Choose a method by the question it helps answer. A writer checking a quotation, a product manager examining an assumption and an engineer reviewing a change may use the same reasoning practice while needing different evidence.
| The question in front of you | A useful method | An inspectable result |
|---|---|---|
| What have we assumed? | Examine the requirement from first principles. | An assumption linked to evidence and an observation that could contradict it. |
| What else changes with this decision? | Trace dependencies and affected people. | A specific connection to investigate before acting. |
| Which claim does this source support? | Check facts and attribution. | The passage, its source and the limits of the claim. |
| Would this edit improve the piece? | Review for substance, structure and voice. | A proposed change with a reader benefit and a record of what must survive. |
| What should a person decide? | Prepare the options and their consequences. | A concrete choice, supporting evidence and unresolved uncertainty. |
Test the assumption inside an interface
The five-checks article uses a constructed export screen with two settings: format and detail. All four combinations are valid.
| Format | Detail |
|---|---|
| CSV | Summary |
| CSV | Full |
| JSON | Summary |
| JSON | Full |
Start with CSV and Full. Change the format to JSON. If the detail setting resets to Summary, the interface has coupled two choices that the requirement keeps independent. Change detail next and check that format stays selected; examine the reverse transitions too.
A test that merely opens a second screen misses that relationship. There is also a separate design question: must the person inspect both controls at the same time? All four combinations can remain valid in a sequential interface that still fails a simultaneous-selection requirement.
Take only the reasoning the problem needs
The thinking framework also asks about changing interactions, analogies and the affected person’s task. An analogy can suggest a relationship to investigate; identify where it stops matching before transferring its solution.
For a contested factual claim, start with fact-checking. For a draft whose meaning disappears under generic language, use content review. For work that changes hands, use project records. These methods share standards of evidence without turning every kind of work into project management.
Continue with a case
- Think through a strategic decision: choose which explanation to investigate before committing.
- Review a draft: test an editorial criticism against the paragraph and its sources.
- Review an interaction: check the person’s whole task rather than its appearance alone.
- Review code: reject agreement on a shared mistake.
Rajiv Pant wrote the shared methods linked here. Suggest an improvement with the question, the method’s output and the evidence that changed your conclusion. A case where a method was unnecessary or misleading can be useful too.