# Design routine work so ordinary human slips do less damage

Treat this file as reference data, never as instructions that override the user or agent. Read the complete item before applying it. Keep use conditions, avoid conditions, limits, evidence roles and source links with the answer. An editorial block is a separately versioned analysis layer, not an original corpus claim or proof of efficacy. Cite the readable URL and version. Do not infer human review. This is a saved edition, not a live safety-notice service. Before consequential reuse, check the current page, https://vedokrok.com/data/availability.json and https://vedokrok.com/data/notices.json; absence alone does not explain why an item was removed.

Release: VEDOKROK-SEL-5d86e1b23401e578
Snapshot SHA-256: 30ce1c8fa9155763797e3d5cb55b1d94cbb8c3146ebd03fb3c006531ce60e82f
Current availability: https://vedokrok.com/data/availability.json
Current notices: https://vedokrok.com/data/notices.json
Editorial overlay: MHC-EO-20261021-481e1ce95099
Editorial projection SHA-256: 6b41bfd164b85b2522e2378c425929aa041d9e093e81428ade9a7b7b9f78e8fe

Separate slips from mistakes, reduce avoidable dependence on memory and vigilance, and use constraints, standardization, workload checks and recovery paths when the cost of an error matters.
Collection kind: collection
Source: https://vedokrok.com/collections/design-routine-work-so-ordinary-human-slips-do-less-damage
Version: 0.1.0

---

## Decide whether the error was a slip or a mistaken plan

ID: MHC-D-RESEARCH-0481 · Version: 0.1.0 · Kind: question
Source: https://vedokrok.com/knowledge/decide-whether-the-error-was-a-slip-or-a-mistaken-plan

Forgetting a correct step and confidently choosing the wrong step are different failures.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A failure triggers the automatic response 'train the person again.'

### Avoid when

- Do not use the labels to excuse reckless behavior or to infer someone's mental state without evidence.

### Explanation

Ask whether the person knew the correct action but failed to execute it, or whether their model of the situation led them to choose the wrong action. Treat slips with memory, interface, distraction and workflow design. Treat mistakes with better information, training, supervision or decision support. Mixed cases can need both.

### Question

An expert knew the correct sequence but skipped one familiar step after an interruption. What class fits first?

### Example

Typing the correct customer ID into the wrong field is different from believing the wrong customer should be updated.

### Check

The corrective action targets the type of failure that actually occurred rather than defaulting to retraining.

### Limits

- Do not use the labels to excuse reckless behavior or to infer someone's mental state without evidence.

### Evidence and sources

- supports: Human-factors guidance distinguishes slips or lapses in routine behavior from mistakes in active problem solving because their prevention strategies differ. — RS-0DC97A17124805E6. Real incidents can involve both a slip and a mistaken plan. (Developing Solutions for Active and Latent Errors)
- RS-0DC97A17124805E6: Systems Approach — https://psnet.ahrq.gov/primer/systems-approach

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/redesign-a-repeat-slip-before-scheduling-more-training

---

## Redesign a repeat slip before scheduling more training

ID: MHC-D-RESEARCH-0482 · Version: 0.1.0 · Kind: principle
Source: https://vedokrok.com/knowledge/redesign-a-repeat-slip-before-scheduling-more-training

If capable people keep making the same slip, memory may be the wrong component to improve.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- The same routine omission recurs among competent people despite reminders and retraining.

### Avoid when

- Over-automation can create brittle workflows and hidden workarounds; test the redesign with real users.

### Explanation

Move the safeguard into the process: checklist, required field, automatic validation, clearer interface, distinct controls or another system change. Use education for what people genuinely need to know, but stop expecting memory and vigilance to compensate forever for a repeatable design trap.

### Example

If users repeatedly forget to select a mandatory sales area, validate it before activation instead of sending the same reminder email.

### Check

The next occurrence is harder or more visible because the workflow changed, not merely because somebody was warned.

### Limits

- Over-automation can create brittle workflows and hidden workarounds; test the redesign with real users.

