bookforge-learning-practice-auditor（Bookforge 学习实践审核员）

Learning Practice Auditor

When to Use

Something about how you or your learners study is producing less retention than expected. You may be putting in hours that do not convert to durable learning. You may be designing training that graduates confident-feeling but underprepared learners.

This skill is the diagnostic hub: it identifies which of five named ineffective practices are present, explains the mechanism creating the illusion of learning, rates the severity, and routes to the specific corrective skill.

Typical entry points:

• Learner: "I study for hours but blank on tests"
• Learner: "I highlight everything but still can't recall it"
• Teacher or trainer: "Why aren't my learners retaining this?"
• Instructional designer: "Is there anything wrong with this training structure?"
• Anyone asking: "Am I studying the wrong way?"

Mode: Hybrid — The agent runs the audit and produces the report. The human and/or agent then invokes the recommended corrective skills.

This skill diagnoses. It does not redesign. After the audit, each corrective action names the skill to invoke next.

Step 1: Intake — Gather Current Practices

Why this step: Anti-pattern detection requires knowing what the learner or designer is actually doing, not what they think they are doing. Many learners describe their method as "studying" without specifying what that means. The intake questions force specificity.

Ask the following. If the user has already provided answers in their message, extract the answers without asking again.

Required

1. What are you trying to learn, and why?

(Subject, course, skill, domain — and the performance event: exam, job task, certification, presentation)

1. Walk me through a typical study session from start to finish.

(What does the learner actually do, in sequence, with the material?)

1. How do you decide when you know something well enough to move on?

(The mastery signal — feeling of familiarity, re-reading smoothly, immediate practice score, etc.)

1. How is the learning structured over time?

(One long session vs. spaced sessions; one topic completed before the next; mixed topics within a session)

Helpful if available

1. Do you have a course outline, training schedule, or study plan I can read?

→ If yes, use Read to scan it for blocked/massed structure signals.

1. What does assessment look like? (Multiple-choice, essay, performance demo, on-the-job transfer)

1. Have you already taken any tests or received feedback on your performance?

If context is still insufficient after asking, note the gap in the audit report and flag it as a finding ("Unable to assess [pattern] — insufficient information about [X]").

Step 2: Anti-Pattern Detection

Why this step: Each anti-pattern has specific detection signals that appear in how the learner describes their practice. This step converts free-form description into a structured finding list.

Check each anti-pattern against the intake answers. Mark each as Detected, Absent, or Insufficient information.

AP-1: Rereading Trap

What to look for:

• "I review my notes / re-read the chapter / highlight the key parts"
• Primary study method is reading or looking at material rather than testing recall
• Mastery signal is "it feels familiar" or "I can follow along when I read it"
• Study sessions described as long reading or annotation sessions

Mechanism (explain this to the user when detected): Each time you reread well-organized text, your processing becomes faster and smoother — a state called *fluency*. That fluency feels like understanding and mastery, but it is really your brain's familiarity with the visual form of the text. It cannot distinguish between "I have encoded this in long-term memory" and "I have seen these words before." On a test, you must retrieve — reconstruct from memory — and fluency does not train that. The 2010 study: students who took a recall test after reading retained 50% more a week later than those who reread. Crammers in a 1978 study forgot 50% of their initial recall within two days; retrieval-practice groups forgot only 13%.

Corrective action: Replace rereading sessions with self-testing sessions → invoke retrieval-practice-study-system

AP-2: Massed Practice Delusion

What to look for:

• "I do all the problems for chapter 5 before moving to chapter 6"
• Training sessions run drill repetitions on one skill until it feels solid, then move on
• Study sessions are concentrated blocks on one topic
• Cramming before an exam described as the primary retention strategy

Sub-pattern — Blocked Practice Trap: Problems or practice tasks are grouped by type. This removes the critical "what kind of problem is this?" sorting decision — which is exactly what real performance requires. College students practicing geometry volumes: blocked practice averaged 89% correct during training but only 20% on a test one week later. Interleaved practice: 60% during training, 63% on the test — a 215% improvement in transfer. (Ch3)

Sub-pattern — Familiarity Trap: After massed repetition, the material feels familiar (System 1 recognition). But the test requires System 2 retrieval and application. Rapid, blocked practice encodes learning in a cognitively simpler neural representation than spaced, varied practice — less flexible, less transferable.

