Step Away from the Chatbot: a Letter to a Student about AI and Creativity

Dear Student, 

I’m glad this letter reached you before you fed that assignment prompt from your Creative Writing professor into ChatGPT. I’d like to share some ideas that may be helpful as you decide whether to follow through on that plan. First of all, I’m sorry you’ve been finding the poetry unit of Introduction to Creative Writing so tedious and uninspiring. It’s true that you probably won’t be writing much poetry in your future career. And it’s also true that the free technology at your disposal—an ever-growing suite of large language models—will produce something likely to receive at least a passing grade for the poem you have been assigned to write. Sure, you know your professor has forbidden the use of artificial intelligence (AI), but you also know she will not be able to definitively prove your guilt (even if she suspects it). All you need to do is copy and paste the assignment prompt into the chatbot, maybe introduce a few typos to throw your professor off the scent, and you’ll be done with this assignment in minutes, free to use your time however you please. It’s all gain and no pain. So, why not do it? 

In this letter, I am going to give you three reasons to close the ChatGPT tab and write the poem yourself. None of these reasons are directly about morality, the grade you will receive, the career you hope to attain, or the likelihood that you will get caught. They are not about the environmental impact of AI, the rights of the human poets whose work has been used without consent to train the models, or academic integrity. To be sure, I think there are good reasons to avoid using AI that fit into some of these categories, but they are not the focus of this letter. Instead, the reasons I will offer have to do with the importance of exercising your own creativity. These reasons will generalize beyond your poetry assignment. They will support the conclusion that you should do other creative things for yourself instead of outsourcing them to AI as well. 

But before I make my case, I will address two bad reasons people sometimes give for avoiding reliance on AI in artistic endeavors, and I will explain why my argument does not rely on these claims. 

Bad Reason 1: The AI-generated product won’t be any good. 

 It may be true that AI-generated products are somehow deficient as art, but there is no need to assume that AI-generated products are altogether bad in order to show why you should write this poem yourself. After all, it is increasingly clear that some people find some AI-generated products—text, visual images, and even music—to be aesthetically valuable. Take, for example, the aesthetic value of humor. Author Simon Rich recently shared the following headlines in the style of The Onion that he generated by prompting the AI model code-davinci-002:

Experts Warn that War in Ukraine Could Become Even More Boring.

Budget of New Batman Movie Swells to $200M as Director Insists on Using Real Batman

Story of Woman Who Rescues Shelter Dog With Severely Matted Fur Will Inspire You to Open a New Tab and Visit Another Website

Rich found code-davinci-002’s ability to generate these joke headlines disturbing precisely because, in his judgment, they are genuinely very funny. Rich is someone whose judgment it would be reasonable to take seriously—he’s a successful humor writer, one of the youngest ever to be hired by Saturday Night Live. But you don’t have to take his word for it. Do you think these are funny? If not, I’d be willing to bet you could get an AI model to generate something you do find funny, beautiful, elegant, or even sublime. These are all ways of being, in some sense, aesthetically good.

Consider a simple poetry prompt like the one your professor has given you: “Write a poem (10–20 lines) that vividly captures the sensory details of a meaningful moment you have experienced. Describe what you saw, heard, smelled, tasted, or touched and show how those details made the moment meaningful to you.” I plugged this text into ChatGPT, and it returned the following: 

The swing creaked in slow rhythm,
chains warm from the sun.
Dandelions scattered across the grass,
their seeds drifting like breath.

My hands gripped the rubber seat,
slick with summer sweat.
Somewhere, a sprinkler ticked,
sputtering arcs of water and laughter.

The air smelled of cut grass and charcoal,
a breeze lifting the edge of my shirt.
I bit into a peach,
its juice running down my chin—
sweet, sun-heavy, and wild.

In that moment,
nothing asked for more than now.
The sky was just sky,
blue and endless.

While this may not be an outstanding poem, it seems to me to be far from bad. Even though I knew it was generated by AI, I found the image it depicted to be quite lovely. It included some commendable elements, such as the evocative description of the peach’s juice as “sun-heavy.” I even found the enigmatic line “nothing asked for more than now” somewhat beautiful. It is certainly better than some human attempts at poetry I have seen. It may even be better than what you yourself would write. The reason I think you should write your own poem for this assignment is not that I think your poem would be aesthetically superior to a poem generated by ChatGPT. Perhaps it would be. But the reasons I will offer below support the idea that you should write the poem yourself, even if it turns out to be less good than what an AI model could generate. 

