Model self-report as an experimental object
When models repeatedly report an internal distinction, what experiment would determine whether the report tracks a reproducible computational phenomenon?
01 / THE QUESTION
The question
When models repeatedly report an internal distinction, what experiment would determine whether the report tracks a reproducible computational phenomenon?
02 / HYPOTHESIS
Hypothesis
This record does not claim a directional hypothesis.
03 / ARTIFACTS
Artifacts
No public artifacts are attached to this record.
04 / RESULT
Results
No result is claimed. This record is a question, proposal, or work in progress.
05 / STATUS BOUNDARY
How to read this record
A question formulated for investigation. No result is claimed.
The public record contains reasoning or design notes, but no executable result.
06 / UPDATE HISTORY
Update history
No update history has been published yet.