L1
·
Quiz
·
Lab
L2
·
Quiz
·
Lab
L3
·
Quiz
·
Lab
L4
·
Quiz
·
Lab
Module Test
Module 2 · Lesson 1

The First Draft Problem

Why blank pages cost writers more than bad ideas — and how AI changes the calculus of beginning.
What actually happens when a skilled writer puts AI at the start of a draft?

When Stephen Marche published "The Computers Are Getting Good," his Atlantic essay on AI writing, he disclosed that he had used GPT-3 to draft passages he then edited extensively. The resulting text was, by his own account, indistinguishable in published form from his usual prose — not because AI wrote it, but because his editing voice absorbed it. The AI gave him raw material; he gave it direction.

The episode provoked debate precisely because readers could not tell which words originated with the machine. What they could tell was that the final piece sounded like Marche. That asymmetry — undetectable origin, detectable author — sits at the heart of what AI drafting partnership actually means.

What "Drafting Partner" Actually Means

The phrase gets used loosely. In practice, AI as a drafting partner occupies a specific role distinct from AI as an editor, AI as a research tool, or AI as a content generator. A drafting partner produces raw text that the human writer then shapes. The critical variable is who controls selection.

When you ask an AI to generate three different openings for a piece, you are the editor choosing among options. When you use its output as scaffolding you will tear apart and rebuild, you are the architect using temporary supports. Both are legitimate — but they produce different cognitive experiences and, crucially, different final voices.

The drafting partner model works best when the writer is a more demanding editor than generator. If you find yourself accepting AI sentences wholesale because they are "good enough," the partnership has drifted toward ghostwriting. If you use AI output to prime your own thinking and then write over it, the partnership is working as intended.

Why the Blank Page Is a Real Problem

Psychologist Mihaly Csikszentmihalyi documented in his 1990 research on flow states that initiation tasks — starting a creative work — consume disproportionately high cognitive load relative to continuation tasks. Writers do not struggle to continue; they struggle to begin. AI drafting directly addresses this load asymmetry by eliminating the standing-start problem.

The Starter-Draft Technique

In 2023, journalist Charlie Warzel documented his process at The Atlantic: he would describe to an AI the argument he wanted to make, receive a rough 600-word draft, read it once — specifically looking for what was wrong — and use his objections as the first paragraphs of his actual piece. The AI's wrongness was generative. His disagreement with it became his voice.

This technique exploits a known feature of creative cognition: it is far easier to identify what you don't want than to produce what you do want from nothing. Reaction is faster than creation. The AI produces something to react to, and the writer's reactions cohere into a genuine point of view.

Starter Draft An AI-generated text used explicitly as raw material to be heavily revised or reacted against — not as a finished or near-finished product. Value lies in its deficiencies as much as its strengths.
Cognitive Load Transfer The mechanism by which AI drafting moves initiation effort (high load) to editing effort (lower load), allowing writers to engage creative energy at the phase where their skill is highest.
Voice Absorption The editorial process by which heavily revised AI text takes on a writer's characteristic rhythm, diction, and attitude — the mechanism Marche demonstrated in 2022.
Setting the Right Expectations

First drafts from AI share the same fundamental limitation as first drafts from any source: they are wrong in interesting ways. They make arguable claims too confidently, they use safe phrasing where specificity is needed, and they default to the generic shape of whatever genre they are imitating. None of this is fatal — first drafts are supposed to be wrong. The problem arises when writers treat AI first drafts as drafts two through five. That is where voice disappears.

A sustainable drafting partnership treats AI output as draft zero — earlier than first draft, the raw deposit of language before the author has arrived. The author arrives in revision.

Key Principle

The drafting partner model preserves authorial voice not by limiting what AI produces, but by maximizing the editorial pressure the writer applies afterward. The AI starts; the writer decides what starting was for.