Bad Reason 2: It is impossible to use AI in any way and still be creative. 

Just as I will not argue that all uses of AI are aesthetically bad, I will not argue that all uses of AI are uncreative. Creativity can certainly be involved in how one prompts an AI model. And as Anthony Cross has argued, there ways of creatively deploying AI in the creation of art that go beyond creative prompting. This is particularly clear when artists work within what Cross calls the “exploration paradigm,” a mode of engaging with AI models that involves a playful interrogation of their capabilities. In this mode, an artist can creatively interact with an AI model in such a way that the interaction itself is part of the piece’s aesthetic value. 

Cross illustrates this mode of creatively engaging with AI with the example of Sofia Crespo’s work Critically Extant. In this series, Crespo exploits the fact that AI models—despite having been trained on millions of images of plants and animals—have impoverished resources for representing species that are rarely photographed. When prompted to produce images of these species, the AI model produced bizarre images that fail to look like plausible specimens. In this way, Crespo creatively reveals something about contemporary AI models, something about the species in question, and something about her own perspective as an artist. Far from making her work less creative, the employment of AI is part of what makes Crespo’s project so creative. This clearly shows that it is possible for someone to be artistically creative while utilizing AI. 

But, if my characterization of your plan is correct, the way you are thinking of using ChatGPT to complete your creative writing assignment is not creative. It is the full-blown outsourcing of all the creative labor that would be involved in writing a poem. So, the fact that someone could deploy AI creatively is not relevant to the question of whether you should simply feed your assignment prompt into ChatGPT and submit the results. 

Good Reason 1: Creativity is not just about bringing things into existence. It’s also about learning. 

I will turn now to the first reason you should write the poem yourself. By giving you this assignment, your professor has invited you to attempt to make a piece of art. Creating a piece of art—even if it doesn’t turn out to be especially good—is an opportunity to learn something. One sort of value that creativity always has is epistemic value. In simple terms, this means that every time someone succeeds at being creative in art or other domains, they thereby learn something. A scientist who succeeds in creatively devising a theory to explain perplexing data learns about a new possible explanation of their observations. A philosopher who creatively conceives of a counterexample to a compelling moral theory learns how the theory may be flawed. A sculptor who creatively incorporates an unexpected substance into their work learns of new aesthetic effects they can achieve within their art form. 

What might you learn by composing a poem for your creative writing class? Well, first of all, if you’ve been given the prompt above about a meaningful experience you’ve had, you’ll need to sort through your memories to identify ones that are especially important to you. As you do, you’ll have the opportunity to make aesthetic judgments about these experiences. That is, you’ll assess how beautiful you find them and what, in particular, you find beautiful about them. This reflective exercise is a creative curation of your own experiences. Kenneth Walden has argued, following Immanuel Kant, that when we make aesthetic judgments, the beauty we perceive is in part a response to the thing we are considering and partly dependent on our own creativity. We don’t just passively observe beauty; we actively participate in creating it through our modes of aesthetic appreciation. 

So, before you’ve even started writing the poem, you’ll have used your creativity to discover something about what moments in your life you’ve found most meaningful and what you value about them, aesthetically. And once you start writing the poem, you’ll have to find creative ways to represent the sensory details of the moment you’ve chosen in words. When you succeed in doing so, you’ll learn what words are apt and inapt to capture your perspective about the experience. You’ll learn about aesthetic effects it is possible to achieve with meter, lineation, and other elements of poetic form. You’ll learn about the contours of your imagination, your vocabulary, and your taste. You’ll learn what matters to you about the moment you chose. In sum, you’ll learn more about the experience, more about the art form, and more about yourself. 

Now, you may object that you will also learn something if you use ChatGPT to generate a poem. And that may be right. My claim is that all creativity involves learning, not that all learning must happen through the use of creativity. But is what you might learn if you use ChatGPT to write your poem as valuable as what you might learn by doing the assignment yourself? 

