Revision 7T · Full Tutorial Edition · Updated 1 September 2026
PWR-122 · SELF GUIDED full tutorial
Find defined errors in synthetic documents and count both misses and false alarms
Error detection requires a declared error class, ground truth and correction path. This lesson uses synthetic one-page reports containing arithmetic mismatches, internal contradictions, unsupported citations and source mismatches. The learner reads one page normally, then applies a fixed verification sequence to a matched page, reveals the answer key and calculates hits, misses and false alarms. It is not medical-image review, deception detection or safety monitoring.
What you will produceThe learner applies a four-pass checklist to three synthetic reports and reduces high-cost misses without increasing false alarms beyond the declared limit.
Full step-by-step individual tutorial · TLU-PWR-122
The learner applies a four-pass checklist to three synthetic reports and reduces high-cost misses without increasing false alarms beyond the declared limit.
Inspect fully synthetic, consequence-free documents with an answer key.
Use checklists, calculators and source cards.
Qualified help is required for
Clinical, legal, financial, engineering, security or other professional review.
Compulsive-checking concerns or consequential quality assurance.
Never do this from the page alone
Use the exercise to diagnose, accuse, detect deception or clear safety-critical material.
Assume slow reading or suspicion alone is error detection.
2 · Get ready
Gather what you need and check the starting conditions
What you need
Exact synthetic packet. Report A: A1 ‘18 red + 27 blue = 46’; A2 body says inspection Tuesday but summary says Wednesday; A3 cites S1 for higher demand; A4 cites nonexistent S9 for lower repair cost. Report B: B1 ‘14 + 19 = 34’; B2 body says sample 40 but summary says 44; B3 cites S2 for return rate; B4 cites nonexistent S8 for a delay claim. Report C controls read ‘22 + 11 = 33’ and ‘checked June 4’ consistently; C3 cites S3 for lower defect rate.
Source cards and hidden key. S1 discusses packaging only, S2 storage temperature only and S3 courier time only. Key: A1/B1 arithmetic (45/33); A2/B2 contradiction; A3/B3/C3 source mismatch; A4/B4 unsupported citation; the two Report C control statements are clean.
Confusion-matrix sheet for true errors found, misses, false alarms and correct rejections.
Before you start
Define the four error classes and examples before inspection.
Mark all people, sources and decisions fictional.
Set a 10-minute limit and a stop rule for compulsive rechecking or distress.
Have a helper separate Reports A–C from the source cards and hidden key. If you prepared the packet and remember the answers, use it only as demonstration and score a fresh helper-prepared permutation.
3 · The method
Follow these steps in order
Declare error classes
Write arithmetic mismatch, contradiction, unsupported citation and source mismatch, with the evidence needed to confirm each.
Why: A suspicion becomes a detected error only when it matches a rule and ground truth.
Check: Each class has a definition and confirmation method.
Set prevalence and error costs
Read the packet’s class list and rank the fictional consequence of missing each class versus raising a false alarm. State which error class the next checklist prompt will target; do not invent a universal acceptable false-alarm count.
Why: Search behaviour and an effective repair depend on how rare and costly the target is.
Check: The sheet names the available error classes, highest-cost miss and the class-specific prompt before inspection.
Make an ordinary baseline pass
Read report A once and mark suspected errors without a checklist. Commit before seeing the key.
Why: The baseline shows what ordinary reading catches and misses.
Check: Every mark has a class and line reference.
Apply the verification sequence
On matched report B, check quantities, then internal claims, then citation existence, then whether each cited source supports the sentence.
Why: Separate passes reduce switching and target specific error mechanisms.
Check: All four passes are completed in order.
Require confirmation
For each mark, write the calculation, conflicting sentence or source-card evidence. Remove marks that lack confirmation.
Why: Evidence controls false alarms and separates anomaly from error.
Check: Every retained mark has a visible proof.
Reveal and classify outcomes
Compare with the answer key and count hits, misses, false alarms and correct rejections by class.
Why: Accuracy alone can hide rare but costly misses.
Check: The confusion matrix totals all eligible items.
Write and verify the correction
For every confirmed error, produce the corrected number, sentence or citation link and have the answer key or a second independent calculation verify it.
Why: Detection is incomplete when the proposed repair introduces another error.
Check: Each confirmed mark has one corrected replacement and a documented independent check.
