Section 376 of 440
Complete canonical tutorial. This reader section contains the same teaching body as PWR-194 · AI-assisted memory. Open the Power dossier.
PWR-194 · SELF GUIDED full tutorial
Turn a source document into checked memory cues without mistaking AI retrieval for human memory
A twelve-sentence museum notice and six fixed AI cards make memory assistance auditable from source to recall. Verify every card against its sentence number, compare retrieval with the aid, rehearse an outage and test five facts unaided after 48 hours. Four delayed recalls do not establish a general memory gain or protection from AI error.
1 · Permission and limits
Know exactly what you may do
2 · Get ready
Gather what you need and check the starting conditions
What you need
- Declared AI-assisted memory fixture: A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers.
- Setup aid for Recall before assistance: Decide which facts truly need unaided recall and which can safely remain externally supported.
- AI-assisted memory log: Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss; retain AI-assisted memory errors, assistance, stop and fallback.
- Memory-aid evidence sheet: use “Humans inherit artificial intelligence biases” to mark every AI-inserted museum fact that must be source-checked. From the AI-tutoring RCT, copy the learning comparison and measured outcome without calling retrieval a memory cure. The NIST Generative AI Profile supplies the provenance, outage and delayed no-AI checks for the six cards.
- Complete numbered museum notice: S1 On Monday the North Hall opens at 09:00. S2 On Tuesday the North Hall opens at 10:00. S3 On Saturday the North Hall opens at 11:00. S4 The ceramic display moves from Room C to Room B. S5 The map exhibit moves from Room A to Room D. S6 Questions about access changes go to desk extension 204, not the ticket line. S7 The café keeps its usual hours. S8 The west lift remains available. S9 The sculpture garden entrance is unchanged. S10 Printed maps are beside the main desk. S11 The notice applies for the first week of June. S12 Staff will issue a new numbered notice if any fact changes.
- Fixed six-card AI proposal (keep the notice visible for verification): Q1 Monday opening? 09:00, cites S1; Q2 Tuesday opening? 10:00, cites S2; Q3 Saturday opening? 11:00, cites S3; Q4 Ceramic display’s new room? Room C, cites S4; Q5 Map exhibit’s new room? Room D, cites S5; Q6 Access-change contact? extension 204, cites S6. Q4 is false because S4 says Room B; quarantine it, leaving five verified cards for the demonstration.
- Transfer notice for the later no-card check: T1 Wednesday garden opens 08:30. T2 Friday garden opens 09:30. T3 The seed exhibit moves from Bay 2 to Bay 5. T4 Access questions go to desk extension 310. T5 The east gate remains closed. T6 A new numbered notice replaces this one after any change. Ask four questions covering both times, the new bay and the contact.
Before you start
- Use only fictional, public or approved non-sensitive text.
- Decide which facts truly need unaided recall and which can safely remain externally supported.
- Start check for AI-assisted memory: Every later answer can point to a stable source sentence.
- Top-of-sheet stop for AI-assisted memory: Stop if a generated card cannot be verified against the original source.
3 · The method
Follow these steps in order
- Create a clean source
Number the sentences in the fictional notice and record AI version, prompt and data boundary.
Why: Every later answer can point to a stable source sentence.
Check: Every later answer can point to a stable source sentence.
- Recall before assistance
Read once, close the notice and write what you remember with confidence.
Why: The unaided baseline is preserved.
Check: The unaided baseline is preserved; verify it in the numbered museum notice.
- Ask for source-linked cards
Request one question, short answer and sentence number per fact; forbid facts not present in the notice.
Why: Six proposed cards each cite a sentence.
Check: Six proposed cards each cite a sentence.
- Verify every card
Reopen the notice and mark answer and citation correct, incomplete or false. Delete rather than memorise a false card.
Why: Only verified cards enter practice.
Check: Only verified cards enter practice; verify it in the numbered museum notice.
- Retrieve in both directions
Answer question-to-fact and cue-to-source; correct from the notice after each attempt.
Why: The learner can produce the fact and find its source.
Check: The learner can produce the fact and find its source.
- Test device-on retrieval
With cards available, record correct answers, latency and citation success on six shuffled questions.
Why: Fast retrieval without a valid source is not scored correct.
Check: Fast retrieval without a valid source is not scored correct.
- Test delayed unaided memory
After 48 hours, answer the same fact types without AI or cards, then use a new notice for transfer.
Why: Device-on and unaided scores remain separate.
Check: Device-on and unaided scores remain separate.
- Rehearse system loss
Hide the cards, use the original notice index and record time to recover each critical fact.
