| name | debugging-trees |
| description | Use this on bug hunts and ambiguous problems. Reconstruct the hypothesis tree and watch how the mental model changes. |
Debugging Trees
What to Extract
For each debugging episode, score:
- how many candidate explanations were generated before the first fix
- anchoring: was the first plausible explanation also the last one tested
- discriminating evidence: tests designed to separate hypotheses, or only
tests confirming the favorite
- cheapest-first ordering: were cheap probes run before expensive ones
- model updates: did contradicting evidence change the working theory
- symptom patches: fixes applied without causal understanding
- search shape: systematic narrowing or random walk
Representative Trees
Reconstruct three to five full trees. For each:
- entry observation
- branches considered, in order
- dead ends and what closed them
- the update moment - what evidence finally moved the mental model
- distance between root cause and where the search started
Mental Model Timeline
Across one long debugging session, snapshot the working theory after each
major turn. The interesting signal is the jump size: incremental updates
versus wholesale replacement versus stubborn survival.
When It Backfires
- Hindsight bias. With today's knowledge every tree looks dumb somewhere.
Judge branch choices against available evidence at that moment.
- Counting the model's hypotheses as the subject's. In co-debugging,
attribute every branch (episode-method).
- Cherry-picking disaster sessions. Include routine debugs; they show the
default tree.
One-Line Memory
Draw the tree. Count the branches. Watch the updates.