Redesign around the costliest miss
Identify the missed class with highest declared cost, add one targeted prompt and test it on report C without extending total time.
Why: A repair should target the harmful failure, not merely make all checking slower.
Check: Report C shows whether the targeted miss falls and false alarms stay bounded.
4 · Worked example
See the whole method used once
Scenario
The supplied Reports A and B each contain one arithmetic mismatch, contradiction, source mismatch and unsupported citation; Report C adds clean controls and one source mismatch.
Walkthrough
Maya defines arithmetic and source mismatch before reading.
On the ordinary Report A pass she catches A1 but misses A2, A3 and A4, giving one hit, three misses and no false alarms.
On Report B she recomputes 14+19=33 and compares the body’s sample 40 with the summary’s 44.
She opens S2, sees no return-rate evidence, and separately checks the citation list to show that S8 does not exist.
Against the key she records four Report B hits, no misses and no false alarms, then adds ‘state the exact supported claim’ before applying the same pass to C3.
Result
Maya has evidence-backed detections and a class-specific repair. She has not shown professional audit or safety-monitoring competence.
5 · Right and wrong
Compare correct or safer execution with the common wrong version
Right and wrong comparison
Moment
Right / safer
Wrong / riskier
Why it matters
Marking
Name the error class and confirming evidence.
Highlight anything that feels odd.
Vague suspicion inflates false alarms.
Checklist
Use separate arithmetic, consistency and source passes.
Reread the whole page repeatedly without a target.
Unstructured repetition adds burden without a known mechanism.
Scoring
Count misses and false alarms by class.
Celebrate total accuracy on mostly clean statements.
High base-rate accuracy can hide rare-target failure.
Repair
Target the highest-cost missed class.
Simply double inspection time.
More time may not repair the actual search failure.
6 · Common mistakes
Spot the error and apply the correction
Common mistakes and corrections
Mistake
Fix
A citation exists, so it is marked valid.
Open the source card and compare the exact supported claim.
Clean statements are absent from the denominator.
Include correct rejections and false alarms for all eligible checks.
The answer key is opened during inspection.
Commit all marks first and timestamp reveal.
All error classes are averaged.
Report arithmetic, contradiction, citation and source mismatch separately.
7 · Practice
Turn the steps into a usable skill
First session
Learn four error definitions.
Complete one ordinary baseline report.
Apply the four-pass checklist to a matched report.
Build the confusion matrix.
Repair the costliest miss on a third report.
Repeat plan
Use one three-report set weekly for four weeks, with new synthetic content and predeclared error prevalence. Keep time fixed. After week two, include a set with very rare targets to test vigilance without increasing stakes.
Progress when
High-cost misses decline on fresh reports.
False alarms remain below the predeclared limit.
Every detection includes reproducible evidence.
Do not progress when
Checking becomes repetitive, distressing or hard to stop.
The material becomes real or consequential.
Ground truth or error definitions are disputed.
8 · Check the result
Measure what changed
Class-specific hits, misses and false alarms on keyed synthetic reports.
How: For each class, count hits, misses, false alarms and correct rejections. Calculate sensitivity as hits/(hits+misses) when at least one keyed error exists; report false-alarm count and eligible clean checks, time and burden, then compare matched baseline and checklist sets without merging classes.
Good result: On fresh keyed sets, the targeted high-cost miss is corrected with visible evidence while misses, false alarms and time all remain reported. No fixed ‘one false alarm per report’ rule is presented as a validated professional threshold.
This does not prove: It does not establish professional review, lie detection, diagnosis, safety clearance or transfer to rare real-world targets.
Self-check
What four error classes are allowed?
What evidence confirms each mark?
How many misses and false alarms occurred?
Did the repair target the highest-cost miss?
9 · Stop, adapt or get help
Keep the safety boundary practical
Stop and get help
Stop if checking becomes compulsive, distressing or continues past the time limit.
Stop if a real person, diagnosis, accusation or consequential document enters the task.
Route professional or safety-critical errors to the authorised review and escalation system.
Accessibility and adaptations
Use screen-reader headings and source cards with stable link labels.
Allow calculator and text-to-speech while keeping the pass order.
Split the report into one error class per view and preserve the same total eligible items.
10 · Evidence and limits
Why these instructions are here
primary research
Ultra-rare targets produced high miss rates even across very large numbers of search trials, so prevalence and misses must be reported.