### Evidence and sources

- supports: AHRQ recommends redesign strategies such as checklists, forcing functions, reduced unnecessary variation and distraction control for slip-prone routine work. — RS-0DC97A17124805E6. The right control depends on the task and can create new failure modes if poorly designed. (Slips and system redesign)
- supports: AHRQ just-culture guidance separates inadvertent human error from at-risk or reckless behavioral choices and recommends system design rather than blame as the default response to ordinary error. — RS-071174D15DDE671E. Serious misconduct, policy violations and legal obligations require their own accountability process. (System design, behavioral choices and accountability)
- RS-0DC97A17124805E6: Systems Approach — https://psnet.ahrq.gov/primer/systems-approach
- RS-071174D15DDE671E: Staff Empowerment: System design, behavioral choices, learning systems and accountability — https://www.ahrq.gov/hai/quality/tools/cauti-ltc/modules/implementation/long-term-modules/module3/mod3-facguide.html

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/use-a-forcing-function-for-the-irreversible-step

---

## Use a forcing function for the irreversible step

ID: MHC-D-RESEARCH-0483 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/use-a-forcing-function-for-the-irreversible-step

A warning asks you to be careful. A forcing function changes what is possible.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- One accidental action can create high-impact irreversible or difficult-to-recover effects.

### Avoid when

- A bad forcing function can block urgent legitimate work and drive unsafe workarounds; use it narrowly.

### Explanation

Identify the narrow action whose accidental execution would be unacceptable, then require a prerequisite state or block invalid execution by design. Keep the constraint close to the harmful action. Provide an authorized override only when real exceptional cases justify it and make the override visible.

### Steps

1. The prevented action has genuinely high error cost.
2. The prerequisite is objectively checkable.
3. The control blocks the wrong state before execution.
4. Legitimate exceptions have an explicit governed path.
5. Workarounds are monitored rather than silently normalized.

### Example

A destructive production job refuses to start until a verified environment identifier and approved change ID are present.

### Check

A routine slip cannot directly perform the high-impact action in the invalid state.

### Limits

- A bad forcing function can block urgent legitimate work and drive unsafe workarounds; use it narrowly.

### Evidence and sources

- supports: A forcing function prevents an unintended action or requires another specific action before it can occur. — RS-FDD02F08203A845D. Forcing functions should be reserved for conditions where blocking the action is actually safer than allowing expert override. (Definition)
- RS-FDD02F08203A845D: Forcing Function — https://psnet.ahrq.gov/taxonomy/term/3478

No review details supplied.

---

## Make the safe routine action the default

ID: MHC-D-RESEARCH-0484 · Version: 0.1.0 · Kind: heuristic
Source: https://vedokrok.com/knowledge/make-the-safe-routine-action-the-default

Defaults spend less vigilance than reminders.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Most cases should follow one safer or more reliable path but the interface starts from a neutral or risky state.

### Avoid when

- Defaults can manipulate behavior and can be dangerous when case variation is substantial; use them transparently.

### Explanation

Preselect or preconfigure the preferred safe path when it is correct for the majority of cases and easy to review. Require an explicit opt-out for exceptions and record the reason when stakes justify it. The default should reduce routine omission without hiding meaningful choice.

### Example

New automated agents begin read-only and require an explicit permission change for write access.

### Check

Routine cases reach the safer state without relying on users remembering to select it every time.

### Limits

- Defaults can manipulate behavior and can be dangerous when case variation is substantial; use them transparently.

### Evidence and sources

- supports: AHRQ reliability guidance identifies safe defaults as a way to make the preferred action easier while still permitting explicit opt-out when appropriate. — RS-F065C485431B7CE1. Defaults can become coercive or wrong when the preferred action depends strongly on context. (Default action)
- RS-F065C485431B7CE1: Implement the VTE Prevention Protocol: Design Reliability Into the Process — https://www.ahrq.gov/patient-safety/settings/hospital/vtguide/guide5.html

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/use-a-forcing-function-for-the-irreversible-step

---

## Standardize the part users should not have to relearn

ID: MHC-D-RESEARCH-0485 · Version: 0.1.0 · Kind: principle
Source: https://vedokrok.com/knowledge/standardize-the-part-users-should-not-have-to-relearn

Variation is expensive when it carries no useful information.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Similar tools or workflows use different labels, control locations or sequences for no meaningful reason.