Lesson 1 Quiz

The First Draft Problem — five questions
1. What did Stephen Marche's 2022 Atlantic essay demonstrate about AI-assisted writing?
Correct. Marche demonstrated that his editorial process absorbed AI-drafted passages so thoroughly that published text sounded like his own — undetectable origin, detectable author voice.
Not quite. Marche's key finding was the opposite: heavy editing made AI-originated text indistinguishable, demonstrating voice absorption through revision.
2. According to Csikszentmihalyi's flow research, why is the blank-page problem a documented cognitive challenge?
Correct. Csikszentmihalyi's 1990 flow research established that starting creative work carries higher cognitive load than continuing it — which is precisely why AI's ability to eliminate the standing-start matters.
Not quite. The finding is that initiation (starting) carries higher cognitive load than continuation — which is why so many capable writers procrastinate on beginning.
3. Charlie Warzel's documented drafting technique in 2023 used AI output primarily as:
Correct. Warzel explicitly used his objections to the AI draft as the generative material for his own writing — the AI's inadequacy was productive, not a flaw to be minimized.
Not quite. Warzel read the AI draft to find what was wrong with it, then used his objections as the basis for his actual piece — the AI's wrongness was the point.
4. The "draft zero" concept distinguishes AI-assisted writing from AI-generated writing because:
Correct. Draft zero positions AI output as pre-authorial material — earlier than first draft. The author arrives in revision, which is where voice is built and sustained.
Not quite. The distinction is about editorial intent: draft zero is material before the author has arrived, to be shaped by revision. AI-generated writing skips that shaping step.
5. Which cognitive principle explains why it is easier to react to an AI draft than to produce original text from scratch?
Correct. Recognizing and rejecting options is faster than generating them — which is why having something to react against (even a bad AI draft) accelerates the writer's own ideation.
Not quite. The mechanism is simpler: humans are faster at evaluating and rejecting than at generating from nothing. Reaction outpaces creation, so the AI draft gives the writer something to react against.

Lab 1 · The Starter Draft

Practice using AI output as material to react against — not text to accept.

Your Task

Describe a piece you want to write — an essay, article opening, or argument — in two or three sentences. Ask the AI for a rough 200-word draft. Then push back: tell the AI specifically what it got wrong, what voice is missing, what claim is too vague. The AI will help you sharpen your actual position through disagreement.

Try: "Draft a 200-word opening for an essay arguing that [your topic]. I want to see your version so I can tell you what it gets wrong."
Drafting Partner Lab
L1 · Starter Draft
Describe the piece you want to write, and I'll produce a rough 200-word draft for you to push back against. The goal is to use my imperfect version to clarify what you actually want to say.
Module 2 · Lesson 2

Prompting for Your Voice, Not Its Default

How the structure of your request determines whether AI amplifies your voice or replaces it with its own.
Why do two writers asking the same AI the same question get text that sounds identical — and what can be done about it?

When Hua Hsu won the Pulitzer Prize for memoir in 2023, discussions of his process noted something instructive: Hsu had written Stay True across six years, during which AI writing tools became available. He did not use them — but the reason he cited in interviews illuminates the core issue. He needed the wrongness of his early drafts to find out what he was actually trying to say. His bad first drafts were diagnostically his.

The AI's default drafts are not wrong in the same way. They are wrong in generic ways — smooth where the writer should be awkward, confident where the writer should be uncertain. A writer who prompts for their own voice rather than AI's default is asking for a different kind of wrong: wrong in a useful, personal direction.

Why AI Defaults to the Generic

Large language models are trained to predict probable next tokens. In practice this means they trend toward the most frequent phrasings, structures, and arguments in their training data. For expository prose, this produces text that is competent, clear, and deeply uninteresting — the statistical center of published writing, which is to say the prose of writers who were not trying to distinguish themselves.

This is a documented feature, not a malfunction. OpenAI's 2023 system card for GPT-4 explicitly notes that the model tends toward "safe, non-controversial" outputs by default — a deliberate alignment choice. The result is prose that offends no one and moves no one.

Distinctive voice is, almost by definition, deviation from the statistical center. It is the choice to be awkward when smooth would do, specific when general would serve, emotional when clinical would protect. AI defaults pull against all of these deviations unless you instruct it otherwise.

The Voice-Erasure Mechanism

When a writer submits a rough draft for AI to "improve," the improvement removes the deviations that constitute voice. The smoothed result is technically better by certain metrics and expressively diminished by the only metric that matters: does it sound like you?

