AI Reveals the Most Human Parts of Writing

 

A woman devotes hours to working on her young adult fantasy novel. When faced with writer's block, she usually contemplates banging her head against the wall or turning to a favorite book for inspiration. Instead, she decides to leverage an AI writing tool. This tool processes her current chapter and generates potential paragraphs for the next part of her story. While these AI-generated suggestions may not always align with her vision, they occasionally contain beautiful sentences or intriguing directions. Even when the AI falls short, it reignites her interest in the narrative, offering a unique perspective.

The rise of high-quality computer-generated text introduces a novel concept—a quasi-writing companion for authors. This AI partner, although not infallible, faithfully attempts to fulfill the writer's requests without seeking credit. Previously, obtaining fluent text on a chosen topic, outside of collaborating with another human writer, was challenging. This prompts writers to ponder which aspects of writing they find tedious, which bring them the joy of creating from scratch, and what elements they cherish most about the writing process.

Over the last five years, I've delved into computer-generated text systems as part of my PhD research in human-AI interaction. The woman mentioned earlier, immersed in her fantasy novel, is based on one of the individuals I interviewed for my study on the social dynamics of writers seeking and integrating support from computers.

To comprehend the potential impact on writing, it's helpful to categorize writing into three components: planning, drafting, and revising. These are viewed not as sequential stages but as cognitive processes that can occur at various points during a writing project. Analyzing these specific facets enables a detailed understanding of how computers might influence writing overall, shaping a future that aligns with our preferences.

Returning to the writer immersed in her novel, she willingly lets the computer assist during writer's block but staunchly believes that crafting the plot is inherently human. The plot represents the story she wishes to convey, embodying her intention and refinement through workshops. Many writers share her sentiment, considering the problem-solving aspect of determining a story's direction as uniquely human.

However, when an AI system proposes a potential solution, it can be perceived as a challenge, with the computer's ideas serving as a starting point for the writer to enhance. For instance, a TV scriptwriter emphasized the need for uniqueness and originality in her field. Although an AI system might suggest a scene, she feels the responsibility to surpass its contribution with something more unexpected and insightful.

Despite this, some writers find joy in planning, considering it a complex, intellectual, and distinctly human aspect of their work. Planning involves activities beyond actual writing, such as deciding plot points, endings, and openings. The challenge of figuring out how a poem should conclude, while difficult, is a rewarding task that writers value.

Moving to the task of translating thoughts into words on the page, often termed "translating" in cognitive psychology research, it aligns with the moment when a writer's mind goes blank. This is precisely the task many computer systems are trained to perform—predicting what comes next. The role of AI as drafting companions represents a departure from traditional writing assistance, becoming a significant selling point for existing tools. These tools, like SudoWrite, Jasper.ai, Lex, and Laika, help writers by continuing where they left off, responding to specific instructions, or generating ideas based on described situations.

Despite the effectiveness of these tools, there is minimal focus on revision. While some writers are eager for AI-generated sentences, others hesitate to let an external entity dictate their words. Writers value the delicate early stages of a project and may be wary of computers disrupting their intentions. The idea of computers providing feedback is not widely explored in commercial systems, but there's potential for technology to generate meaningful questions about a writer's work.

As we navigate the nuanced ways computers impact writing, protecting authenticity and intention remains crucial for writers. Viewing computer-generated text as a dance with the writer's text is a helpful analogy—writers are open to collaboration as long as they retain control. The hope is that computer-generated text expands opportunities for individuals to express their visions, whether through planning, drafting, revising, or other means. While fully computer-generated books may not surpass human-authored works soon, the nuanced conversation around the various ways computers influence writing will continue to evolve.

Considering computer-generated text from a reader's perspective adds another layer to the discussion. When will computers blur the line between a person's voice and a machine's output? The concept of communicative intent becomes crucial. If a novel is mostly written by a computer based on a writer's idea, does it still reflect the writer's story? These questions will be answered culturally, reflecting our perceptions of authenticity and truth. Examining where computers can contribute will aid in shaping this understanding, ensuring it aligns with our core values in writing