Structured dissent

Peer review dies socially before it dies procedurally. In a team that has quietly normalized adopting the system's output, asking a colleague "did you check that?" starts to feel like an accusation — and the inhibiting peer pathway closes exactly when the reinforcing one is running hottest. The group-decision literature has documented this failure family for decades: cohesive groups converge on the confident member's answer, and dissent that depends on individual courage arrives too rarely to matter.

Structured dissent institutionalizes the challenge instead: a rotating second-look duty with protected time, a named devil's-advocate role on consequential calls, and blameless flagging channels so the question is the role's job rather than a personal judgment. The same design applies to human-and-agent teams — a worker assigned to challenge the agent pool's outputs on a schedule is a peer check that autonomy expansion cannot quietly erode.

What it buys on the diagram: the peer-check pathway strengthens, and peer contagion is capped — a claim no longer travels on a colleague's confidence alone. What it costs is honest: protected hours, and the discomfort of institutionalized disagreement.

The procedural half is a record: require the reason for agreeing or disagreeing with the system's suggestion to be written down at the moment the call is made. That turns the second look into an auditable trace, so supervision can examine how the person and the tool worked together instead of counting how often they agreed. The failure it is written against is documented — staffing reduced on the assumption that the tool is efficient turns review into rubber-stamping, and an organization that penalises deviation makes its own oversight performative. Neither cited chapter offers an effect size, so this page states a design prescription and no measured gain.

The failure mode to design against is ritual dissent — a devil's-advocate who performs the role while everyone waits for the meeting to end. Rotation, real cases, and tracking how often the second look actually changes an outcome keep the role live; a challenger who never changes anything is either auditing a perfect team or has stopped challenging.