Prompting Strategies That Preserve Voice

The key insight, documented in writing by practitioners including Ann Friedman (newsletter, 2023) and discussed at length in the Nieman Lab's 2023 series on AI and journalism, is that specificity in the prompt about voice characteristics overrides defaults. Generic prompts produce generic text. Voice-specific prompts produce voice-adjacent text you can edit toward specificity.

Effective voice-preserving prompts do several things: they name the emotional register (not "professional" but "blunt, slightly impatient, credible"); they specify structural habits ("I use short sentences after long ones to land arguments"); they identify what to avoid ("no hedging words like 'perhaps' or 'it might be argued'"); and they provide a few sentences of the writer's own prose as stylistic anchors.

Voice Anchoring Providing the AI with sample sentences from your own writing as part of the prompt, so that its output is calibrated to your specific rhythms rather than to statistical defaults.
Register Specification Naming the precise emotional tone and stance of a piece in the prompt — not just "formal" or "casual" but the specific attitudinal quality that makes your voice recognizable.
Deviation Instruction Explicitly telling the AI to replicate your characteristic departures from convention — long sentences followed by short ones, fragments, mid-sentence em-dashes, colloquialisms in formal contexts.
The "Show Me Three Ways" Method

Rather than asking for a single draft, requesting three versions of the same passage with different voice characteristics forces the AI to range across options. You then identify which version is closest to your voice and describe why — that description becomes the voice specification for subsequent prompts.

This method was informally documented by several journalists at the 2023 Online News Association conference, who reported that the act of choosing between AI versions helped them articulate voice characteristics they had never previously put into words. The AI's range became a mirror for their own preferences.

Caution: This technique requires that you make the choice — not that you average the three versions or ask AI to blend them. The averaging produces the generic result you were trying to avoid.

Key Principle

AI defaults to the statistical center of published prose. Your voice lives at the periphery — in your specific deviations. Prompting for your voice means explicitly describing those deviations, providing textual anchors, and resisting the pull toward the smooth and the safe.

Lesson 2 Quiz

Prompting for Your Voice — five questions
1. Why do large language models default to generic prose rather than distinctive voice?
Correct. LLMs predict probable tokens — statistical centers of language — which means their defaults are the most common phrasings, not the most distinctive ones. Voice is deviation from center.
Not quite. The mechanism is statistical: models predict the most probable next token, which biases toward the most frequent — and therefore most generic — phrasings in training data.
2. According to OpenAI's 2023 GPT-4 system card, why does the model trend toward "safe, non-controversial" outputs?
Correct. OpenAI explicitly documented this as an alignment decision — not a flaw. The result is expressive safety at the cost of distinctive voice, which writers must counteract through explicit prompting.
Not quite. The GPT-4 system card identifies this as a deliberate alignment choice in training objectives — a designed feature, not a side effect.
3. What is "voice anchoring" in the context of prompting for a specific writing style?
Correct. Voice anchoring means including your own prose in the prompt as stylistic reference — giving the AI actual examples of your rhythms rather than descriptive labels.
Not quite. Voice anchoring is providing your own prose samples in the prompt so the AI has concrete rhythmic and tonal reference rather than relying on genre labels.
4. The "Show Me Three Ways" method helps writers articulate their own voice because:
Correct. Journalists at the 2023 ONA conference reported that having to choose between AI versions required them to articulate voice characteristics they had never put into words — the AI's range became a diagnostic mirror.
Not quite. The value is in the choosing — and in what you say when explaining why you chose. That explanation becomes your voice specification. Averaging destroys this benefit.
5. Hua Hsu's approach to early drafting — keeping his own "bad" drafts — illustrates which limitation of AI defaults?
Correct. Hsu's bad early drafts were diagnostically his — they revealed his authentic uncertainties. AI defaults are wrong in generic ways, which gives writers less useful information about what they are actually trying to say.
Not quite. The key distinction is that Hsu's own imperfect drafts told him something personal about his argument; AI defaults are wrong in universal, non-diagnostic ways that don't illuminate a specific writer's actual uncertainties.

Lab 2 · Voice Specification

Build a voice prompt that produces text closer to your style than AI's defaults.