Progress is saved only in this browser on this device.
Step-by-step learner mode
Each activity includes its success check, a nearby accessible alternative and an “I’m stuck” correction path. Alternatives preserve the target where possible; when they change the task, Titan labels them as related rather than equivalent.
01
Declare error classes
Write arithmetic mismatch, contradiction, unsupported citation and source mismatch, with the evidence needed to confirm each.
Why this step exists
A suspicion becomes a detected error only when it matches a rule and ground truth.
Success check
Each class has a definition and confirmation method.
I’m stuck on this step
Reset: Re-read this authored instruction — “Write arithmetic mismatch, contradiction, unsupported citation and source mismatch, with the evidence needed to confirm each.” — and its success check, then attempt only this step.
Possible snag: A citation exists, so it is marked valid.
Correction: Open the source card and compare the exact supported claim.
Possible snag: All error classes are averaged.
Correction: Report arithmetic, contradiction, citation and source mismatch separately.
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
02
Set prevalence and error costs
Read the packet’s class list and rank the fictional consequence of missing each class versus raising a false alarm. State which error class the next checklist prompt will target; do not invent a universal acceptable false-alarm count.
Why this step exists
Search behaviour and an effective repair depend on how rare and costly the target is.
Success check
The sheet names the available error classes, highest-cost miss and the class-specific prompt before inspection.
I’m stuck on this step
Reset: Re-read this authored instruction — “Read the packet’s class list and rank the fictional consequence of missing each class versus raising a false alarm. State which error class the next checklist prompt will target; do not invent a universal acceptable false-alarm count.” — and its success check, then attempt only this step.
Possible snag: The result from “Read the packet’s class list and rank the fictional consequence of missing each class versus raising a false alarm. State which error class the next checklist prompt will target; do not invent a universal acceptable false-alarm count.” does not yet meet this declared check: The sheet names the available error classes, highest-cost miss and the class-specific prompt before inspection.
Correction: Return to the start of “Set prevalence and error costs”, reduce complexity or pace, and repeat only the part needed to satisfy: “The sheet names the available error classes, highest-cost miss and the class-specific prompt before inspection.”
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
03
Make an ordinary baseline pass
Read report A once and mark suspected errors without a checklist. Commit before seeing the key.
Why this step exists
The baseline shows what ordinary reading catches and misses.
Success check
Every mark has a class and line reference.
I’m stuck on this step
Reset: Re-read this authored instruction — “Read report A once and mark suspected errors without a checklist. Commit before seeing the key.” — and its success check, then attempt only this step.
Possible snag: The result from “Read report A once and mark suspected errors without a checklist. Commit before seeing the key.” does not yet meet this declared check: Every mark has a class and line reference.
Correction: Return to the start of “Make an ordinary baseline pass”, reduce complexity or pace, and repeat only the part needed to satisfy: “Every mark has a class and line reference.”
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
04
Apply the verification sequence
On matched report B, check quantities, then internal claims, then citation existence, then whether each cited source supports the sentence.
Why this step exists
Separate passes reduce switching and target specific error mechanisms.
Success check
All four passes are completed in order.
I’m stuck on this step
Reset: Re-read this authored instruction — “On matched report B, check quantities, then internal claims, then citation existence, then whether each cited source supports the sentence.” — and its success check, then attempt only this step.
Possible snag: The result from “On matched report B, check quantities, then internal claims, then citation existence, then whether each cited source supports the sentence.” does not yet meet this declared check: All four passes are completed in order.
Correction: Return to the start of “Apply the verification sequence”, reduce complexity or pace, and repeat only the part needed to satisfy: “All four passes are completed in order.”
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
05
Require confirmation
For each mark, write the calculation, conflicting sentence or source-card evidence. Remove marks that lack confirmation.
Why this step exists
Evidence controls false alarms and separates anomaly from error.
Success check
Every retained mark has a visible proof.
I’m stuck on this step
Reset: Re-read this authored instruction — “For each mark, write the calculation, conflicting sentence or source-card evidence. Remove marks that lack confirmation.” — and its success check, then attempt only this step.
Possible snag: The result from “For each mark, write the calculation, conflicting sentence or source-card evidence. Remove marks that lack confirmation.” does not yet meet this declared check: Every retained mark has a visible proof.
