Use this skill when a writer already has draft fragments or prototype text and wants the system to propose interactive follow-up directions for expanding, reorganizing, or refining the argument. It is especially useful for rapid draft prototyping where generated text and interface suggestions co-evolve.
Paradigm: P11 — Artifact-Driven Prompt Enhancement
Domain: argumentative writing
Source: Zhang et al. (2023). VISAR: A Human-AI Argumentative Writing Assistant with Visual Programming and Rapid Draft Prototyping.
The user has an existing draft, outline, or generated argument prototype and needs AI-derived suggestions tied to that artifact to decide what to elaborate, strengthen, rebut, or reorganize next.
A→G, T→G, G→I, H→I→Aug, T→Aug
| Symbol | Role |
|---|---|
| H | writer iterating on an argumentative draft |
| T | high-level revision goal such as strengthen evidence or improve counterargument |
| I | selection among AI-suggested revision actions, highlighted regions, or follow-up controls |
| Aug | revised instruction combining the writer's goal with chosen artifact-grounded suggestions |
| G | LLM-based assistant that analyzes the current draft and proposes next-step refinements |
| A | existing draft, paragraph prototype, or argument outline used as contextual grounding |
| Widget | Binds To | Example |
|---|---|---|
| draft region highlighter | artifact span selection for targeted revision | Select the paragraph containing the weakest evidence to request strengthening. |
| AI revision suggestion chips | artifact-derived follow-up actions | Chips such as 'add counterargument', 'clarify warrant', 'expand evidence', or 'improve conclusion'. |
| argument gap inspector | detected weaknesses in draft structure | A panel flags that Claim 2 lacks supporting evidence and offers a one-click fix path. |
| prototype variant switcher | alternative draft directions | Toggle between a more assertive introduction and a more balanced introduction before committing to one. |
Return type: composite
A user has a rough argumentative draft about remote work policy. The system analyzes the draft and surfaces suggestion chips like 'add opposing viewpoint' and 'strengthen evidence in paragraph 2.' The user clicks the weak paragraph, chooses 'add opposing viewpoint,' and the system forms an augmented instruction grounded in that selected text and revision goal, then produces a stronger revised version.