Mechanism: Rapid repetition draws on short-term working memory. Durable encoding requires a consolidation process that unfolds over hours and days. Massed practice produces fast visible gains that feel like mastery, but the gains are not consolidated. In the surgical microsurgery study: residents who had 4 lessons spaced one week apart outperformed those trained in a single day on *all* measures — elapsed time, movement efficiency, and success rate — one month later. 16% of the massed-practice group damaged tissue beyond repair.

Corrective action: Replace massed and blocked practice with spaced, interleaved, and varied practice → invoke desirable-difficulty-classifier to classify present and absent difficulty strategies and get redesign recommendations.

AP-3: Illusions of Knowing Cluster

What to look for:

• Mastery signal is "it feels familiar" or "I recognized the answer when I saw it"
• High confidence before a test, low score after
• "I studied this — I don't know why I blanked"
• Learner or trainer expresses surprise at gaps revealed on assessment
• Anyone relying on "just reading through" as a review method

Note: This anti-pattern overlaps with AP-1 and AP-2. Log it separately when the primary problem is *miscalibrated confidence*, not just the study method used.

Mechanism: You rely on System 1 signals — fluency, familiarity, ease of processing — to judge mastery. These signals are unreliable. The Dunning-Kruger research: students at the 12th percentile in logic ability believed they were at the 68th percentile on average. Fluency illusion: re-reading a clear text increases the sensation that it is simple and already known. Hindsight bias: after learning the answer, you feel you "knew it all along," which makes you underestimate future study needs.

Corrective action: Run a calibration cycle to identify the specific distortion and install reliable mastery signals → invoke learning-calibration-audit

AP-4: Learning Styles Myth

What to look for:

• "I'm a visual learner, so I only use videos"
• Training is segmented by assessed learning style
• Learner or designer says matching style to instruction is important for retention
• Learner avoids certain modalities because they are "not their style"

The Pashler Validity Test: For a study to prove learning styles theory, it must: (1) classify learners by style, (2) randomly assign them to style-matched vs. style-mismatched instruction on the same content, (3) show that style-matched learners outperform style-mismatched learners. In the 2008 Pashler et al. review: no published study meets this design standard. Those that do flatly contradict the theory. Studies found that matching instruction to the *nature of the content* (visual for geometry, verbal for poetry) benefits *all* learners regardless of their style preference.

Mechanism: Learning preferences are real — but preference does not equal differential effectiveness. A learner's self-reported "style" predicts what they *prefer*, not what produces the best retention. Designing instruction around style matching misses the variance drivers that actually matter: prior knowledge, retrieval practice, spacing, and structure-building ability. More than 70 competing style theories exist with contradictory dimensions and no cross-validation — itself a strong signal the construct lacks scientific coherence.

Corrective action: Drop style-based instructional segmentation; shift to content-nature matching and apply evidence-based strategies to all learners regardless of style preference → invoke desirable-difficulty-classifier

AP-5: Errorless Learning Myth

What to look for:

• Training shows the answer before asking the learner to attempt it
• Flash cards are answer-first or answer-alongside
• Instructor corrects immediately before learner generates a response
• Stated rationale is "we don't want to practice errors"

Mechanism: Generating an answer — even a wrong one — before seeing the correct answer strengthens the subsequent memory of the correct answer more than passive presentation. This is the *generation effect*: word pairs where learners had to fill in a missing letter (foot → s_ _e) produced better later recall than pairs presented intact (foot → shoe). Errors made in *generative attempts* followed by corrective feedback are part of the encoding process, not its enemy. Preventing the attempt removes the cognitive effort that creates durable encoding.

Key distinction: Uncorrected misconceptions are harmful. The solution is corrective feedback — not preventing the attempt.

Corrective action: Restructure practice to require generation before exposure — pre-tests, cued retrieval, retrieval-first flash cards → invoke retrieval-practice-study-system

Step 3: Classify Severity

Why this step: Not all detected anti-patterns require the same urgency. Severity determines what to fix first and what to route immediately vs. note for later.

For each detected anti-pattern, assign a severity:

Step 4: Produce the Audit Report

Why this step: A written report makes findings actionable. The learner or designer can share it, return to it, and use the corrective routing as a checklist.

