Share

Review as Control

The Review That Isn't There

By Sevda Polat— August 12, 2026

Feature image for article: The Review That Isn't There

A survey of 86 deployed agent systems found that 74% use human verification to ensure quality. None of them compared outcomes with the review step against outcomes without it. Separately, a meta-analysis of 106 human-AI experiments found that adding a human reviewer to a stronger AI performer tends to make the result worse. The step is staffed, budgeted, and cited whenever regulators ask how quality is maintained. Whether it improves anything remains, in most organizations, untested.

Review as Control

The Review That Isn't There

By Sevda Polat— August 12, 2026

A survey of 86 deployed agent systems found that 74% use human verification to ensure quality. None of them compared outcomes with the review step against outcomes without it. Separately, a meta-analysis of 106 human-AI experiments found that adding a human reviewer to a stronger AI performer tends to make the result worse. The step is staffed, budgeted, and cited whenever regulators ask how quality is maintained. Whether it improves anything remains, in most organizations, untested.

Designing Effective Review

Designing Review That Resists Its Own Collapse

By Sevda Polat— August 12, 2026

Feature image for article: Designing Review That Resists Its Own Collapse

An FDA-authorized pathology tool requires the doctor to record a diagnosis before the AI reveals its own. Commit first, then compare: that ordering contains most of what you need to know about designing review that resists rubber-stamping. Three areas of design detail decide whether a checkpoint holds: what evidence the reviewer sees, when review gets triggered, and how you find out whether any of it is still working.

Designing Effective Review

Designing Review That Resists Its Own Collapse

By Sevda Polat— August 12, 2026

An FDA-authorized pathology tool requires the doctor to record a diagnosis before the AI reveals its own. Commit first, then compare: that ordering contains most of what you need to know about designing review that resists rubber-stamping. Three areas of design detail decide whether a checkpoint holds: what evidence the reviewer sees, when review gets triggered, and how you find out whether any of it is still working.

Adjacent Thinking

Human-AI Teams Usually Underperform the Better Solo Player

A meta-analysis of 106 studies finds most human-AI pairings lose to whichever performer was already stronger alone.

When the Model Grades Its Own Homework

LLM suggestions shift annotator labels toward the model's own outputs, inflating the evaluated model's reported performance.

A Safety Engineering Vocabulary for Agent Oversight Timing

Alert Fatigue Degrades Review Quality Across the Board

Longer AI Explanations Raise Confidence Without Improving Accuracy

The Out-of-the-Loop Problem Hasn't Gone Away

Adjacent Thinking

Human-AI Teams Usually Underperform the Better Solo PlayerA meta-analysis of 106 studies finds most human-AI pairings lose to whichever performer was already stronger alone.

When the Model Grades Its Own HomeworkLLM suggestions shift annotator labels toward the model's own outputs, inflating the evaluated model's reported performance.

Quick links

A Safety Engineering Vocabulary for Agent Oversight Timing

Alert Fatigue Degrades Review Quality Across the Board

Longer AI Explanations Raise Confidence Without Improving Accuracy

The Out-of-the-Loop Problem Hasn't Gone Away

Review Failure Modes

Human review is the go-to safeguard when agents handle consequential work. But the review step has its own failure modes, well-documented in fields like clinical alerting and code review, and increasingly visible in agent workflows.

Research points to at least six. The reviewer might arrive too late, see only what the agent chose to surface, or lack real authority to intervene. Cognitive anchoring and sheer approval volume degrade attention from the inside. And without outcome feedback, reviewers never learn to calibrate. These failures interact, and they tend to accumulate without drawing attention to themselves.

Review Failure Modes

Human review is the go-to safeguard when agents handle consequential work. But the review step has its own failure modes, well-documented in fields like clinical alerting and code review, and increasingly visible in agent workflows.

Research points to at least six. The reviewer might arrive too late, see only what the agent chose to surface, or lack real authority to intervene. Cognitive anchoring and sheer approval volume degrade attention from the inside. And without outcome feedback, reviewers never learn to calibrate. These failures interact, and they tend to accumulate without drawing attention to themselves.

