Practice Drill: Model Revision and Reconstruction
A no-notes drill for revising an overgeneralised world-model relation using evidence.
Chinese version: /zh-hant/practice/drills/model-revision-and-reconstruction/
Subskill
Revise and reconstruct one overgeneralised relation in a world model rather than merely replacing an answer.
Performance condition
Without notes, complete the task in 8 minutes. Use the claim below and the new observation. Write six numbered answers.
Initial claim: “Working from home always makes people more productive.”
New observation: A team reports that some members finish more focused individual work at home, while others lose time because of unsuitable workspace, interruptions, unclear coordination, or work requiring rapid collaboration.
Required response
- State the initial model and its implied conditions.
- Identify the mismatch between the model and the observation.
- Name the earliest failed assumption, missing variable, or boundary condition.
- Write a more accurate conditional relation.
- Explain what remains useful in the old model and how the new relation connects to it.
- Predict what should happen in one changed case, and name an observation that would require another revision.
Success criteria
- Separates the observation from the model’s interpretation.
- Identifies a specific failure, not only “there are many factors.”
- Produces a conditional relation rather than an unconditional reversal.
- Retains any valid part of the old model while narrowing its scope.
- Makes a prediction that follows from the revised relation.
- Names plausible counterevidence or a condition requiring further revision.
Common error
Replacing “always more productive” with “working from home never helps”. This reverses an overgeneralisation without reconstructing the causal model.
Constraint
One attempt only before checking the model note. Repeat after two days with a different claim and no template.
Next action
- If successful: Save the unedited response as an attempt, then complete a second case from study, health, or work.
- If unsuccessful: Locate the earliest missing part: initial model, mismatch, failed assumption, revised relation, integration, or test. Retry only that part with a simpler claim.