Loading the investigation…
Completed investigation · explanation testing
Follow the investigation
Qualitative researchProposed an explanation
New case evidenceCompared competing explanations
Reviewed resultKept the explanation unconfirmed
What the new case established
This is a useful inconclusive result.
The evidence documented burdens, possible responses, and the final access policy. It did not reveal the decision process needed to say why that option was chosen.
One investigation, several owners
How the analysis moved from interviews to a new case
Each method kept control of its own judgment. The Workbench preserves the question, artifact identity, result, and handoff.
What changed
What the new sources added
What remains open
Why the explanation is still unconfirmed
Why the public causal conclusion was withheld
One chronology statement exceeded its evidence.
Independent challenge
A second model checked the bounded findings
This challenge is advisory. A researcher still decides whether any wording changes.
New connected execution · 23 September 2026
One agent carried the reviewed explanation to the publication guardrail
Next evidence
What would make the explanation testable
These are evidence requests, not another software step or an invitation to rerun the same sources.
Evidence packet
The different-case sources used
The original Open Science interviews generated the explanation. They were not counted as evidence in this case test.
Inspect artifact identity, digests, and method custody
Internal IDs and hashes are kept here so they do not replace the plain-language result. “Producer-retained reference” means the Workbench has an exact ID and digest but does not copy the original run bytes.
Open the matching JSON projection