Go read that poem from ChatGPT again. What did you learn? The poem portrays someone sitting on a swing set on a hot summer day, eating a peach. Imaginatively engaging with that depiction may well be a rich experience for you. You have probably never imagined how dandelion seeds drifting through the air could resemble breath, for instance. Perhaps you imagine this and find it beautiful. In doing so, you may have a new aesthetic experience as a result of reading the poem.

But valuable though the experience may be, what you learn by reading a poem you have ChatGPT generate for you (assuming you, in fact, take the time to read it before submitting the text to your professor) will be different in character from what you would learn by writing a poem yourself. I will explore that difference in the next section. 

Good reason 2: Creativity offers a path to figuring ourselves out. 

When you exercise your own creativity, you have a uniquely good opportunity to understand yourself better. This is, in part, because being creative necessarily involves self-disclosure. To be creative is to reveal something about who you are. When you are creative, you have a kind of ownership over your actions, and a kind of responsibility for them. This kind of responsibility is captured well by Gary Watson’s notion of attributability. Watson argues that when what I do is attributable to me, it discloses something about what I am committed to, and what matters to me. Things that are attributable to me are “inescapably my own.” When you toil for hours over the perfect phrase to capture the exuberant mood of your high school graduation and land on “the gymnasium lights glittered in flashes of triumph,” that creative characterization is inescapably your own. It reveals something about how you saw that moment in your life and what you found meaningful about it. 

But if, instead, you ask ChatGPT, as I did, to write a single line of poetry that captures the excitement of graduating high school, and it returns “caps fly like startled birds as we leap into the sky of what’s next,” the line of text is not attributable to you. It does not reveal how you saw the moment or what you found meaningful about it. It may happen to accord with your feelings about the moment. But it is not revelatory. It is not yours. It does not afford you the opportunity to learn about yourself in particular.

Learning about yourself is important. Jordan MacKenzie has argued that knowing who we are involves knowing what we value, what character traits we have, what we are capable of, how we feel, what makes us happy, and what we believe, among other things. This kind of self-knowledge is, on MacKenzie’s understanding, a matter of self-love. 

How might creativity enable this sort of self-knowledge, and ultimately a healthy form of self-love? One possibility is that, if you adopt a creative practice (such as writing poetry) and commit to it over time, you may eventually become good enough at it to see yourself in a new light. This is true in part because work that is more creative is correspondingly more self-disclosive. One way that highly creative achievements might show you something important about yourself can be found in Thi Nguyen’s work on the nature of games. Nguyen argues that there is a kind of “harmony between self and challenge” one can experience in the context of playing a game like chess:

When a chess player discovers an elegant move that lets them escape a trap while simultaneously adding pressure to their opponent, the harmony of the move—the elegant fit between the challenge and the solution—is available both to the player themselves and to outsiders. But there is something more that is available only to the player: a special experience of harmony, of a fit between one’s awareness, one’s problem-solving and decision-making abilities, and the elegance of the output. It’s not just that the solution fit the situation; it’s that one’s abilities fit the demands of the situation. 

A similar sort of harmony, I contend, can be experienced in the case of creative artistic achievements. If you hone your poetry skills through sustained creative work, you might find that you are able to rise to the creative challenges of the art form. The feeling of harmony Nguyen describes is an aesthetic experience, and also an instance of learning about yourself. In experiencing the beauty of your own agency, you learn more about what you are capable of. But if ChatGPT is the composer of your poem rather than you, you will not have risen to the occasion and creatively met the challenges of the art form. You will have handed those challenges off to an algorithm. As a result, you will not unlock the experience Nguyen characterizes here. You will miss out on the opportunity to glimpse the beauty of your own artistic agency. 

Creativity offers a path to understanding who you are. By choosing not to take that path, you sacrifice the discoveries it would enable you to make. Moreover, by choosing not to take that path, you forego an opportunity to allow others to see who you are as well. This brings me to my final point. 

Good reason 3: Creativity enables meaningful connection with others.