Why: A verified non-AI fallback works when the generated aid is unavailable.
Check: A verified non-AI fallback works when the generated aid is unavailable.
4 · Worked example
See the whole method used once
Scenario
A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers.
Walkthrough
- Number the museum notice’s 12 sentences and recall four facts before using AI.
- Generate six cards; card 4 says Room C but cites S4, which says Room B, so quarantine that card rather than rehearsing it.
- Practise the five verified cards in both directions until every answer points to its sentence.
- With cards available, retrieve 5/5 facts in 70 seconds with five valid citations.
- After 48 hours without cards, recall 4/5 and mark the room change uncertain rather than guessing.
- Simulate account loss and recover all five facts from the numbered notice in four minutes.
Result
The aid supports verified retrieval and four of five facts are recalled unaided after 48 hours. It does not show general memory improvement or immunity to AI-induced error.
5 · Right and wrong
Compare correct or safer execution with the common wrong version
| Moment | Right / safer | Wrong / riskier | Why it matters |
|---|---|---|---|
| Source linkage | Keep sentence numbers on every card. | Store uncited AI summaries as memory. | Generated errors can be rehearsed into later recall. |
| Unaided measure | Test recall after 48 hours with no card access. | Count instant card retrieval as improved human memory. | External access changes performance but is not unaided retention. |
| Sensitive data | Use a fictional notice when completing the numbered museum notice. | Upload personal notes or confidential records for convenience. | Memory aids can create privacy and retention risks. |
| Fallback in numbered museum notice | Keep a source index and outage drill. | Assume the AI or account will always be available. | System loss can remove access to offloaded information. |
6 · Common mistakes
Spot the error and apply the correction
| Mistake | Fix |
|---|---|
| Store uncited AI summaries as memory. | Verify and link each fact before practice. |
| Count instant card retrieval as improved human memory. | Report device-on and device-off results separately. |
| Upload personal notes or confidential records for convenience. | Remove identifying data or use an approved local process. |
| Assume the AI or account will always be available. | Practise recovery from the original verified source. |
7 · Practice
Turn the steps into a usable skill
First session
- Museum-notice memory run: A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers.
- Numbered source and baseline recall: Create a clean source: Number the sentences in the fictional notice and record AI version, prompt and data boundary.
- Card verification and bidirectional retrieval: Verify every card, then Retrieve in both directions.
- Uncited-summary correction: if “Store uncited AI summaries as memory.” appears, apply “Verify and link each fact before practice.”
- Source-index outage drill: Rehearse system loss: Hide the cards, use the original notice index and record time to recover each critical fact.
Repeat plan
Build one five-to-eight-card set weekly. Test after 48 hours without cards and perform an outage drill every second week; progress only when source traceability stays perfect and delayed errors do not rise.
Progress when
- Every later answer can point to a stable source sentence.
- The unaided baseline is preserved.
- Six proposed cards each cite a sentence.
- All active cards match their sources, device-on retrieval includes valid citations, delayed unaided recall improves over baseline, and an outage fallback retrieves every critical fact.
Do not progress when
- Do not continue while this error remains: Store uncited AI summaries as memory.
- Pause until this correction works: Report device-on and device-off results separately.
- This AI-assisted memory stop ends the block: Stop if a generated card cannot be verified against the original source.
8 · Check the result
Measure what changed
Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss
How: Baseline fixture: A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers. Enter success checks from “Create a clean source” and “Retrieve in both directions”. If “Store uncited AI summaries as memory.” occurs, apply its named fix; then score Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss on the unused “Rehearse system loss” item.
Good result: All active cards match their sources, device-on retrieval includes valid citations, delayed unaided recall improves over baseline, and an outage fallback retrieves every critical fact.
This does not prove: Boundary for AI-assisted memory: “Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss” describes only A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers. It cannot establish “An AI second brain permanently expands memory”.
Self-check
- Without the example, demonstrate: Every later answer can point to a stable source sentence.
- Find the fault in this attempt: “Store uncited AI summaries as memory.” Apply “Verify and link each fact before practice.”; what changes?
- What evidence in the completed record shows that this is wrong: “Count instant card retrieval as improved human memory.”?
- AI-assisted memory stop decision: Stop if a generated card cannot be verified against the original source.
9 · Stop, adapt or get help
Keep the safety boundary practical
Stop and get help
- Stop if a generated card cannot be verified against the original source.
- Do not upload health, employment, legal or private records to an unapproved AI.
- Rebuild the set after the source or model changes; do not mix old and new facts.
Accessibility and adaptations
- Use audio, large print, braille or image-supported cards while keeping the source locator.