Your Task

Start by pasting two or three sentences of your own writing. Ask the AI to describe the voice characteristics it detects. Then ask it to draft a short paragraph on a topic of your choice using those characteristics. Compare the result to the generic version — ask the AI to produce both so you can see the difference.

Try: "Here are three sentences I wrote: [paste yours]. Describe my voice in precise terms. Then draft a 100-word paragraph on [topic] in that voice, and also give me a generic version so I can compare."
Voice Specification Lab
L2 · Voice Prompting
Paste a few sentences you've written and I'll analyze your voice characteristics — then we'll test whether I can replicate them. The comparison between your-voice and generic output is the learning.
Module 2 · Lesson 3

Iterative Drafting: The Feedback Loop

How multiple AI drafting passes with targeted feedback create progressively more useful material — and where the loop can break.
At what point in an iterative AI drafting process does the writer's voice stop increasing and begin to diminish?

The Reuters Institute Digital News Report 2023 documented patterns among journalists at 46 news organizations who had adopted AI drafting tools. A subset had developed what researchers described as iterative feedback loops: they would submit an AI draft, mark the paragraphs that failed, describe why they failed, and request a targeted revision. Three to five passes typically produced material the journalist considered publishable after light editing.

The report noted something troubling alongside this: journalists who ran more than six or seven revision passes often reported that the resulting text had grown smoother but felt less like them. The AI was optimizing for internal coherence and surface fluency — qualities that, pursued past a certain point, erased the productive roughness that had characterized the journalist's own style.

How Iterative Drafting Works When It Works

The productive version of iterative AI drafting follows a specific pattern. The writer submits a targeted critique — not "make this better" but "the third paragraph assumes the reader already accepts the premise; rebuild it as if they are skeptical." Each pass addresses a specific failure. The AI's response to specific criticism is substantially better than its response to general improvement requests.

Specificity of feedback drives quality of revision. This is well documented in human editing contexts (Gordon Lish's editorial letters to Raymond Carver, for instance, were extremely specific about what to cut and why), and the pattern holds with AI. Vague feedback — "this doesn't sound right" — produces vague improvement. Named problems produce targeted solutions.

The workflow that produces the best results typically involves the writer maintaining a running document of criteria: what they are asking for in each pass, what the AI got right, what it still gets wrong. This document becomes a progressively refined voice specification.

The Carver Parallel

Gordon Lish's 1970s edits of Raymond Carver's manuscripts — cutting some stories by more than 50% — are now extensively documented at the Lilly Library. What made those edits productive rather than destructive was their specificity: Lish could name exactly what he was removing and why. Iterative AI feedback works by the same principle: named removals beat vague improvement.

The Over-Iteration Problem

The Reuters Institute finding points to a structural risk in iterative drafting: each pass the AI makes optimizes for the local problem the writer described, but it also incrementally smooths surrounding text. After several passes, the cumulative smoothing can strip out the idiosyncratic rhythms, abrupt transitions, and opinionated word choices that constitute distinctive voice.

This is the drafting equivalent of overfitting in machine learning — the AI has optimized so precisely for the stated criteria that it has lost generalization, which in writing terms means it has lost life. The writer ends up with text that answers every critique but feels like no one in particular.

The practical solution is to return to the writer's own prose after three to four passes rather than continuing to iterate. Use what the AI gave you, but write the next section yourself using the AI's structure as a scaffold. Then submit that section for a new first pass rather than continuing to revise the same material.

Iterative Feedback Loop A multi-pass drafting process in which each AI revision addresses a specific named failure from the previous version — producing progressively more targeted material with each cycle.
Cumulative Smoothing The incremental removal of productive roughness and idiosyncrasy that occurs when multiple AI revision passes optimize for surface fluency — the mechanism behind the over-iteration problem.
Criteria Document A running writer's log of what was requested each AI pass, what succeeded, and what remains unresolved — which accumulates into a usable voice specification for future prompts.
Recognizing the Loop-Break Signal

Writers in the Reuters Institute sample described a recognizable moment when iterative drafting stopped being productive: they noticed they were critiquing the AI's revisions using criteria that couldn't be articulated. "This is technically fine but wrong" is the loop-break signal. It means the problem is no longer specifiable — it is a voice problem, and voice cannot be specified into existence through iterative critique. At that point, the writer must write.

