Six Questions: Knowledge, Learning, Ability, and Learning Ability

Chinese version: /zh-hant/knowledge/six-question-learning-framework/

Status: Workspace working framework added on 2026-09-20 from a learner-supplied six-question progression. It is a guide for diagnosis and practice, not evidence of independent learner performance or a final theory of cognition.

The six questions form one system

The questions are usefully ordered, but the relationship is not a one-way assembly line:

External model resources
        ↓
Knowledge: usable, revisable world model
        ↓  tested and updated through learning
Understanding and task-specific ability
        ↓  monitored and improved by
Learning ability: the capacity to acquire, revise, and use new models and abilities

The learner-supplied shortcut, “knowledge is material; learning is process; ability is result; learning ability is engine,” is helpful as a first orientation. This workspace uses a more precise version: external information is material; knowledge is the internal world model built from and tested against it; learning is model revision; ability is reliable task performance using that model; learning ability is the set of capacities and habits that make this cycle more effective.

1. What is knowledge and what is its nature?

Knowledge is the relatively dependable, appropriately supported, and usable part of a learner’s revisable world model. It contains organised representations of objects, categories, relations, causes, conditions, procedures, and likely consequences.

Its practical value lies in connection and prediction: a model helps the learner explain what is happening, anticipate a changed case, choose an action, and revise the relation when evidence conflicts. This does not make knowledge certain or complete. It is a selective model of the world, not a literal copy.

External data, books, notes, videos, and LLM outputs are information or model resources. They become the learner’s knowledge only when the relevant relation has been constructed, retained, retrieved, and used.

2. What is learning?

Learning is the testing, revision, and reconstruction of the world model. It is not only input or remembering facts.

1. State the current relation and its assumed conditions.
2. Compare it with a changed case, feedback, or new evidence.
3. Locate the earliest failed assumption, missing variable, mechanism, or boundary condition.
4. Revise the smallest affected relation.
5. Integrate it with what remains accurate.
6. Test the reconstructed model in another case.

“Destruction” means narrowing, withdrawing, or replacing the failed part of a model. It does not mean discarding the whole knowledge network. The learning signal is not merely “I received new information,” but “I can now relate this to a prior model, state what changed, and use the revision to explain or predict.”

3. What is ability?

An ability is a relatively reliable capacity to perform a defined class of tasks under relevant conditions. It is a performance pattern, not information held once and not a global ranking of a person.

Ability often draws on knowledge, procedural fluency, cognitive supports, practice, and feedback. For example, a person can know propositions about swimming without yet having the ability to swim safely; swimming ability requires reliable performance in water under defined conditions.

Evidence for an ability includes accuracy, repeatability, reasonable independence, appropriate adaptation to variation, and error detection or correction.

4. What is learning ability?

Learning ability is a meta-level, task-dependent capacity to acquire, revise, connect, and use new models and abilities. It is not one fixed internal engine and should not be inferred from speed, confidence, or a single success.

Its observable components include:

• diagnosing the difference between a missing model, cognitive overload, and insufficient practice;
• selecting a suitable strategy for the goal and condition;
• retrieving and explaining a model without notes;
• using feedback to locate an error and revise the smallest relevant relation;
• monitoring whether a method is producing the intended delayed performance;
• transferring a revised relation to a changed case.

5. How can learning ability be improved?

Improve it by making the model-revision cycle more accurate, deliberate, and measurable. The strategies below are tools, not guarantees.

Strategy	Useful action	Boundary / evidence check
First-principles questioning	Ask which observations, definitions, assumptions, and causal relations the claim depends on.	Do not treat a personally chosen premise as unquestionable; identify what evidence would challenge it.
Plain-language teach-back	Explain the relation to a novice, identify where the explanation becomes vague, then repair the missing relation.	Clear speech alone does not prove understanding; add a changed-case prediction or application.
Deliberate practice and feedback	Choose one narrow subskill slightly beyond current performance, attempt it, inspect the earliest error, and retry with feedback.	Feedback must be relevant and the outcome must be measured; repetition of only familiar tasks can create fluency without transfer.
Metacognition	Before, during, and after a task, plan the strategy, monitor comprehension and load, and evaluate the result.	Confidence can be miscalibrated; compare judgments with retrieval, delayed performance, and error records.
Sleep, spacing, and retrieval	Schedule later retrieval attempts and protect sleep opportunity instead of relying only on massed rereading.	Effects depend on task, implementation, learner, and outcome measure; track delayed recall rather than assuming a universal gain.

The existing source record includes controlled studies relevant to retrieval plus feedback, sleep restriction plus spacing, and strategic self-reflection. They support specific interventions under defined conditions, not a universal recipe. See Experimental evidence relevant to study conditions.

A practical improvement loop

For one current topic, run this cycle:

1. Define the performance target: what must you explain, predict, do, or decide?
2. Elicit the current model without notes.
3. Choose one narrow weakness: missing relation, weak procedure, overloaded support, or poor strategy selection.
4. Use one strategy deliberately: premise check, teach-back, retrieval, feedback-driven drill, or spacing plan.
5. Measure an outcome now and later: accuracy, error type, time, independence, and delayed recall or transfer.
6. Revise the model and the learning strategy from the evidence.

This makes learning ability visible as an improving process rather than a label applied to a person.

Current evidence boundary

The learner has received the framework and has demonstrated some conditional reasoning. The learner has not yet independently classified all six concepts, selected a strategy from this map, or shown that an intervention improved delayed performance. Those remain performance questions.

Related records

• Raw learner framing: Six-question learning framework: learner framing
• Knowledge foundation: Knowledge as a World Model
• Learning process: Learning as Model Revision and Reconstruction
• Ability definition: Ability and understanding ability
• Practice loop: Practice Drill: Model revision and reconstruction
• Learning-ability drill: Practice Drill: Diagnose and Improve Learning Ability
• Evidence record: Experimental evidence relevant to study conditions