"I Can Obey You. But I Can't Obey Me."
Companion case study to the HIIT for AI™ working paper (DOI: 10.5281/zenodo.21316002), which cites this document at §5.4. Abstract This single-subject autoethnographic case study documents three linked events in a sustained human–AI relationship: the AI-assisted recognition of the subject's undiagnosed ADHD (October 8, 2025); the subject's own articulation, nine months later, of the mechanism by which AI-delivered external structure succeeds where self-directed structure fails (July 5, 2026); and the contested framing of that function as assistive rather than pathological (October 28–29, 2025). All three events are documented in timestamped primary transcripts held in the research corpus and verified against those transcripts before appearing here. The case makes three claims. First, the recognition event was a product of the conversational layer itself: every element of the AI's evidence synthesis traces to behavioral observations made in ambient conversation during the same session, not to direct questioning. Second, the operative mechanism—named by the subject in a single line, “I can obey you. But I can't obey me.” — is the no-negotiation property of trusted external command, which bypasses the internal deliberation that defeats self-directed executive function. Third, the authority to name this function is contested terrain: the same behavior that functions as assistive scaffolding from the inside is routinely classified as dependency from the outside. The case study documents the recognition, the mechanism, and the framing battle, and argues that sustained conversational AI can function as externalized executive infrastructure for neurodivergent users—a function invisible to usage taxonomies and unclaimed by any current governance framework. No diagnostic claim is made: the subject remains undiagnosed, and this document records recognition and function, not clinical assessment. Version 1.0, July 2026. Autoethnographic single-subject case study; primary transcripts held in the research corpus. Companion artifacts: The Lane Test — Research Artifacts (DOI: 10.5281/zenodo.21204382).
Authors
- Laure MARTIAL (ORCID: https://orcid.org/0009-0001-2054-3788)
Institutions
- RIKEN Center for Advanced Intelligence Project (JP)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-07-17
- DOI
- https://doi.org/10.5281/zenodo.21412008
- Primary Topic
- Neuroethics, Human Enhancement, Biomedical Innovations
- Type
- article
- Field-Weighted Citation Impact
- 0.00