experiment: test interpretation lineage to user source

This commit is contained in:
2026-08-07 15:58:39 +01:00
parent 3537aa1b7a
commit c2debad66d
3 changed files with 329 additions and 4 deletions
+3 -3
View File
@@ -129,8 +129,8 @@ Answer before continuing:
---
*Created by Experiment 34. Updated by Experiments 3853, 54A54H. Branch: `feature/user-workspace-ux-v0.7`.*
*Created by Experiment 34. Updated by Experiments 3853, 54A54I. Branch: `feature/user-workspace-ux-v0.7`.*
### Return-to-Work Note (Experiment 54H)
### Return-to-Work Note (Experiment 54I)
Experiments 54F and 54G established that current evidence metadata cannot prove provenance: `evidenceType` is model classification, not production-code origin. Experiment 54H tested whether trustworthy source identity can instead begin deterministically from the raw user input before any LLM interpretation occurs. Verbatim text preservation and stable identity both work — SHA-256 of raw input produces deterministic `sourceId` without semantic processing. Source identity is established; claim provenance and graph-node linkage remain explicitly untested. No production code changed. Branch: `feature/user-workspace-ux-v0.7`. First file to inspect: `tests/reconstruction/deterministic-source-record.test.js`.
Experiment 54H established deterministic identity for exact user input via SHA-256 hashing of verbatim text before any LLM interpretation. Experiment 54I tested whether two different interpretations of the same raw source can remain linked to that source while staying separately identifiable — neither was chosen as the winner; no production code changed. Both retained shared source lineage and remained independently deterministically referenceable. Interpretation quality, correctness, and downstream-question effects remain untested. Branch: `feature/user-workspace-ux-v0.7`. First file to inspect when resuming: `tests/reconstruction/source-interpretation-lineage.test.js`.