Large Language Models (LLMs) are transforming natural language generation and spreading into uses such as narrative, scriptwriting, and literature. LLMs offer new modes of creativity through their capacity to generate language in multiple stylistic variations and contexts a way that may enhance creativity. LLMs are exciting and challenging for writers considering the use of AI in their creative writing. They certainly have their advantages and can contribute positively to creativity, however, LLMs do not readily demonstrate elements of cultural awareness, consistency in storytelling, or faithfully represent styles of writing of humanity. Humans, no matter how they are generated, still possess emotional intelligence, contextual awareness, and the ability to create meaningful narratives. The biggest challenge is harmonizing contributions by both human and machine without losing the integrity of the story. In this paper, we present the Collaborative Authoring Framework for Enhanced Storytelling (CAFES) as an answer to this challenge. This framework allows authors to direct the narrative, tone, and character development with LLMs acting as co-authors. CAFES employs adaptive prompt engineering, iterative human-in-the-loop editing, and reinforcement-guided fine-tuning to enhance the coherence and alignment of machine-generated content. Experimental creative writers using CAFES reported better story inspiration, fewer instances of writer’s block, and improved narrative flow compared to baseline LLM-assisted writing. Finding a balance between narrative voice and machine-generated recommendations improved author happiness. Ultimately, CAFES demonstrates how LLMs can collaborate with humans to generate novel narratives by combining human and AI creativity.
Moti Ranjan Tandi (Thu,) studied this question.