We've been having the wrong argument
Some people generate a story from a single prompt. Others spend months developing characters, directing scenes, rejecting drafts and refining the narrative. Both may describe their work as AI-assisted writing. Their creative contributions are not necessarily equivalent.
We need better questions. What did the writer create? What did AI contribute? Who made the consequential decisions? And is the story actually worth reading?
1. Understanding VET
Computer scientist Meredith Ringel Morris developed the VET framework to examine polarized claims about artificial intelligence.
- Valence: Is AI portrayed as beneficial or harmful?
- Effectiveness: How capable and reliable is AI?
- Trajectory: How quickly will AI capability, adoption and societal effects change?
Her paper identifies four recurring narratives: AI Hype, AI Doom, AI Denial and AI Normalcy. None should replace careful examination of actual evidence. VET is an analytical taxonomy, not an experiment establishing that this exercise improves critical thinking. The writing examples below are PennedCanon’s applications; they are not findings from Morris’s paper.
2. Applying VET to writing
"AI writing is all garbage."
That's a claim about effectiveness. But which workflow are we evaluating? One-shot generation, AI-assisted editing, or extensive human-directed collaboration? Those processes should not be treated as identical.
"AI will replace novelists."
That's a claim about capability and trajectory. It also assumes something about readers, publishers, markets and the value of individual human voices.
"Using AI means you're not creative."
This raises a separate question about human contribution. Story conception, characterization, original expression, selection, arrangement and substantive revision may all matter. But directing AI does not automatically establish authorship of everything the model generates.
3. Four questions worth separating
Literary merit: Is the finished story compelling?
Creative contribution: What did the human actually do?
Ethics and transparency: Was the process represented honestly?
Copyright: Which elements contain protectable human expression?
Answering one question does not settle the others.
4. What copyright guidance says
The U.S. Copyright Office's 2025 AI copyrightability report distinguishes human-authored expression from material generated by AI. Human-created expressive elements, creative modifications, and sufficiently original selection or arrangement may qualify for protection. Prompts alone generally do not establish authorship of the resulting AI-generated expression.
A provenance record can help document contributions. It cannot guarantee copyright protection or substitute for human authorship.
Source: U.S. Copyright Office — Copyright and Artificial Intelligence.
5. Provenance: showing the work behind the work
A useful record preserves original concepts, human-written passages, AI-generated alternatives, important editorial instructions, rejected directions, and substantive revisions.
It should not pretend that a percentage of typed words captures the entire creative process. A single editorial decision can fundamentally reshape a novel.
6. A real example: Sovereign Mercy
The development of Robert Wescott’s Sovereign Mercy shows why originating a premise, proposing a character, directing a revision and writing its prose deserve separate attribution.
The premise: an original author message
In the original chat, Story Origins — Original Premise and Early Twist Proposals, Wescott proposed a government-created AI whose purpose was to monitor, control and, if necessary, unplug other AIs. His message also supplied the wealthy benefactor who is secretly an AI and needs the government to build the more powerful intelligence. The responding assistant developed those ideas into further plot proposals. This directly supports crediting the foundational premise and benefactor twist to Wescott; it does not attribute every later development to him.
The school-bus opening and Sam Pike: retrospective attribution
The later chat Sam Pike Character Choice attributes the school-bus opening scenario and Samuel “Sam” Pike to ChatGPT, with subsequent collaborative development. The original premise message contains neither the bus nor Pike. The attribution is preserved in the project’s creative genealogy, but it is a retrospective account, not a recovered first-generation message naming the character. Whether Wescott’s own bus-driving background influenced that proposal remains undocumented. Later human direction does not retroactively turn an AI-originated proposal into a human-originated one.
Chapters 19–21: a documented revision account
The retained project record describes Wescott challenging an earlier outage sequence: why would Tessa pursue this particular resident, how did she reach the building, and what could the encounter establish? It records his willingness to discard the scene or the outage itself if local repairs could not make the investigation meaningful.
His governing requirement was that a pursuit produce consequential understanding, corroborate another line of inquiry or justify a larger question. He also required Chapters 19–21 to be developed together, with clean handoffs and distinct dramatic outcomes. The same record credits ChatGPT with proposing the replacement emergency involving overlapping transport reserves and with generating the joint drafts and revisions. Wescott supplied the critique, governing direction and selection; the record does not recast those decisions as his personal composition of all the resulting prose.
Evidence reviewed November 7, 2026: the original premise message and its assistant response; the retrospective Sam Pike Character Choice chat; and retained project entries “Robert Wescott — substantive originating direction, critique and selection” and “AI/ChatGPT — specific proposal and execution.” The chapter example is supported by those preserved project accounts. The earlier exchanges in Draft Chapter 19 were not retrievable during this review, so this page does not present their paraphrases as verified quotations or claim a fresh comparison of every original draft. These are the founder’s private development records, not independent validation.
The public founder case study documents another, separate example: Apple Books annotations connected to saved revisions, including clothing continuity across Chapters 1 and 2. It reports seven editorial comments recorded as addressed with revision links. That case study is not the source for the origin and chapter-development claims above.
These examples illustrate distinct human and AI contributions. They do not establish the copyright status of every passage or guarantee the completeness of the record.
7. Cognitive offloading deserves attention
Research raises legitimate concerns about dependence on AI. Gerlich (2025) reported associations involving AI use, cognitive offloading and critical-thinking measures. These observational relationships do not prove that AI causes long-term deterioration in every user. The study used surveys and interviews and acknowledges self-report and sampling limitations. A September 2025 correction replaced a duplicated table; the author stated that the conclusions were unchanged.
Kosmyna and colleagues (2025) studied essay-writing tasks under different assistance conditions. Their findings raise questions about engagement, recall and perceived ownership, but should not be generalized automatically to long-form fiction development. The revised preprint reports 54 participants in the first three sessions and 18 in a fourth. It is not a longitudinal test of novelists’ skills or a demonstration of permanent cognitive harm.
One practical safeguard is to define a scene's dramatic purpose yourself, evaluate generated alternatives critically, and continue practicing writing skills independently.
8. Who gets to tell a story?
Someone may have extraordinary experiences but lack the confidence or technical writing skills to construct a novel. AI can reduce barriers to organizing and expressing those experiences. That doesn't eliminate the need for craft. It creates an opportunity to develop it.
Are we protecting human storytelling, or sometimes protecting the barriers to participating in it?
9. Responsible collaboration
Start with creative intention. Exercise judgment. Document important decisions and human expression. Disclose meaningful assistance honestly when appropriate. Respect other creators and remain accountable for the finished work.
No process guarantees good fiction. But a documented process can be examined, criticized and improved.
10. Try the Claim Clinic
Rather than debating whether AI writing is inherently good or bad, examine a specific claim and the assumptions behind it.
Open the Claim Clinic exercise.
References and attribution
Meredith Ringel Morris (2026), VET: A Framework for Analyzing AI Discourse. Original publication, May 28, 2026; arXiv revision, June 9, 2026. The arXiv version identifies its license as CC BY-NC-ND 4.0. This page links to the paper and offers independent commentary; it does not reproduce its figures.
Gerlich (2025), AI Tools in Society; September 10, 2025 correction.
Kosmyna et al. (2025), Your Brain on ChatGPT, research preprint, revised December 31, 2025.
This guide independently applies VET to fiction writing. Morris does not endorse PennedCanon or its authorship and provenance recommendations.