It is hard to see ourselves accurately, and it is also hard to see others accurately. For this reason, relating authentically to other human beings is no trivial achievement. Creativity is particularly valuable in enabling human connection, in part, because it is especially revelatory about who we actually are. To see why, recall my point above that creativity always results in learning. What we learn when we are creative is not entirely accidental. In order to be creative, one must deliberately seek to learn something. In this way, creativity necessarily involves curiosity. We can fake curiosity, but only to a degree. Lani Watson proposes the following test for determining whether someone’s curiosity is genuine: “In general, it is misplaced to ascribe curiosity about X to someone who, when offered information about X, at no cost to themselves, nonetheless declines it.” So, if my student tells me they’re curious about my research, but then declines my offer to explain it to them, it would be reasonable for me to conclude that their initial expression of curiosity was insincere (maybe they were just trying to get on my good side by feigning interest). 

But when someone does something creative, they reveal genuine curiosity about something. It may be difficult from the outside to tell what they wanted to learn. Did your classmate write a poem about attending her first baseball game because she was curious about the aesthetic properties of that experience or because she was curious about what it would take to earn an A on the assignment? We’d need to look very closely for evidence of one motivation or the other. But if the poem succeeded in being creative, that means she successfully explored something. By examining her poem closely, we will likely find clues about her true motivation. Because our creativity discloses our curiosity, it discloses an aspect of who we are that philosophers have called the inquisitive self. This is the part of you that is motivated by particular questions, interested in particular issues, and disposed to investigate in particular ways. 

If I tell you that I am interested in clouds, you would probably believe me. And if you see me reading a book about meteorology, you’ll have some evidence that my interest is genuine. But you’d learn much more about my interest in clouds if you encounter some creative work in which I explore that interest. When a meteorologist shares their creative theory about the mechanism underlying the formation of rare morning glory clouds, it reveals something very specific about their cloud curiosity. When Joni Mitchell sings of “cloud illusions,” attentive listeners learn what she finds puzzling about literal and metaphorical clouds. In both cases, you may find yourself relating to that interest and to the inquiry the person has undertaken. You might see the inquirer as a fellow traveler. This is one way that our creativity helps connect us to other people.

By contrast, imagine that you prompt ChatGPT to write a haiku about clouds, and it returns the following: 

Clouds drift silently—

stories written on the wind,

vanish in the blue.

Your request for a poem brought something new into existence. But does this poem give anyone the opportunity to learn anything important about who you are? When you read it, do you learn anything important about anyone else? If you are inclined to answer these questions in the negative, as I am, perhaps the reason is that this poem is not a creative product. ChatGPT did not produce this poem out of its own curiosity. And though your choice of prompt may have been motivated by some sort of curiosity, there’s nothing distinguishing about you in the final product. 

If this poem leaves you cold, perhaps that is because it takes the form of a creative product, but lacks one of the most valuable hallmarks of creativity: a window into who someone else is as an inquirer. The result is a sort of emptiness that can make encounters with AI-generated products feel uncanny and alienating. This point was captured by actor, comedian, and screenwriter Simon Pegg in an interview about the 2023 Writers Guild of America strike:

AI has had no childhood trauma. AI has never had a boyfriend or girlfriend, never had its heart broken. It’s never been through anything that would give it the impetus to create art. When it has its own personality and its own experiences, it can write a film about what it’s like to have no body. But for the moment it can’t do that, and so to rely on it would be to just make everything mediocre…We have to fight mediocrity in order to create great art. 

As Pegg suggests here, because AI lacks its own personality and its own experiences, it has nothing original to draw upon in creating art. It cannot, therefore, enable the experience of human connection that art offers when it resonates with us. But unlike ChatGPT, you do have a personality. You have experiences. You have a perspective. You have questions that interest you. You have what it takes to be creative. If you choose to make use of these resources, and you share what you create with others, you will afford them an opportunity to connect with you. 

Conclusion

Though I am not your professor, I am a professor. And I know I speak for many of my colleagues when I say that one of the best parts of being a professor is seeing students grow and flourish. Sadly, our view of what is happening with individual students has been somewhat obscured by the widespread availability of AI. Because AI is so good at mimicking the products of human creativity, your professor may never know for sure whether the poem you submit to her is the product of your own creativity. But you will know. I hope you like what you learn. 

Sincerely,

Dr. Lindsay Brainard 

Acknowledgements:

I am grateful to Ian Cruise and Keshav Singh for helpful feedback on earlier drafts of this letter.

Note: 

This is a shortened version of a piece to appear in Living Debates in Aesthetics, edited by Sarah Worth.