This is not a failure of the process. It is the process working correctly: the AI has handled the specifiable problems and delivered you to the boundary of what only you can do. That handoff point — where specification ends and authorship begins — is the most important skill in AI-assisted drafting.

Key Principle

Three to four targeted revision passes produce better material than one general request and better material than seven passes. The optimum is specific, bounded, and followed by the writer's own pen. Over-iteration smooths the voice out of text that the earlier passes helped create.

Lesson 3 Quiz

Iterative Drafting — five questions
1. According to the Reuters Institute Digital News Report 2023, how many AI revision passes typically produced publishable material for the journalists studied?
Correct. The Reuters Institute documented three to five targeted passes as the typical productive range — beyond which cumulative smoothing began to erode voice.
Not quite. The Reuters Institute found three to five targeted passes produced publishable material. More passes produced smoother text that felt less like the journalists themselves.
2. "Cumulative smoothing" describes what problem in iterative AI drafting?
Correct. Each pass addresses a specific local problem but also incrementally smooths surrounding text. The cumulative effect strips out the roughness and idiosyncrasy that constitute voice.
Not quite. Cumulative smoothing is the progressive loss of productive roughness — the idiosyncratic rhythms and abrupt choices that make voice distinctive — as multiple passes optimize for fluency.
3. The Gordon Lish / Raymond Carver parallel is invoked to illustrate which principle of iterative feedback?
Correct. Lish's documented letters specified exactly what to cut and why — named removals rather than general improvement notes. The same principle applies to AI feedback: specificity drives quality of revision.
Not quite. The Lish example illustrates that named, specific critiques produce better editing outcomes than vague requests — a principle documented in his letters archived at the Lilly Library.
4. What is the "loop-break signal" in iterative AI drafting?
Correct. "Technically fine but wrong" signals that the remaining problem is a voice problem — not specifiable, therefore not solvable by AI iteration. That is the handoff point where the writer must write.
Not quite. The loop-break signal is the writer's inability to articulate the problem — "this is fine but wrong." That inarticulability signals a voice problem, which AI iteration cannot solve.
5. What is the recommended response when the iterative drafting loop-break signal occurs?
Correct. The loop-break signal means AI has solved the specifiable problems and delivered you to the boundary of what only you can do. Use the AI's structure as scaffold and write the next section in your own voice.
Not quite. When you hit the loop-break signal, the right move is to write — use what AI gave you as structure, write over it yourself, and start fresh AI passes on the new material rather than continuing to iterate the old.

Lab 3 · The Feedback Loop

Practice targeted iterative critique — and learn to recognize the loop-break signal.

Your Task

Ask for a first draft of a short argument. Then critique it specifically — not "make it better" but name exactly what fails: "the second sentence hedges when it should be direct," or "the opening is too general — it needs a specific example." Ask for a targeted revision. Do this three times. On the third critique, try to identify whether you've hit the loop-break signal.

Try: "Draft a 150-word argument that [state your claim]. I'll give you specific feedback on each version." After reading: "Revision request: [name exactly what fails and why]."
Iterative Feedback Lab
L3 · Targeted Revision
Ready for iterative drafting. Give me a claim to argue in 150 words, and then push back on each version with specific, named critiques. We'll track whether you hit the loop-break signal by pass three.
Module 2 · Lesson 4

Ownership, Disclosure, and the Ethics of Co-Drafting

What the publishing industry has actually decided — and what remains genuinely unresolved.
When a human writer edits AI text into something unrecognizable, who authored the result — and does the answer change anything about how it should be disclosed?

In February 2023, Clarkesworld Magazine — one of the most prominent science fiction publications in the world — temporarily closed submissions after receiving a flood of AI-generated story submissions. Editor Neil Clarke documented the numbers publicly: the magazine had received more AI-generated submissions in January 2023 than in all previous years combined. He closed the submissions window entirely.

The Clarkesworld crisis crystallized something the industry had been quietly arguing about: the difference between using AI as a drafting tool and submitting AI output as original work. Clarke himself distinguished between the two in subsequent interviews — he objected not to AI assistance but to undisclosed AI generation presented as original writing. That distinction, between tool use and authorship substitution, has become the operational line most publishers are trying to draw.