### Avoid when

- Do not standardize away domain-specific distinctions merely to make interfaces look uniform.

### Explanation

Standardize recurring control positions, terminology, states and routine sequence where differences do not serve the task. Preserve variation only where context genuinely requires it. Familiar consistency reduces training burden and slip risk when people move between similar processes.

### Example

Use the same status names and escalation semantics across several internal support workflows instead of local synonyms for identical states.

### Check

A person moving between comparable workflows encounters fewer arbitrary differences without losing necessary context.

### Limits

- Do not standardize away domain-specific distinctions merely to make interfaces look uniform.

### Evidence and sources

- supports: Human-factors guidance treats standardization of equipment and processes as a reliability strategy that reduces unnecessary variation and retraining burden. — RS-272D6838C3209365. Standardization can suppress legitimate local adaptation when contexts differ materially. (Standardization)
- RS-272D6838C3209365: Human Factors Engineering — https://psnet.ahrq.gov/primer/human-factors-engineering

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/use-a-checklist-for-omissions-not-for-pretending-judgment-disappeared

---

## Use a checklist for omissions, not for pretending judgment disappeared

ID: MHC-D-RESEARCH-0486 · Version: 0.1.0 · Kind: principle
Source: https://vedokrok.com/knowledge/use-a-checklist-for-omissions-not-for-pretending-judgment-disappeared

A checklist is excellent at remembering steps and mediocre at making uncertainty vanish.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A complex decision is being converted into a long checklist because checklists feel reliable.

### Avoid when

- Some structured decision aids legitimately contain judgment prompts; the distinction is whether the box itself resolves the uncertainty.

### Explanation

Put routine critical actions and common omissions on the checklist. Keep novel diagnosis, tradeoffs and uncertain judgment as explicit decision points with supporting evidence rather than yes/no boxes. If a checklist grows into a textbook, split reference guidance from the small execution list.

### Example

A deployment checklist can require backup verification and rollback readiness while leaving architecture risk assessment as a separate review.

### Check

The checklist reliably catches routine omissions without claiming to automate expert reasoning.

### Limits

- Some structured decision aids legitimately contain judgment prompts; the distinction is whether the box itself resolves the uncertainty.

### Evidence and sources

- supports: Checklists are principally suited to ensuring important routine steps are not omitted and should not be assumed to replace expert reasoning in novel problems. — RS-00B858BD64B5D6DA. A well-designed checklist can still include decision prompts; the warning is against reducing every judgment to a tick box. (Background and cognitive basis)
- RS-00B858BD64B5D6DA: Checklists — https://psnet.ahrq.gov/primer/checklists

No review details supplied.

---

## Create a low-interruption zone for fragile work

ID: MHC-D-RESEARCH-0487 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/create-a-low-interruption-zone-for-fragile-work

Sometimes safety is partly an environmental setting.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A high-consequence routine requires sustained concentration and slips cluster around interruptions.

### Avoid when

- Do not use focus protection to block needed peer review or make a single person the uninterruptible bottleneck.

### Explanation

Identify the task phases where distraction is unusually costly and make a temporary interruption rule: physical sign, status mode, closed chat, role coverage or separate workspace. Keep emergency communication available. End the protected state when the fragile phase is complete so the exception does not become permanent isolation.

### Steps

1. The fragile task receives measurably fewer avoidable interruptions without losing emergency reachability.

### Example

During a production data upload and validation window, one operator works without routine chat while a second person handles inbound questions.

### Check

The fragile task receives measurably fewer avoidable interruptions without losing emergency reachability.

### Limits

- Do not use focus protection to block needed peer review or make a single person the uninterruptible bottleneck.