Structure the report as follows:

## Learning Practice Audit — [Subject / Program Name]

### Summary
[1-3 sentence overview: how many anti-patterns found, dominant severity level, top recommendation]

### Findings

#### [Anti-Pattern Name] — Severity: [Critical / Moderate / Low]
**Detected signals:** [What in their description triggered this finding]
**Mechanism:** [Why it feels effective but produces weak retention — in plain language]
**Evidence:** [1-2 specific research findings with numbers]
**Corrective action:** [Named replacement practice] → invoke `[skill-name]`

[Repeat for each detected anti-pattern]

### What to Do Next
[Prioritized list of corrective skills to invoke, in order of severity]
1. [Corrective skill] — addresses [anti-pattern(s)]
2. ...

### What Is Already Working
[Any detected evidence-based practices — note and reinforce these]

Examples

Example 1 — Individual Student (Critical: Rereading + Massed Practice)

User description: "I study organic chemistry by reading the chapter twice, then doing all the practice problems for each reaction type before moving on. I feel pretty solid during my sessions but I've failed two midterms."

Audit findings:

• AP-1: Rereading Trap — Critical. Primary review method is re-reading. The fluency gained from reading the chapter twice is not producing retrievable memory. The learner is using feeling-solid as the mastery signal, which tracks fluency not recall.
• AP-2: Massed Practice Delusion (Blocked Practice) — Critical. All practice problems for one reaction type before moving to the next = blocked massed practice. This eliminates the "which type of problem is this?" discrimination skill that the exam actually tests. The 20% vs. 63% result on blocked vs. interleaved geometry test is directly applicable.
• AP-3: Illusions of Knowing — Moderate. "I feel pretty solid during sessions but failed two midterms" = textbook miscalibration. The mastery signal is unreliable.

Corrective routing:

1. retrieval-practice-study-system — replace re-reading with self-quizzing and build spacing schedule
2. desirable-difficulty-classifier — redesign the problem practice to interleave reaction types
3. learning-calibration-audit — install reliable mastery signals to replace the fluency-feeling

Example 2 — Corporate Trainer (Moderate: Learning Styles Design)

User description: "We have our training team assess employees' learning styles before onboarding. Visual learners get diagrams-first modules; auditory learners get voice-over walkthroughs. We invested in this system two years ago."

Audit findings:

• AP-4: Learning Styles Myth — Critical. The entire curriculum segmentation is organized around a theory with no empirical validation (Pashler et al., 2008). Investment in assessment tools and differentiated modules is producing no learning benefit relative to matched instruction. The content-nature matching finding is relevant: geometry-style content (spatial workflows, process maps) should be visual for *all* learners, regardless of assessed style.

Corrective routing:

1. desirable-difficulty-classifier — audit the onboarding modules for evidence-based difficulty strategies that will benefit all learners regardless of style
2. Redirect the investment in style assessment toward spaced-practice scheduling and retrieval-based knowledge checks

Example 3 — Teacher (Low: Partial Massed Practice)

User description: "I do a mixed review at the end of each unit. I mostly teach one topic per day but sometimes bring in earlier material. Students self-quiz once a week using flashcards."

Audit findings:

• AP-2: Massed Practice — Low. One topic per day is still somewhat blocked, but the weekly mixed review and flashcard self-quizzing are evidence-based practices that partially offset this. The dominant strategy is already better than pure massed practice.
• No other anti-patterns detected from this description.

Recommendation: This practice is already above average. The flashcard self-quizzing is the highest-value element — reinforce and expand it. For the next redesign cycle: consider interleaving topics within sessions, not just end-of-unit reviews.

Reference

Full evidence profiles, all research citations, and detailed sub-pattern descriptions for each anti-pattern: → references/anti-pattern-reference-table.md

For mechanism details on illusions of knowing, see also: → learning-calibration-audit references

License

This skill is licensed under CC-BY-SA-4.0. Source: BookForge — Make It Stick: The Science of Successful Learning by Peter C. Brown, Henry L. Roediger III, Mark A. McDaniel.

Related BookForge Skills

Install related skills from ClawhHub:

• clawhub install bookforge-retrieval-practice-study-system
• clawhub install bookforge-desirable-difficulty-classifier
• clawhub install bookforge-learning-calibration-audit

Or install the full book set from GitHub: bookforge-skills