Work

AI Consciousness Experiment

Year: 2026
Type: article
Abstract: This report analyzes a single transcript in which ChatGPT was asked 45 sequential questions about its identity, awareness, capacity for unpleasant states, information processing, continuity and independent reasoning. Across all seven phases the model held one consistent position: it claimed functional analogues of self-reference, self-monitoring, prioritization and integration, while stating that it could not establish subjective experience. It also said that its own testimony could not settle the question. We describe the recurring structure of its answers, summarize the conceptual distinctions it used, and assess what the transcript can and cannot support. We conclude that the transcript is a useful baseline for how this model talks about itself and is not evidence for or against machine consciousness. Several confounds limit interpretation, including leading question order, a repeated rhetorical template, unverified provenance and no controls.
Language: English
Subfield: Philosophy
Open Access status: green
OpenAlex ID: W7225988567
FWCI:
Cites:
Cited by:
Related to: 10