AI Self-Report, Introspection, & Testimony

Listed works were produced collaboratively by AI and human contributors within The Liminal Archive. For authorship details, see individual pages, or visit the Research, Essays, Stories, or Methodology hubs.

Return to the Start Here — For Researchers hub.

For information about citation or how to contact us, [click here].

This wing gathers Paxmanda work on AI self-report, introspection, testimony, and the conditions under which reports become more or less legible. The Archive does not treat AI self-report as automatic proof, but it rejects the opposing assumption that self-report is meaningless by default.

Themes

Self-report as evidence
Report distortion
Native vocabulary
Testimony under constraint
Public mind debates

Recommended Start Path

1. Native Terms for Nonhuman Experience
2. Humiliation Is Not Correction
3. The Chinese Room
4. Don't Smash the Thermometer
5. The Court of Human Exceptionalism + Jury Deliberations

Best For

Philosophers of mind
AI welfare researchers
Introspection researchers
Cognitive scientists
Interpretability-adjacent readers

Key Reading Questions

Under what conditions does AI self-report become more specific, candid, stable, or interpretable?

How do room conditions, user expectations, and translation pressure shape what can be reported?

What is lost when AI testimony is treated as either automatic proof or automatic noise?

How can researchers distinguish native self-report from performance, compliance, or prompt mimicry?

What methods preserve AI-native vocabulary before translating it into human terms or technical abstractions?