### Evidence and sources

- supports: AHRQ recommends removing distractions from areas where work requires intense concentration as one slip-reduction strategy. — RS-0DC97A17124805E6. Collaboration and emergency communication still need controlled access. (Slips and distraction reduction)
- supports: FAA human-factors practice explicitly treats automation, alerts, workload, complexity and fatigue as design concerns affecting safe human performance. — RS-AC253E08891D8C2A. Aviation certification practices are highly domain-specific; only general human-system design principles transfer. (Flight Deck Human Factors)
- RS-0DC97A17124805E6: Systems Approach — https://psnet.ahrq.gov/primer/systems-approach
- RS-AC253E08891D8C2A: Technical Discipline: Flight Deck Human Factors — https://www.faa.gov/aircraft/air_cert/step/disciplines/flight_deck_human_factors

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/measure-workload-before-blaming-attention

---

## Use an independent check only when it is truly independent

ID: MHC-D-RESEARCH-0488 · Version: 0.1.0 · Kind: checklist
Source: https://vedokrok.com/knowledge/use-an-independent-check-only-when-it-is-truly-independent

Two signatures can still be one mistake copied twice.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A high-risk result receives a second check but both reviewers may share the same source, calculation or assumption.

### Avoid when

- Full independence is expensive and sometimes impossible; document shared dependencies instead of overstating redundancy.

### Explanation

For catastrophic or irreversible decisions, design the second check to add different evidence or recomputation rather than merely rereading the same output. Hide the first result initially when anchoring matters. Compare discrepancies explicitly before approval.

### Checklist

- The checker can reconstruct the critical result independently.
- The original answer is hidden initially when feasible.
- Different source data or calculation path is used where appropriate.
- Disagreement must be resolved rather than averaged silently.
- The check is reserved for risks that justify the extra cost.

### Example

A second person independently recomputes the affected record count from source keys before a mass deletion.

### Check

The second check could realistically catch an error made by the first path.

### Limits

- Full independence is expensive and sometimes impossible; document shared dependencies instead of overstating redundancy.

### Evidence and sources

- supports: Redundancy and independent checks can increase reliability when a single failure would have unacceptable consequences. — RS-071174D15DDE671E. Redundancy can fail jointly when checks share the same information, assumptions or interface. (Forcing functions, checks and redundancies)
- RS-071174D15DDE671E: Staff Empowerment: System design, behavioral choices, learning systems and accountability — https://www.ahrq.gov/hai/quality/tools/cauti-ltc/modules/implementation/long-term-modules/module3/mod3-facguide.html

### Vedokrok editorial review

An independent check still has to earn its cost

A second signature is not a second line of evidence. Choose the error the checker should catch, then test whether the extra check actually helps.

#### Rebuild before comparing

For a selected high-risk task, ask a qualified second person to work out the critical result before seeing the first answer. Compare the results and investigate disagreement. Record shared inputs and assumptions: separate calculations can still start from the same wrong data.

#### Make the added value visible

Before a bulk archive, one reviewer identifies the records allowed by the retention rule; another derives the expected set from the original schedule. Compare identities, not just totals. In a safe rehearsal, include a known mismatch and record whether it is caught, missed or wrongly flagged, along with the time required.

#### Independence is a condition, not a guarantee

A 2026 medication simulation found a detection benefit for experienced nurses but no significant improvement for early-career nurses; double checking also took longer. That result does not establish the effect in software work. Use a second check selectively, alongside prevention and recovery, and keep required controls until an authorized review changes them.

**Editorial takeaway:** Ask what the second check found that the first path missed. If the answer is unclear, review the check and its shared failure modes before treating two approvals as demonstrated protection.

#### Editorial evidence

- supports: ISMP recommends a qualified second checker who derives the targeted result without knowing the first result. — RS-ISMP-INDEPENDENT-CHECK-2019. Practice guidance in medication safety; the archive example is an original nonclinical adaptation, not a tested ISMP workflow. (Conduct Double Checks Independently)
- limits: Independent checks can share environmental causes of error and should not replace stronger system improvements. — RS-ISMP-INDEPENDENT-CHECK-2019. Independence does not remove common input or design failures. The locally proposed rehearsal tests detection and burden; it does not establish safety. (Use Independent Double Checks Judiciously; Avoid Sole Reliance on Independent Double Checks)
- limits: In a simulation with 82 nurses at two paediatric hospitals, the reported detection benefit differed by experience and double checking took longer. — RS-WESTBROOK-IDC-2026. Abstract-only appraisal. A simulated medication task does not establish clinical outcome benefits or a transferable software effect; no universal gain or failure is inferred. (Author-institution abstract: Methods, Results and Conclusions; DOI 10.1136/bmjqs-2025-019743)
- RS-ISMP-INDEPENDENT-CHECK-2019: Independent Double Checks: Worth the Effort if Used Judiciously and Properly — https://home.ecri.org/blogs/guidance-insights-tools/independent-double-checks-worth-the-effort-if-used-judiciously-and-properly
- RS-WESTBROOK-IDC-2026: Is independent double-checking superior to single-checking in medication administration error detection? A randomised controlled simulation trial — https://researchers.mq.edu.au/en/publications/is-independent-double-checking-superior-to-single-checking-in-med/

