The Agency of A.I.

Or, A.I. and it’s relationship with Art

This article features an interview from Riya Cyriac. She wanted to make it clear that the opinions she states in her interview are hers alone and have nothing to do with her workplace.

Winter in Georgia is a light affair. You don’t need more than a sweater and a jacket to stay warm, but last winter was colder than usual. It brought snow in January. Not fluffy cotton tufts but hard icy pellets, crunching under boots like hail. That winter, in December, I attended the graduation party for a family friend. Hosted at their childhood home, the memories of yesteryears greeted me at the foyer, the cookouts and late dinners and summer camps. We were all eating in the basement when the celebratory graduate said something that piqued my interest. He said that he only really made it through that final year thanks to ChatGPT. It was not unusual to hear a college student admit to a bit of cheating to get through classes. We’ve all looked through Quizlet for quiz answers or looked for textbook answer sheets. What was new was the salience of that source.

ChatGPT was unleashed onto the public in November 2022 and has altered every sector of life. The internet is now inundated with A.I. pictures, videos, and text. Job boards are flooded with A.I. trainer positions and people are using ChatGPT as their therapist. In the background people are ringing the alarm and claiming that A.I. will steal everyone’s job. Including authors. With lawsuits against A.I. companies from prominent authors such as Ta-Nehisi Coates, Michael Chabon, and Junot Díaz, the fear is palpable over what A.I. could bring for the future of literature. But what exactly is A.I. and how does it interact with the arts?

“I call myself like an A.I. skeptic”

“So A.I. doesn’t exist actually,” said Angela Collier, a YouTuber with a PHD in physics. In her video “A.I. doesn’t exist but it will ruin everything” she explained what constitutes an A.I. What we refer to now as A.I.—ChatGPT, Gemini, Grok, DeepSeek—should more accurately be described as machine learning systems or LLM’s (Large Language Models). This means they are trained on a data set for a specific purpose, not that they are actively making intelligent inferences based on the data. 

“ChatGPT’s goal is to make texts that look as if a human wrote it” said Dr. Collier in her video. These models are more generative than intelligent, as they aren’t concerned with the veracity of their product, but that there is a product to begin with. That’s why the phenomena of A.I. hallucinations exist. As reported by The New York Times, “sometimes, they just make stuff up” and while they are getting better with math equations, these hallucinations have yet to be solved. With the ever growing usage of A.I. by everyday people, it seems these technicalities and faults aren’t deterring the general public from treating ChatGPT like an all-knowing friend to talk to. That’s also true in the realm of art.

“I call myself like an A.I. skeptic,” said Riya Cyriac, an Associate Product Manager at Salesforce and founder of the nonprofit The Young Writer’s Initiative. In her undergraduate thesis “The Metrics That Matter: Rethinking How We Measure Co-Creation with GenAI for Creative Writing”, Riya outlines how researchers often use A.I. centric metrics instead of human centric when conducting studies on A.I. generated writing. “That’s the only way they [researchers] know how to measure if their product has any value” explains Riya. “It’s much harder for a researcher to capture these qualitative things.” Problems emerge with this focus A.I. metrics. 

“I support A.I. when it augments humanity… not when it actively erodes us.”

Productivity is prioritized in these A.I. studies. This is fine for an A.I. tool which Riya described as an “efficiency tool first and foremost”, however when you combine that with the writers who would use this tool, it’s not compatible. Writers and the arts in general don’t tend to prioritize the final product. There is a product produced, but they are more concerned with what they learned while making the product and the quality of the product rather than the quantity or efficiency of the product-making process. In her thesis, Riya analyzes 40 Paris Review interviews with acclaimed writers and synthesizes metrics critical to how writers assess themselves. Notably she found that these authors prioritized their creative process and struggle more than the actual product. 

The idea of creative struggle is a friction point between A.I. and writers. As an efficiency tool, A.I is designed to relieve the cognitive load of tasks. For writers, the creative struggle is paramount to the process as it’s through this struggle that a story takes shape. In other professions, struggle is a pain point; in the arts it’s the price of distilling the soul. 

“You are genuinely using your critical thinking skills and your brain to unearth something that’s new,” said Riya. “That’s the process of creation.” There are studies that say A.I. can negatively affect your cognitive skills. Researchers at MIT conducted a study using LLM’s, search engines, and a brain-only group to write essays. There were signs that A.I. tools “may unintentionally hinder deep cognitive processing, retention, and authentic engagement with written material”; in essence you think and retain less when using A.I. with your work.

A.I. can’t perform complex tasks like creativity. That is, it cannot create something “original” as a human can, it imitates. Riya demonstrated this by trying to use ChatGPT as an editor. She found that ChatGPT was a passive agent, acting only when prompted and generally it giving positive feedback. It cannot give opinions based on it’s experiences or feelings. This is an issue for something as organic as a workshop, when work needs criticism and constructive feedback, something current A.I. doesn’t have the complexity for. Apple released a study highlighting this fact in June, stating Large Reasoning Models face “a complete accuracy collapse beyond certain complexities.”

Still, these are issues for writers attempting to use A.I. for their work, not for an ordinary person who wants to generate a short story for fun. It seems more like the tech world is developing their LLM’s for the latter, not concerned with the professionals. “A.I. developers already lost the game,” said Riya in reference to attracting writers to A.I. But there are ways to still focus on use cases less devastating to the writer ecosystem. 

Riya saw a way forward. “What matters now is that it’s developed ethically” she said describing herself is a human-first product manager. “I support A.I. when it augments humanity... not when it actively erodes us.” To avoid this erosion it’s important for the product managers behind the developers to be cognizant of the effects the features of A.I. can have on the user as the developers “don’t think about the human side.” Riya advocates for more consideration for human centric metrics in A.I. research, so that these tools aren’t developed in opposition to the interest of those who may use them. 

She’s not alone in this. There are many studies that go into the topic of co-creation between A.I. and humans. One such study by students from the University of Edinburgh proposes the UCCC (User-centered Co-Creativity) model in which the user can choose how much input the A.I. tool has on their work before the tool starts to augment the creative process, giving the artist more control over how A.I. influences the end product. As A.I. continues to evolve at breakneck pace, it’s clear that considerations for human users and it's affect on society need to develop just as fast. If not, A.I. may start consuming us instead.

---

I was not very surprised or incensed when my friend revealed his usage of ChatGPT in his college work. It seemed to me at the time much like the usual ways college students got around work. But now, with the willful and gleeful outsourcing and automating of thinking to these A.I. models, I wonder exactly when convenience becomes a crutch. When Sam Altman posted an A.I. story earlier this year and said it got the “vibe of metafiction”, it seemed to me much like a college student who’d rather pass a class than thoughtfully engage with coursework even for a sliver of time. 

Riya shared my sentiments, declaring that “being a writer isn’t a right”, it’s something to work towards because it’s worth it. It’s not simply that writers don’t pay mind to efficiency, it’s also that exploring and expanding yourself to form a story takes time and patience and frustration and grit. Writing is not for the weak. Every writer who’s trudged to the end of a story will attest to that. It’s not that we’re all masochists. Riya puts it best: “Creative writing… that’s part of human joy.” 

---

About the author: Tosin Ebunola is a writer with a Bachelor of Arts in Creative Writing from Mercer University. She was previously shortlisted in the Young Creatives Nottingham competition for Creative Writing. She lives in Georgia and is a reader for the literary magazine WayWords.