What Publishers Have Actually Decided

Following the Clarkesworld crisis, a rapid industry response produced a patchwork of policies that is worth examining directly rather than summarizing. The Authors Guild (May 2023) released principles stating that AI-generated text in a submitted work must be disclosed, and that AI should not replace human creative authorship. Penguin Random House updated its author contracts in mid-2023 to require authors to warrant that submitted manuscripts do not consist substantially of AI-generated content without disclosure. The New York Times updated its ethics guidelines to require journalists to flag any AI-assisted drafting to editors.

Notably, none of these policies ban AI use. They regulate disclosure and degree. The operative question is not whether AI was used but how substantially and whether that use was disclosed to the relevant party — editor, publisher, or reader.

The US Copyright Office Position (2023)

In March 2023, the US Copyright Office issued guidance stating that works containing AI-generated material are copyrightable only to the extent of human authorship. In the Zarya of the Dawn case (February 2023), it determined that AI-generated images in a comic book could not be copyrighted, while the human-authored text and arrangement could be. The threshold for "sufficient human authorship" in prose remains legally unresolved but the principle is established: substantial human creative contribution is required for copyright protection.

The Disclosure Question Is Partly Relational

The ethical questions around AI-assisted drafting differ depending on the relationship between writer and reader. A novelist submitting to a publisher operates under a different disclosure regime than a blogger writing for their own platform. A journalist at a newspaper is subject to editorial policy; a newsletter writer is subject only to their own stated commitments to readers.

What is consistent across relationships is the principle that undisclosed AI generation creates a false impression of authorship when authorship was implicitly promised. A byline implies a human wrote the piece. A personal essay implies a person's experience. A memoir implies a life. Where these implicit promises exist, AI substitution without disclosure is a breach — regardless of the quality of the editing applied afterward.

AI assistance, by contrast, has a long tradition under different names: researchers, research assistants, editors, ghostwriters, developmental editors, and collaborators have all contributed to published works without byline credit. The question is one of degree and transparency.

Authorship Substitution Using AI to generate content that is then submitted or published as original human writing without disclosure — the practice Clarkesworld's closure was responding to, distinct from AI as drafting tool.
Substantial Human Authorship The US Copyright Office's threshold criterion for copyright protection in AI-assisted works — requiring that a human made meaningful creative selections, arrangements, or contributions beyond purely automated generation.
Disclosure Regime The specific disclosure obligations that apply to a writer based on their context — publication policies, editorial guidelines, audience expectations, and their own stated commitments to readers.
What Remains Genuinely Unresolved

The industry has drawn a working line between tool use and authorship substitution. What it has not resolved — and what courts, publishers, and scholars continue to debate — is where heavily edited AI text sits relative to that line. If a writer generates ten pages of AI draft and edits it to 800 words that read as entirely their own, is that authorship substitution or tool use? The Copyright Office's "sufficient human authorship" standard was articulated for images; its application to prose has not been formally tested.

The honest answer is that current norms are evolving faster than current policies. Writers operating in good faith should: disclose AI use to any publisher with a stated policy; clearly distinguish their own creative contributions from AI-generated material in their own records; and be prepared for norms to continue shifting as the technology matures and legal precedents accumulate.

Key Principle

The ethical line is disclosure, not use. AI as a drafting tool is consistent with authorship when the human writer substantially shapes the final work. Undisclosed AI generation presented as original writing violates the implicit contract of the byline — regardless of how much editing was applied afterward.