Late Arrival

The Reviewer Shows Up After the Window Closes

A Taobao field study found that when human agents took over after customer frustration had already built, chat duration rose 40.8% and satisfaction dropped nearly a full point compared to matched human-handled cases. By that point, the viable options had already narrowed.

Framing Bias

Reviewers See the Agent's Summary, Not the Situation

A Nature Machine Intelligence study found people evaluating LLM explanations performed barely above chance at distinguishing correct from incorrect answers. Longer explanations raised confidence without improving discrimination. The explanation shaped the reviewer's judgment rather than informing it.

Authority Gap

Review Without Authority to Act Is Just Observation

When a reviewer can see a decision but cannot reject, pause, or reroute it, the oversight is ceremonial. A common pattern: approval requests that time out into automatic execution. A position paper warns this generates false legitimacy without actual control.

Confidence Anchoring

The Agent's Confidence Anchors the Reviewer's Own Judgment

An AAAI experiment found participants accepted high-confidence AI advice more often, even when it was wrong. Well-calibrated confidence improved accuracy by 20%, but miscalibrated confidence produced gains indistinguishable from zero. Anchoring sets in before the reviewer forms an independent view.

Approval Fatigue

Too Many Items and the Reviewer Stops Noticing

Computerized prescribing alert systems show override rates from 46% to 96%. Code review experiments find defect detection dropping from 59% to 35% as patch complexity grows. At high volume, review becomes pattern-matching against a default of "approve."

Missing Feedback

Reviewers Never Learn Whether Their Approvals Were Right

An ACL 2025 study found model suggestions shifted annotator labels toward the AI's output while raising self-reported confidence. When those labels became evaluation data, reported model performance inflated. Without outcome feedback, "human-approved" describes a process step, not a quality signal.

Review Design Patterns

Human review is the default answer to "how do we keep agents from doing something wrong?" But review quality depends heavily on how it's structured. A reviewer who sees the agent's recommendation before examining the evidence is already anchored. One who shares the agent's full context may miss systematic drift. One who arrives after the meaningful choices have closed adds a step without adding a check.

Six design patterns target specific failure modes in review workflows. Some protect reviewer independence, others restructure what the reviewer sees and when, and a few govern how long an agent's authority persists. Each costs something, and each degrades differently under volume.

Review Design Patterns

Human review is the default answer to "how do we keep agents from doing something wrong?" But review quality depends heavily on how it's structured. A reviewer who sees the agent's recommendation before examining the evidence is already anchored. One who shares the agent's full context may miss systematic drift. One who arrives after the meaningful choices have closed adds a step without adding a check.

Six design patterns target specific failure modes in review workflows. Some protect reviewer independence, others restructure what the reviewer sees and when, and a few govern how long an agent's authority persists. Each costs something, and each degrades differently under volume.

Blind Review

Blind Review Catches Drift First Reviewers Will Miss

A second reviewer evaluates the work without seeing the agent's output or the first reviewer's assessment. The goal is detecting systematic drift, not duplicating effort. Production workflows typically sample 10-15% of items, selecting cases likely to expose correlated errors rather than random variation.

Seeded Disagreements

Seeded Disagreements Show Whether Reviewers Catch Real Errors

Known-answer cases get quietly inserted into a reviewer's queue. Performance on these planted items reveals whether the reviewer is catching errors or waving things through. The ongoing cost is maintaining a gold standard that tracks the agent's current failure modes, not last quarter's.

Evidence First

Showing Evidence Before Recommendations Weakens the Anchoring Effect

When reviewers see the AI recommendation first, it shapes how they read the supporting evidence. Withholding the recommendation until the reviewer forms an initial judgment reduces this anchoring. Research found reviewers were 65% more likely to revise answers in a human-first workflow.

Bounded Authority

Expiring Permissions Force Agents to Re-Earn Their Authority

Agent permissions should expire and require renewal rather than persist by default. Sensitive operations may warrant token lifetimes of 60-300 seconds. The harder design question: defining validity conditions beyond clock time, like price changes or shipment status, and monitoring them during execution.