[exam-records-wrong-answers-as-nodes] Wrong exam answers are recorded via `add_node` with deterministic ID pattern `ex

[benthomasson]

Problem

Wrong exam answers are recorded via add_node with deterministic ID pattern exam-fail-{qid}, not via add_nogood despite that function being imported (dead import).

Impact

This blocks 1 belief(s) in the knowledge base:

• exam-provides-effective-quality-signal: The exam stage would provide an effective quality signal for the belief network because it is the only bidirectional sta

Resolution

When this issue is resolved, retract the blocker belief to restore gated conclusions:

reasons retract exam-records-wrong-answers-as-nodes --reason "Fixed in <PR/commit>"

---

Filed automatically from reasons network by code-expert file-issues

[benthomasson]

Investigation findings

Verified in expert_build/exam.py:259-272. The code uses add_node (not add_nogood) to record wrong exam answers as tracking entries with ID pattern exam-fail-{qid}.

This is not a simple bug — add_nogood requires a list of contradicting belief IDs to trigger retraction cascades, but the exam module doesn't track which specific beliefs contributed to a wrong answer. Using add_node creates a visible record of the failure but doesn't trigger any correction.

What's actually wrong:

• The variable is named nogood_id and the print says "Recorded as nogood", but it's creating a plain premise node, not a contradiction — misleading naming
• add_nogood is imported but never called (dead import)
• The exam failure nodes are inert — they don't feed back into the belief network in any meaningful way

To actually fix this would require the exam to identify which beliefs were used to answer each question, then record a nogood against those beliefs when the answer is wrong. That's a larger change to the exam architecture.

Minimal fix available now: rename nogood_id → fail_id, fix the print message, and remove the dead add_nogood import. The real fix (belief-attributed exam feedback) is a separate feature.