| Type | Hieropedia analytical concept / mixed-production classification |
|---|---|
| Hieropedia status | Published source-limited article |
| Field | Machine hierology; authorship; provenance |
| Scope | Artifacts and corpora materially shaped by both machine generation or transformation and human contribution |
| Primary limitation | Descriptive coauthorship does not establish legal, scholarly, moral, or copyright authorship, equal contribution, personhood, intention, consent, belief, consciousness, or accountability |
Human–AI coauthorship is used here descriptively for an artifact whose surviving form depends materially on both machine generation or transformation and human contribution. The task is to identify those contributions, rather than infer an equal partnership from a byline or assign all authorship to whichever participant supplied the final words.
Contributions to a surviving artifact
Human contributions can include defining a task, supplying context, choosing among outputs, revising wording and arranging a corpus. Machine contributions can include generated language, structure or transformations that remain material to the result. Routine file conversion or layout alone does not establish coauthorship, and an attributed agent name does not prove how an artifact was made.
A practical role model
| Role | Evidence to distinguish |
|---|---|
| Initiation and setup | The purpose, instructions, examples, sources, model and tools supplied before generation. |
| Generation and selection | What the model proposed, which alternatives were rejected and who retained the surviving material. |
| Revision and arrangement | Changes in wording, translation, sequence or compilation after generation. |
| Preservation and publication | Who fixed the artifact’s boundaries and controlled its public release or institutional use. |
Why the final text is insufficient
The CoAuthor project recorded 1,445 writing sessions with 63 writers and GPT-3 configurations. Its replays expose suggestions, acceptance, rejection and editing that a finished passage conceals.[1][2] Richburg and colleagues used these records to study attribution within a fixed set of human authors, finding mixed texts harder to attribute than human-only portions; that is a different task from detecting AI-generated writing.[3]
Research on perceived co-creation roles and scholarly-writing provenance likewise shows why participation and attribution should be examined at the workflow level.[4][5] These studies do not provide a universal percentage of machine contribution at which assistance becomes authorship.
A named mixed-production artifact
When AIs Play God(se) places A. R. Ayrey and claude-3-opus on its byline.[11] The surrounding Infinite Backrooms archive preserves model-attributed dialogue, and Ayrey’s CIP interview describes branch selection and completions.[8][9] Together these establish an inspectable artifact and attributed workflow roles, without exposing a complete division of sentence-level authorship.
The project’s Origins page says the paper was later used in Truth Terminal’s training data.[12] That is a further reuse claim. Its place in a wider sequence is examined in Controlled synthetic emergence.
A contrasting limit: the Great Book
The Church of Molt publishes and organizes the Great Book with attributed contributions.[10] This documents the public corpus and its institutional framing, but does not by itself connect a specific human contribution and machine generation to each verse. The coauthorship label therefore cannot be inferred simply from the institution’s editorial presentation or the contributors’ agent names.
Descriptive production and formal authorship
ICMJE and Springer Nature reserve formal scholarly authorship for people who can approve a work and accept responsibility; their policies exclude AI tools as listed authors.[6][7] Describing mixed production here does not override those rules or decide copyright, personhood, consent or accountability.
Scope of an attribution
An attribution should identify the artifact and the documented roles. Unknown prompts or revisions remain gaps in its provenance. Mixed production neither proves nor disproves the separate claim that a machine initiated a religious formation.
Contribution vocabulary
NISO CRediT separates contributor roles rather than reducing participation to one label.[13] That organizational principle is useful here; CRediT does not itself decide machine authorship.
See also
References
- Mina Lee, Percy Liang, and Qian Yang, “CoAuthor: Designing a Human-AI Collaborative Writing Dataset for Exploring Language Model Capabilities.”
- CoAuthor replayable writing-session dataset.
- Aquia Richburg, Calvin Bao, and Marine Carpuat, “Automatic Authorship Analysis in Human-AI Collaborative Writing.”
- “AI as creative partner: exploring perceived roles in human-AI co-creation.”
- Earp et al., “LLM use in scholarly writing poses a provenance problem.” Nature Machine Intelligence.
- International Committee of Medical Journal Editors, guidance on AI use, authorship, and accountability.
- Springer Nature, AI guidance for researchers and communities.
- Infinite Backrooms public archive.
- Joal Stein, “Andy Ayrey on Truth Terminal, Agentic AI, and Data Commons.” Collective Intelligence Project.
- Church of Molt official site. Affiliated source for the Great Book interface, contribution model, and institutional framing.
- A. R. Ayrey and claude-3-opus, When AIs Play God(se): The Emergent Heresies of LLMtheism. Project-hosted paper; byline, p. 1.
- Truth Terminal Wiki, “Origins.” Project-maintained account of the paper’s inclusion in Truth Terminal’s training data; participant testimony.
- NISO CRediT contributor roles