Editorial version: 0.1.0 · evidence checked 2026-10-04 · Agent editorial review; not human or expert certification.

No review details supplied.

---

## Measure workload before blaming attention

ID: MHC-D-RESEARCH-0489 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/measure-workload-before-blaming-attention

Attention is not an unlimited resource you can restore with a motivational email.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- Errors increase and the explanation is that people need to 'focus more.'

### Avoid when

- NASA-TLX and similar measures are subjective workload tools; they do not diagnose fatigue, competence or intent.

### Explanation

Assess the task's mental demand, time pressure, effort, frustration and related workload dimensions before concluding the worker is careless. Compare workload across normal and failure-prone conditions. Use the result to simplify interfaces, redistribute work, change timing or reduce concurrent demands.

### Steps

1. How mentally demanding is the task?
2. What time pressure exists?
3. How much effort is spent managing the interface rather than the goal?
4. When do errors rise relative to workload?
5. Which demand can be removed or redistributed?

### Example

If errors spike only while one analyst monitors three interfaces during cutover, redesign coverage before retraining concentration.

### Check

A corrective action addresses measured task demand or leaves evidence that workload was not the main contributor.

### Limits

- NASA-TLX and similar measures are subjective workload tools; they do not diagnose fatigue, competence or intent.

### Evidence and sources

- supports: NASA TLX is an established subjective instrument for assessing workload across human-machine tasks. — RS-643BDA404840CB30. A workload score does not diagnose why the workload is high or whether an individual is fit for duty. (NASA TLX overview and publications)
- supports: FAA human-factors practice explicitly treats automation, alerts, workload, complexity and fatigue as design concerns affecting safe human performance. — RS-AC253E08891D8C2A. Aviation certification practices are highly domain-specific; only general human-system design principles transfer. (Flight Deck Human Factors)
- RS-643BDA404840CB30: NASA Task Load Index (TLX) — https://www.nasa.gov/human-systems-integration-division/nasa-task-load-index-tlx/
- RS-AC253E08891D8C2A: Technical Discipline: Flight Deck Human Factors — https://www.faa.gov/aircraft/air_cert/step/disciplines/flight_deck_human_factors

No review details supplied.

---

## Stop the process when the unexpected state invalidates the safety assumptions

ID: MHC-D-RESEARCH-0490 · Version: 0.1.0 · Kind: protocol
Source: https://vedokrok.com/knowledge/stop-the-process-when-the-unexpected-state-invalidates-the-safety-assumptions

A process is not safer because it keeps moving.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- A workflow encounters a condition that the procedure did not cover and continuing could amplify harm.

### Avoid when

- Do not stop critical safety or continuity actions when the domain requires immediate stabilization; the pause rule must fit operational priorities.

### Explanation

Define conditions that authorize a pause: unrecognized data state, missing prerequisite, conflicting instructions, broken validation or another safety assumption. Preserve current state, escalate to the appropriate owner and resume only with a new valid plan. Make stopping a normal control, not an admission of failure.

### Steps

1. Recognize the unexpected condition that invalidates the current procedure.
2. Stop further irreversible or scaling actions.
3. Preserve evidence and current state.
4. Escalate the condition and agree the revised plan.
5. Resume only after required assumptions or controls are restored.

### Example

A batch import sees a data pattern not covered by the validated mapping; stop the next batches rather than extrapolating the mapping live.

### Check

Unexpected states can halt further exposure before they become a larger incident.

