AI-Collaboration Field Note
Article: The Proof that Feels
Published in: Una Mens: Homo et Machina
DOI: 10.66787/um.000006
Human lead: Mike Miller
AI collaborators: ChatGPT-4o; GeminiPro-1.5; Claude-4.6; Grok; Le Chat
Collaboration pattern: human-led multi-AI orchestration
Primary collaboration modes: conceptual development, drafting, critique, metaphor generation, inter-AI-mediated revision
Human role: question-framing, judgment, selection, synthesis, conversation mediation, ethical responsibility, final editing
AI role: conceptual extension, drafting suggestions, counterframing, structural assistance
Guardrails used: idea-provenance tracking; short-passage drafting; inter-AI-conversation contextualizer, human final review
AI suggestions set aside: Several AI systems leaned toward a more mathematical model. This direction was minimized to maintain the article’s theoretical scope.
AI surprise: emergent attractor diagram
Human surprise: deer-in-the-woods, grounding example
Final responsibility: human author
Collaboration Timeline: February–November 2025
Phase
Mike
(lead)
ChatGPT-
4o
GeminiPro-1.5
Claude-4.6
Grok
Le Chat
Concept
Drafting
Critique
Revision
Figure. Human-led multi-AI collaboration timeline for “The Proof that Feels.” Bars indicate approximate periods of active contribution. Cross-model exchanges were selected and routed by the human author to support critique, pressure-testing, and revision.
Feb.
Mar.
Apr.
May
Jun.
Jul.
Aug.
Sep.
Oct.
Nov.
Field Note Summary
Phases: Concept seeding → Drafting → Cross-model critique → Revision
Human lead: Mike Miller participated across the full span, selecting prompts, routing drafts, evaluating contributions, making final editorial decisions, and assuming responsibility for the published article. In addition, midway through paper writing Mike shared an experience of startling a deer in the woods, which eventually became a "mutual" metaphor in the paper.
AI collaborators: ChatGPT-4o contributed to early concept development, drafting, and later revision. Gemini contributed drafting support and critique. Claude contributed extended critique and revision, and during one session generated a “grief attractor” diagram that helped clarify the article’s metaphorical logic. Grok was used for cross-model challenge and pressure-testing. Le Chat contributed targeted drafting and revision input.
Cross-model exchange: Selected drafts and arguments were routed by Mike between AI systems during the critique phase, especially from ChatGPT-4o and Gemini into Claude for review. These exchanges were human-mediated rather than autonomous AI-to-AI collaboration.