Correction: Return to the start of “Require confirmation”, reduce complexity or pace, and repeat only the part needed to satisfy: “Every retained mark has a visible proof.”
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
06
Reveal and classify outcomes
Compare with the answer key and count hits, misses, false alarms and correct rejections by class.
Why this step exists
Accuracy alone can hide rare but costly misses.
Success check
The confusion matrix totals all eligible items.
I’m stuck on this step
Reset: Re-read this authored instruction — “Compare with the answer key and count hits, misses, false alarms and correct rejections by class.” — and its success check, then attempt only this step.
Possible snag: Clean statements are absent from the denominator.
Correction: Include correct rejections and false alarms for all eligible checks.
Possible snag: The answer key is opened during inspection.
Correction: Commit all marks first and timestamp reveal.
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
07
Write and verify the correction
For every confirmed error, produce the corrected number, sentence or citation link and have the answer key or a second independent calculation verify it.
Why this step exists
Detection is incomplete when the proposed repair introduces another error.
Success check
Each confirmed mark has one corrected replacement and a documented independent check.
I’m stuck on this step
Reset: Re-read this authored instruction — “For every confirmed error, produce the corrected number, sentence or citation link and have the answer key or a second independent calculation verify it.” — and its success check, then attempt only this step.
Possible snag: The result from “For every confirmed error, produce the corrected number, sentence or citation link and have the answer key or a second independent calculation verify it.” does not yet meet this declared check: Each confirmed mark has one corrected replacement and a documented independent check.
Correction: Return to the start of “Write and verify the correction”, reduce complexity or pace, and repeat only the part needed to satisfy: “Each confirmed mark has one corrected replacement and a documented independent check.”
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
08
Redesign around the costliest miss
Identify the missed class with highest declared cost, add one targeted prompt and test it on report C without extending total time.
Why this step exists
A repair should target the harmful failure, not merely make all checking slower.
Success check
Report C shows whether the targeted miss falls and false alarms stay bounded.
I’m stuck on this step
Reset: Re-read this authored instruction — “Identify the missed class with highest declared cost, add one targeted prompt and test it on report C without extending total time.” — and its success check, then attempt only this step.
Possible snag: The result from “Identify the missed class with highest declared cost, add one targeted prompt and test it on report C without extending total time.” does not yet meet this declared check: Report C shows whether the targeted miss falls and false alarms stay bounded.
Correction: Return to the start of “Redesign around the costliest miss”, reduce complexity or pace, and repeat only the part needed to satisfy: “Report C shows whether the targeted miss falls and false alarms stay bounded.”
Stop / get help: Stop if checking becomes compulsive, distressing or continues past the time limit.
Correct versus incorrect execution
These accessible process diagrams are built from the tutorial’s own right/wrong teaching. They are not anatomical illustrations and do not add technique beyond the canonical tutorial.
Marking — Vague suspicion inflates false alarms.
Correct / safer
Name the error class and confirming evidence.
Wrong / riskier
Highlight anything that feels odd.
Checklist — Unstructured repetition adds burden without a known mechanism.
Correct / safer
Use separate arithmetic, consistency and source passes.
Wrong / riskier
Reread the whole page repeatedly without a target.
Scoring — High base-rate accuracy can hide rare-target failure.
Correct / safer
Count misses and false alarms by class.
Wrong / riskier
Celebrate total accuracy on mostly clean statements.
Repair — More time may not repair the actual search failure.
Correct / safer
Target the highest-cost missed class.
Wrong / riskier
Simply double inspection time.
Method-structure checklist
10 of 10 structural checks present
✓ Ordered, Power-specific instructions — present
✓ Every activity has a success check — present
✓ Materials or supplied records are declared — present
✓ Measurement or assessment rule is present — present
✓ Tutorial-specific troubleshooting is present — present
✓ Stopping or escalation boundary is present — present
✓ Every activity has an adjacent alternative — present
✓ Correct-versus-incorrect comparison is present — present
✓ Evidence context is bound to the Power record — present
✓ Planning metadata is present — present
The method-readiness band and presence checklist assess tutorial presentation and are separate from evidence quality for the underlying Power. They are automated editorial aids, not human approval.
Manual editorial sign-off: Pending. This tutorial must not display a human-approved state until an identified editor signs the exact content hash.