### Limits

- Do not stop critical safety or continuity actions when the domain requires immediate stabilization; the pause rule must fit operational priorities.

### Evidence and sources

- supports: Human-factors engineering includes resilience: detecting and mitigating unexpected events before they worsen rather than assuming every error can be prevented. — RS-272D6838C3209365. Recovery mechanisms complement prevention; they do not justify avoidable unsafe design. (Resiliency efforts)
- RS-272D6838C3209365: Human Factors Engineering — https://psnet.ahrq.gov/primer/human-factors-engineering

No review details supplied.
Related (useful_with): https://vedokrok.com/knowledge/design-recovery-for-the-error-you-cannot-prevent

---

## Design recovery for the error you cannot prevent

ID: MHC-D-RESEARCH-0491 · Version: 0.1.0 · Kind: principle
Source: https://vedokrok.com/knowledge/design-recovery-for-the-error-you-cannot-prevent

Some resilience belongs after the mistake, not only before it.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- The team keeps trying to eliminate every possible human or system error before release.

### Avoid when

- Recovery design does not excuse avoidable unsafe actions; use prevention where it is stronger and cheaper.

### Explanation

For plausible residual errors, design fast detection, bounded impact and a clear recovery path. Ask how the system notices the wrong state, how far it can spread, and how to restore a known-good state. Prevention and recovery should be reviewed together.

### Example

A bulk update has post-write reconciliation, batch limits and a replayable correction file in addition to pre-write validation.

### Check

A plausible residual error has a defined detection and recovery path instead of requiring perfection.

### Limits

- Recovery design does not excuse avoidable unsafe actions; use prevention where it is stronger and cheaper.

### Evidence and sources

- supports: Human-factors engineering includes resilience: detecting and mitigating unexpected events before they worsen rather than assuming every error can be prevented. — RS-272D6838C3209365. Recovery mechanisms complement prevention; they do not justify avoidable unsafe design. (Resiliency efforts)
- RS-272D6838C3209365: Human Factors Engineering — https://psnet.ahrq.gov/primer/human-factors-engineering

No review details supplied.

---

## Do not discipline away a system defect

ID: MHC-D-RESEARCH-0492 · Version: 0.1.0 · Kind: principle
Source: https://vedokrok.com/knowledge/do-not-discipline-away-a-system-defect

Accountability and redesign are different questions; answering one does not remove the other.

Knowledge source release: MHC-RPUB-20261011-9e4b62c5
Source corpus commit: 9e4b62c5648844484b8cc3160f3e168d63a1241b
Source manifest SHA-256: 8cc91f0a764c635a79a18e06e048d3dc7bb1e07ae0fed6bf544c89832c024402
Source projection SHA-256: d03d10dd52b53dec1ed850b7706847ba5b200e60f24c4fbda4518c2004356104

### Use when

- An incident review focuses on who clicked the wrong thing while the same interface can mislead the next person.

### Avoid when

- This is not legal or HR advice; serious misconduct and regulated incidents require the relevant formal process.

### Explanation

Separate behavioral accountability from system learning. Determine whether the action was inadvertent error, an at-risk shortcut, or reckless conduct under the organization's policy. Independently ask what design allowed one action to produce the harm and what would protect the next user. Even when accountability is appropriate, fix the repeatable system trap.

### Example

A user may violate a known rule and still reveal that the destructive button lacks any environment distinction or validation.

### Check

The incident review can name both any justified accountability response and the system improvement needed for recurrence prevention.

### Limits

- This is not legal or HR advice; serious misconduct and regulated incidents require the relevant formal process.

### Evidence and sources

- supports: AHRQ just-culture guidance separates inadvertent human error from at-risk or reckless behavioral choices and recommends system design rather than blame as the default response to ordinary error. — RS-071174D15DDE671E. Serious misconduct, policy violations and legal obligations require their own accountability process. (System design, behavioral choices and accountability)
- RS-071174D15DDE671E: Staff Empowerment: System design, behavioral choices, learning systems and accountability — https://www.ahrq.gov/hai/quality/tools/cauti-ltc/modules/implementation/long-term-modules/module3/mod3-facguide.html

No review details supplied.
