| name | cognitive-loop-map |
| description | Use this to reconstruct how someone repeatedly moves from intent to learning, and how that loop shifts by task type. |
Cognitive Loop Map
The Big Idea
The deliverable is HOW work gets thought through, not which adjective
describes the thinker. Reconstruct transitions, count dominant paths, then
map where the loop bends under different conditions.
The Loop
Trace each episode through:
INTENT -> PROBLEM REPRESENTATION -> PLANNING -> INFORMATION GATHERING ->
DECISION -> EXECUTION -> MONITORING -> CORRECTION -> VALIDATION -> LEARNING
For every transition, find real instances. Note the common path and the
common skips. A skipped VALIDATION stage is a finding, not an accident.
The Condition Table
Then split by condition. The loop usually changes shape:
| Condition | What to compare |
|---|
| Unfamiliar task | Where does uncertainty go first |
| Familiar task | Which stages get compressed or skipped |
| Bug | Entry point, hypothesis discipline |
| Architecture work | Framing depth, option generation |
| Frontend / visual work | Iteration pattern, verification style |
| Production incident | Speed vs evidence trade |
| Research task | Depth, stopping rule |
| High-stakes operation | Deliberation, reversibility checks |
| Failed first attempt | Hand off to failure-loops |
| Different AI model | Same human, changed loop - attribute carefully |
Method
- Pick 20-40 representative episodes across conditions.
- Mark each stage transition as present, skipped, or outsourced to AI.
- Draw the dominant loop per condition.
- Name the two or three transitions that most often go wrong.
When It Backfires
- Inferring the loop from polished final summaries instead of raw moves.
- Confusing the model's loop with the human's. Attribute first
(episode-method).
- Presenting one idealized loop when there are really three: calm mode,
pressure mode, delegated mode.
One-Line Memory
Map transitions, not traits.