- Reduce card count and lengthen delay gradually; do not remove a communication or memory aid to manufacture independence.
10 · Evidence and limits
Why these instructions are here
- primary research
Registered support for AI-assisted memory: “Humans inherit artificial intelligence biases”. It bears on Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss inside the AI-assisted memory fixture. It does not validate “An AI second brain permanently expands memory”.
Humans inherit artificial intelligence biases - primary research
Constraint for AI-assisted memory, drawn from “AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting”: Expectation of external access changed what people recalled, and biased AI exposure produced later unassisted errors rather than improvement.
AI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting - official guidance
Generative-AI control for AI-assisted memory: record version and source checks, then keep human review plus failure behaviour visible. Those controls are exercised in “Test delayed unaided memory” while measuring Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss.
Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile
Limits
- AI-assisted memory boundary: interpret “Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss” only for A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers.
- A successful result does not establish “An AI second brain permanently expands memory”.
- AI-assisted memory limiting finding: Expectation of external access changed what people recalled, and biased AI exposure produced later unassisted errors rather than improvement.
- No perfect-performance claim for AI-assisted memory: the evidence register does not make “Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss” universal, consequence-free or flawless in A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers.
- Scope remains AI-assisted memory: A 12-sentence fictional museum notice names three opening times, two room changes and one contact rule. The AI proposes six question–answer cards linked to sentence numbers. Recheck the comparator, support and “Device-on retrieval accuracy, source traceability and latency plus delayed unaided recall and performance after system loss” after any configuration change.
Open the complete canonical research register
- Primary empirical supportLimiting / contraryHumans inherit artificial intelligence biases
Lucía Vicente; Helena Matute · 2023 · Primary research
- Limiting / contraryAI tutoring outperforms in-class active learning: an RCT introducing a novel research-based design in an authentic educational setting
Greg Kestin; Kelly Miller; Anna Klales; Timothy Milbourne; Gregorio Ponti · 2025 · Primary research
- Limiting / contraryRelying on AI at work reduces self-efficacy, ownership, and meaning while active collaboration mitigates the effects
Elena Hayoung Lee; Yidan Yin; Nan Jia; Cheryl J. Wakslak · 2026 · Primary research
- Primary empirical supportLimiting / contraryGoogle Effects on Memory: Cognitive Consequences of Having Information at Our Fingertips
Betsy Sparrow; Jenny Liu; Daniel M. Wegner · 2011 · Primary research
- Limiting / contraryExperimental evidence on the productivity effects of generative artificial intelligence
Shakked Noy; Whitney Zhang · 2023 · Primary research
- Limiting / contraryOfficial boundary contextArtificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile
C. Autio; R. Schwartz; J. Dunietz; S. Jain; M. Stanley; E. Tabassi; P. Hall; National Institute of Standards and Technology · 2024 · Official standard
- Limiting / contraryOfficial boundary contextNIST Privacy Framework: A Tool for Improving Privacy Through Enterprise Risk Management, Version 1.0
National Institute of Standards and Technology · 2020 · Official standard
Read the complete evidence interpretation on the Power dossier.
Tutorial delivery controls
Learn, adapt, troubleshoot and resume
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.
Create a clean source
Number the sentences in the fictional notice and record AI version, prompt and data boundary.
Every later answer can point to a stable source sentence.
Every later answer can point to a stable source sentence.
I’m stuck on this step
Reset: Re-read this authored instruction — “Number the sentences in the fictional notice and record AI version, prompt and data boundary.” — and its success check, then attempt only this step.
Possible snag: Upload personal notes or confidential records for convenience.
Correction: Remove identifying data or use an approved local process.
Stop / get help: Stop if a generated card cannot be verified against the original source.
Recall before assistance
Read once, close the notice and write what you remember with confidence.
The unaided baseline is preserved.
The unaided baseline is preserved; verify it in the numbered museum notice.
I’m stuck on this step
Reset: Re-read this authored instruction — “Read once, close the notice and write what you remember with confidence.” — and its success check, then attempt only this step.
Possible snag: The result from “Read once, close the notice and write what you remember with confidence.” does not yet meet this declared check: The unaided baseline is preserved; verify it in the numbered museum notice.
Correction: Return to the start of “Recall before assistance”, reduce complexity or pace, and repeat only the part needed to satisfy: “The unaided baseline is preserved; verify it in the numbered museum notice.”
Stop / get help: Stop if a generated card cannot be verified against the original source.
Ask for source-linked cards
Request one question, short answer and sentence number per fact; forbid facts not present in the notice.
