AI Control Profiles

This research introduces a control-profile taxonomy that replaces binary “assisted vs. generative” framings with a spectrum of human creative input and directional authority.

The Problem

Current discourse around AI in publishing uses imprecise language: a story is either “AI-assisted” or “AI-generated,” obscuring the actual work of human authorship that occurred. A writer who prompts an AI model for a scene outline has far more creative control than a writer who copy-pastes the AI’s output verbatim—yet both fall under “AI-assisted.”

Control Profiles

We propose five profiles that measure the quality and depth of human creative direction:

Profile 1: Human Direction (90-100%)

The human author controls plot, character, voice, theme, and execution. AI is used only for tactical assistance (brainstorming, editing suggestions, copywriting refinement).

Example: Author outlines the entire story, writes the first draft, uses Claude for line editing and dialogue polish.

Profile 2: Collaborative (70-90%)

The human author controls high-level vision (plot, character, theme) but delegates tactical execution to AI within clearly defined constraints.

Example: Author writes the story outline and character briefs, prompts an AI model to draft scenes, then substantially edits the output.

Profile 3: Directional (50-70%)

The human author sets the direction (concept, broad outline, key constraints) but accepts substantial AI-generated material with light revision.

Example: Author provides a plot skeleton and tone guidelines; AI drafts the full manuscript; author edits for continuity and voice.

Profile 4: Prompt-Guided (30-50%)

The human author provides prompts and accepts most of the AI output, making selective edits and revisions.

Example: Author writes detailed prompts; AI generates scenes; author cherry-picks best sections and stitches them together.

Profile 5: Minimal Direction (0-30%)

The human author provides minimal input (a title, a genre, or a few keywords) and publishes largely unedited AI output.

Example: Author requests a short story in a given genre; AI generates it; author publishes with only metadata edits.

Disclosure Framework

Each published work should include metadata indicating which profile(s) were used and where. This allows readers to understand the role of human authorship.

Why It Matters

Clear control profiles enable:

  • Reader trust — transparency about how stories were made
  • Author accountability — credit for actual creative work
  • Market clarity — labels that help readers choose according to their preferences
  • Industry standards — vocabulary that enables policy and norms

Further Reading:

  • Control, Disclosure, and the Future of Publishing
  • The Author’s Role in AI-Generated Narrative