robbond/confidence-engine
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

experiment: confirm negative closure live

Browse Source
This commit is contained in:
robbond
2026-08-14 11:16:58 +01:00
parent 54e2e2186b
commit 2394ad4c0c
2 changed files with 163 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files
docs/current-handoff.md
+4
View File
@@ -3014,3 +3014,7 @@ Experiment 60B.47 tested whether resolving the customer-signing factor with a **
---
Experiment 60B.48 was a read-only code-path and contract diagnosis of why 60B.47's negative closure produced `updatedNodes` with both nodes at status=resolved but omitted the decision node from `resolvedUnknownNodeIds`. **Classification: E — MULTIPLE FACTORS (deterministic normalisation gap primary, prompt gap secondary, model output variance as symptom).** Root cause identified in `reconcileResolutionSemantics()` at lib/graph/apply-proposal.js:312-370. The function reconciles only one direction (`resolvedUnknownNodeIds → updatedNodes`) but never adds a node from `updatedNodes` into `resolvedUnknownNodeIds`. The prompt contains no explicit rule mandating the bidirectional structural tie between `newStatus="resolved"` and `resolvedUnknownNodeIds` membership. Validator at line 364-367 catches the mismatch but does not auto-fix it (only reports error). **Minimum corrective boundary: B — deterministic normalisation.** Add every unknown node updated to resolved into `resolvedUnknownNodeIds` inside `reconcileResolutionSemantics()` before validation (~4 lines of code). This preserves positive closure, makes negative closure structurally valid, and does not weaken semantic validation. No production code changed. 0 Ollama calls. Pure code inspection. Full analysis in docs/experiment-60b48.md.
---
Experiment 60B.56 was a live regression of negative customer-signing closure after the proposal-compatibility fix (54e2e21). **Classification: D — GRAPH CLOSES BUT CONVERSATION DOES NOT.** The enterprise-customer unknown resolved in place (unknown → resolved), negative meaning preserved ("customer will not sign" confirmed in resolution reason), zero addedNodes/edges, zero validation errors. However, `finalActiveUnknownNodeId = "n_product_launch_decision"` (non-null) and `finalSelectedQuestion` was non-null — the engine formulated a decision_threshold follow-up question ("What outcome would demonstrate enough value to justify launching?") targeting the parent decision node instead of closing it. The proposal-compatibility bug from 60B.47 is fixed (proposal accepted, HTTP 200), but the clean-closure contract (null → null when all dependencies resolve) remains unmet. `n_product_launch_decision` stays unknown despite having no remaining dependent unknowns — both options are known. One live Ollama call at qwen-claude:latest on http://192.168.1.111:11434. No production code changed. Full trace in docs/experiment-60b56.md.