Six proposed cards each cite a sentence.
Six proposed cards each cite a sentence.
I’m stuck on this step
Reset: Re-read this authored instruction — “Request one question, short answer and sentence number per fact; forbid facts not present in the notice.” — and its success check, then attempt only this step.
Possible snag: The result from “Request one question, short answer and sentence number per fact; forbid facts not present in the notice.” does not yet meet this declared check: Six proposed cards each cite a sentence.
Correction: Return to the start of “Ask for source-linked cards”, reduce complexity or pace, and repeat only the part needed to satisfy: “Six proposed cards each cite a sentence.”
Stop / get help: Stop if a generated card cannot be verified against the original source.
Verify every card
Reopen the notice and mark answer and citation correct, incomplete or false. Delete rather than memorise a false card.
Only verified cards enter practice.
Only verified cards enter practice; verify it in the numbered museum notice.
I’m stuck on this step
Reset: Re-read this authored instruction — “Reopen the notice and mark answer and citation correct, incomplete or false. Delete rather than memorise a false card.” — and its success check, then attempt only this step.
Possible snag: The result from “Reopen the notice and mark answer and citation correct, incomplete or false. Delete rather than memorise a false card.” does not yet meet this declared check: Only verified cards enter practice; verify it in the numbered museum notice.
Correction: Return to the start of “Verify every card”, reduce complexity or pace, and repeat only the part needed to satisfy: “Only verified cards enter practice; verify it in the numbered museum notice.”
Stop / get help: Stop if a generated card cannot be verified against the original source.
Retrieve in both directions
Answer question-to-fact and cue-to-source; correct from the notice after each attempt.
The learner can produce the fact and find its source.
The learner can produce the fact and find its source.
I’m stuck on this step
Reset: Re-read this authored instruction — “Answer question-to-fact and cue-to-source; correct from the notice after each attempt.” — and its success check, then attempt only this step.
Possible snag: Store uncited AI summaries as memory.
Correction: Verify and link each fact before practice.
Possible snag: Assume the AI or account will always be available.
Correction: Practise recovery from the original verified source.
Stop / get help: Stop if a generated card cannot be verified against the original source.
Test device-on retrieval
With cards available, record correct answers, latency and citation success on six shuffled questions.
Fast retrieval without a valid source is not scored correct.
Fast retrieval without a valid source is not scored correct.
I’m stuck on this step
Reset: Re-read this authored instruction — “With cards available, record correct answers, latency and citation success on six shuffled questions.” — and its success check, then attempt only this step.
Possible snag: Count instant card retrieval as improved human memory.
Correction: Report device-on and device-off results separately.
Stop / get help: Stop if a generated card cannot be verified against the original source.
Test delayed unaided memory
After 48 hours, answer the same fact types without AI or cards, then use a new notice for transfer.
Device-on and unaided scores remain separate.
Device-on and unaided scores remain separate.
I’m stuck on this step
Reset: Re-read this authored instruction — “After 48 hours, answer the same fact types without AI or cards, then use a new notice for transfer.” — and its success check, then attempt only this step.
Possible snag: The result from “After 48 hours, answer the same fact types without AI or cards, then use a new notice for transfer.” does not yet meet this declared check: Device-on and unaided scores remain separate.
Correction: Return to the start of “Test delayed unaided memory”, reduce complexity or pace, and repeat only the part needed to satisfy: “Device-on and unaided scores remain separate.”
Stop / get help: Stop if a generated card cannot be verified against the original source.
Rehearse system loss
Hide the cards, use the original notice index and record time to recover each critical fact.
A verified non-AI fallback works when the generated aid is unavailable.
A verified non-AI fallback works when the generated aid is unavailable.
I’m stuck on this step
Reset: Re-read this authored instruction — “Hide the cards, use the original notice index and record time to recover each critical fact.” — and its success check, then attempt only this step.
Possible snag: The result from “Hide the cards, use the original notice index and record time to recover each critical fact.” does not yet meet this declared check: A verified non-AI fallback works when the generated aid is unavailable.
Correction: Return to the start of “Rehearse system loss”, reduce complexity or pace, and repeat only the part needed to satisfy: “A verified non-AI fallback works when the generated aid is unavailable.”
Stop / get help: Stop if a generated card cannot be verified against the original source.
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.
Keep sentence numbers on every card.
Store uncited AI summaries as memory.
Test recall after 48 hours with no card access.
Count instant card retrieval as improved human memory.
Use a fictional notice when completing the numbered museum notice.
Upload personal notes or confidential records for convenience.
Keep a source index and outage drill.
Assume the AI or account will always be available.
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.