Presentation Type

Poster

Presentation Type

Submission

Keywords

artificial intelligence, ChatGPT, creative writing, authorship, pre-writing, post-writing

Department

Creative Writing

Major

Creative Writing and French Studies

Abstract

This study articulates how large language models (LLMs) can aid creative writing as supplements, not replacements to human authorship. Alongside Dr. John Struloeff and research assistant Tiffanie Richerme (‘25), our project adds to the ethical implementation of artificial intelligence in literary settings, particularly institutions of higher education studying creative writing. Initial testing compared LLMs’ narrative outputs across three text generators—Google Gemini, OpenAI’s ChatGPT, and xAI’s Grok. Each was asked to generate short stories evaluated on creativity and authenticity. These prompts asked LLMs to incorporate dialogue realism, avoidance of archetypal character tropes, and effective narrative pacing. Each was limited in creative liberty, as revealed by a tendency to resolve narratives and eliminate the possibility for vague, interpretative endings, even when prompted to abstain thematic closure. ChatGPT illustrated strongest human-like prose and was selected as the central model studied. ChatGPT was employed as a tool for pre-writing and post-writing, helping storyboard and proofread for grammatical errors or thematic plot-holes. The model generated a three-phase questioning framework to inspire character development, central conflict, and thematic reflection. Pre-writing phases proved more effective in expanding character and plot than post-writing phases devoted to editorial and thematic mechanics. While pre-writing questions sparked inspiration, post-writing process tended to constrain the writer to revert thematic understandings to pre-existing, conventional conclusions devoted to clarity rather than stylistic experimentation. These findings suggest that LLMs can be effective in prompting creative inquiry for initial writing processes. Beyond this, LLMs tend to restrain authentic authorship by preserving tropic character arcs with resolved endings.

Faculty Mentor

Dr. John Struloeff

Location

Waves Cafeteria

Start Date

10-4-2026 1:00 PM

End Date

10-4-2026 2:00 PM

Share

COinS
 
Apr 10th, 1:00 PM Apr 10th, 2:00 PM

Supplement, Not Replacement: Investigating LLMs as Tools in the Creative Writing Process

Waves Cafeteria

This study articulates how large language models (LLMs) can aid creative writing as supplements, not replacements to human authorship. Alongside Dr. John Struloeff and research assistant Tiffanie Richerme (‘25), our project adds to the ethical implementation of artificial intelligence in literary settings, particularly institutions of higher education studying creative writing. Initial testing compared LLMs’ narrative outputs across three text generators—Google Gemini, OpenAI’s ChatGPT, and xAI’s Grok. Each was asked to generate short stories evaluated on creativity and authenticity. These prompts asked LLMs to incorporate dialogue realism, avoidance of archetypal character tropes, and effective narrative pacing. Each was limited in creative liberty, as revealed by a tendency to resolve narratives and eliminate the possibility for vague, interpretative endings, even when prompted to abstain thematic closure. ChatGPT illustrated strongest human-like prose and was selected as the central model studied. ChatGPT was employed as a tool for pre-writing and post-writing, helping storyboard and proofread for grammatical errors or thematic plot-holes. The model generated a three-phase questioning framework to inspire character development, central conflict, and thematic reflection. Pre-writing phases proved more effective in expanding character and plot than post-writing phases devoted to editorial and thematic mechanics. While pre-writing questions sparked inspiration, post-writing process tended to constrain the writer to revert thematic understandings to pre-existing, conventional conclusions devoted to clarity rather than stylistic experimentation. These findings suggest that LLMs can be effective in prompting creative inquiry for initial writing processes. Beyond this, LLMs tend to restrain authentic authorship by preserving tropic character arcs with resolved endings.

 

To view the content in your browser, please download Adobe Reader or, alternately,
you may Download the file to your hard drive.

NOTE: The latest versions of Adobe Reader do not support viewing PDF files within Firefox on Mac OS and if you are using a modern (Intel) Mac, there is no official plugin for viewing PDF files within the browser window.