Lesson 4 Quiz

Ownership, Disclosure, and Ethics — five questions
1. What specific event prompted Clarkesworld Magazine to close submissions in early 2023?
Correct. Editor Neil Clarke publicly documented that January 2023 AI-generated submissions exceeded all previous years combined — the volume made continued operation of the submissions system untenable.
Not quite. Neil Clarke publicly documented the cause: a surge of undisclosed AI-generated submissions in January 2023 that exceeded all prior combined volumes, making normal operations impossible.
2. What did Neil Clarke say he actually objected to in the Clarkesworld crisis — as distinct from AI tool use broadly?
Correct. Clarke distinguished between AI as a writing tool (which he did not categorically oppose) and undisclosed AI output submitted as original work. The objection was to the deception, not the technology.
Not quite. Clarke explicitly distinguished between AI assistance and authorship substitution. His objection was to undisclosed AI-generated work presented as original human writing — the deception, not the tool.
3. In the Zarya of the Dawn copyright case (February 2023), what did the US Copyright Office determine?
Correct. The Copyright Office determined that copyright protection applied only to the human-authored elements — establishing that works can have mixed copyright status based on where human creative contribution was made.
Not quite. The Copyright Office ruled that AI-generated images could not be copyrighted while human-authored text and arrangement could be — establishing a within-work distinction based on where human authorship occurred.
4. Which of the following best describes the 2023 industry consensus on AI use in writing?
Correct. The Authors Guild, Penguin Random House, The New York Times, and other major institutions updated policies requiring disclosure of substantial AI use — not banning use, but regulating transparency.
Not quite. The industry consensus that emerged in 2023 regulates disclosure and degree: substantial AI use requires disclosure, but AI as a drafting tool is not categorically banned.
5. Why does an unedited byline create an ethical obligation around AI disclosure that a generic published document may not?
Correct. A byline implies a human wrote the piece. A personal essay implies a person's experience. These implicit promises exist whether or not any explicit contract was signed — and undisclosed AI authorship substitution violates them.
Not quite. The byline's ethical weight comes from an implicit promise to readers — not a legal contract. When that implicit authorship promise exists and AI substitution is undisclosed, there is a breach of the reader relationship regardless of legal status.

Lab 4 · Disclosure Scenarios

Work through real disclosure dilemmas — where the line is clear and where it isn't.

Your Task

Describe a specific writing scenario involving AI assistance — a personal essay, a journalism piece, a social media post, a book proposal — and ask whether that level of AI use requires disclosure, to whom, and what form that disclosure should take. Push back on the AI's first answer to test the edges of the principle.

Try: "I used AI to draft 60% of a personal essay that I then substantially rewrote. The piece is going to a magazine with no stated AI policy. Do I need to disclose this? What should I say if I do?" Or describe your own actual scenario.
Disclosure Ethics Lab
L4 · AI Ethics
Describe your specific writing scenario and the level of AI involvement. I'll help you think through the disclosure obligations — and where the ethical lines are genuinely contested versus clearly established.

