not slide into status ranking or personal judgment.

It does not supply the fixed standard being reviewed against (performance_benchmark, Competency Framework) or set the autonomy consequences of the result (decision_rights_boundary, Decision Rights by Competence).

Instantiates: Competence Calibration Feedback — supplies the outside-view evidence and feedback that self-assessment structurally lacks.
Sibling mechanisms: Benchmarked Feedback · Calibration Conversation · Competency Framework · Exemplar Comparison · Confidence Rating Scale · Calibration Exercise · Skills Assessment · Reflective Error Log · Simulation or Case Test · Decision Rights by Competence · Supervised Practice

Notes¶

Peer review is only a calibration mechanism when it has both an anchor and a safety frame; strip either away and it reliably decays into status judgment. It supplies evidence and feedback but borrows its standard from a Competency Framework or shared conventions — without that anchor, "peer opinion" and "peer calibration" are not the same thing.

References¶

[1] The halo effect is the tendency for an overall impression of a person to color judgments of their specific qualities. In peer review it is the mechanism by which a reviewer's regard for the author — high or low — contaminates their assessment of the actual work, which is why anchoring and blind review matter. ↩