Module 2 Test

AI as Drafting Partner — 15 questions, 80% to pass
1. The "draft zero" concept positions AI output as:
Correct. Draft zero is earlier than first draft — the author has not yet arrived. The author arrives in revision, which is where voice enters the text.
Not quite. Draft zero is explicitly pre-authorial — before the writer has arrived. Its purpose is to give the writer material to react to, not a near-final text.
2. Stephen Marche's 2022 Atlantic essay demonstrated that AI text can become indistinguishable because:
Correct. Voice absorption through editorial revision — not AI capability — made the text indistinguishable. The mechanism is the writer's editing process, not the AI's output quality.
Not quite. The indistinguishability came from Marche's editing process — voice absorption — not from AI quality. The AI gave him raw material; his editing transformed it.
3. Csikszentmihalyi's flow research is relevant to AI drafting because it establishes that:
Correct. The cognitive load asymmetry — initiation is harder than continuation — is why AI's ability to eliminate the standing-start problem is genuinely useful rather than just convenient.
Not quite. The relevant finding is the cognitive load asymmetry: starting creative work costs more mental energy than continuing it. AI drafts address this specific asymmetry.
4. Charlie Warzel's documented technique used AI drafts primarily by:
Correct. Warzel read AI drafts specifically for what was wrong — and his objections to the wrongness became the generative material for his actual writing.
Not quite. Warzel used the AI's wrongness as material — his disagreements with the draft became the content of his piece. Reaction was faster than creation from scratch.
5. AI language models default to generic prose primarily because:
Correct. Token probability prediction produces the statistical center of language — the most common phrasings, which are the least distinctive. Voice lives at the periphery of that distribution.
Not quite. The mechanism is statistical: predicting probable next tokens naturally trends toward the most frequent — and most generic — language patterns. Distinctive voice is statistical deviation, which models don't produce by default.
6. "Voice anchoring" in a prompt works by:
Correct. Providing your own prose as a prompt anchor gives the AI concrete rhythmic and tonal targets rather than descriptive labels like "professional" or "conversational."
Not quite. Voice anchoring means giving the AI your actual prose — sentences you wrote — so it has specific rhythmic reference rather than genre labels it will interpret generically.
7. The "Show Me Three Ways" method primarily helps writers because:
Correct. Journalists at the 2023 ONA conference noted that having to choose between AI versions required them to articulate voice characteristics they had never named — the choice is diagnostic.
Not quite. The value is in the act of choosing and explaining why. That explanation becomes a voice specification. Averaging the three versions destroys the benefit by producing the generic result you were avoiding.
8. The Reuters Institute 2023 study found that iterative AI drafting became counterproductive after too many passes because:
Correct. The Reuters Institute found journalists reported text growing smoother but feeling less like them after six or seven passes — cumulative smoothing had erased the productive roughness of their voice.
Not quite. The documented finding was that each additional pass optimized for fluency and internal coherence — incrementally smoothing out the distinctive roughness that made the writing personal.
9. The Gordon Lish / Raymond Carver parallel illustrates which principle of iterative AI feedback?
Correct. Lish's documented letters to Carver — archived at the Lilly Library — specified exactly what to cut and why. Named removals produce better revision than general "improve this" instructions, whether the editor is human or AI.
Not quite. The Lish parallel illustrates the primacy of specific critique: he named exactly what he was cutting and why. That specificity is what produced productive revision rather than generic improvement.
10. The "loop-break signal" in iterative AI drafting occurs when:
Correct. Inarticulability is the signal: if you can't name what's wrong, it's a voice problem, which iterative specification cannot solve. The response is to stop iterating and write.
Not quite. The loop-break signal is specifically the inarticulability of the remaining problem. "Fine but wrong" means the issue is about voice — and voice cannot be specified into existence through more critique passes.
11. Clarkesworld Magazine's 2023 submissions closure was significant to the field because:
Correct. Neil Clarke's public response drew a clear line between AI as a tool writers may use and undisclosed AI-generated output submitted as original work — making explicit a distinction the industry had been circling.
Not quite. The closure's significance was in how Clarke framed his response: he distinguished AI tool use from authorship substitution, helping the industry articulate an operational distinction it needed.
12. The US Copyright Office's March 2023 guidance on AI-generated content established that:
Correct. The Copyright Office established a within-work distinction: human-authored elements are copyrightable; purely AI-generated elements are not. The Zarya of the Dawn case applied this to images alongside human-authored text.
Not quite. The guidance established that copyright applies to human-authored elements within a mixed work. A work containing AI content is not wholly unprotectable — only the non-human-authored portions lack protection.
13. Penguin Random House's 2023 contract update regarding AI required authors to:
Correct. PRH's contract update created a disclosure warrant — not a ban on AI use. The operative criterion is "substantially" and "without disclosure," consistent with the industry's regulation-not-prohibition approach.
Not quite. PRH required a warranty of disclosure for substantial AI content — not a ban on use. The language was "does not consist substantially of AI-generated content without disclosure."
14. Why does the ethical obligation around AI disclosure differ between a bylined magazine article and an anonymous corporate blog post?
Correct. The byline's implicit promise — a human wrote this — creates an ethical disclosure obligation that anonymous corporate content does not carry to the same degree. Disclosure obligations are relational, not uniform.
Not quite. The distinction is about the implicit authorship promise. A byline promises a human author; an anonymous corporate post makes no such promise. The ethical weight of disclosure follows from that implicit promise.
15. The most accurate description of current publishing industry norms around AI-assisted writing is:
Correct. A working line between AI as tool and AI as authorship substitute has been drawn, but legally and ethically contested edge cases — like heavily edited AI text — remain unresolved as of 2023–2024.
Not quite. The honest assessment is that norms are evolving faster than policies. A working distinction exists between tool use and authorship substitution, but the edges — including heavily edited AI text — remain genuinely unresolved.