“Language Weaves Its Tapestry”: Crafting Found Poetry Using AI Tools

Ruth Li

34 “Language Weaves Its Tapestry”: Crafting Found Poetry Using AI Tools

Ruth Li

Abstract

In this chapter, I offer a pedagogical approach to teaching creative writing in the era of AI technologies. I share an activity that engages students in crafting found poetry using AI tools such as ChatGPT. Found poems are collage-like assemblages drawn from other sources such as newspaper or magazine articles. By immersing students in crafting found poems using AI, I inspire creative experimentation with AI-generated texts. Even as writing processes become automated and writing styles standardized, I posit that this found poem activity nurtures students’ careful attunement to the materiality of language, to its shape and texture as words unfold on a page. I share example student-written found poems that illustrate the creative possibilities of this approach. Furthermore, I address the ethical concerns of composing AI found poems, including patchwork plagiarism, and share ways to encourage students to reflect critically on their writerly choices. Ultimately, I explore the possibilities and tensions of leveraging AI technologies to stimulate critical thinking, lyric experimentation, and creative play.

Keywords: found poetry, creative writing, generative AI

Introduction/Background

The advent of generative AI technologies such as ChatGPT invites scholars and teachers to rethink what it means to create. Verse by Verse, a “Semantic Experience” created by Google AI, is “an experimental AI-powered muse that helps you compose poetry inspired by classic American poets.” The production of a machine-based model of poetry raises questions: What does it mean to write poetry in the wake of AI technologies? What is the role of the human in artificial intelligence? In what ways can we productively leverage AI tools to enhance critical thinking and creative expression?

In responding to these questions, I offer an innovative pedagogical approach to teaching creative writing in the era of AI technologies. I share an activity that engages students in crafting found poetry using AI tools. As described by the Academy of American Poets, “found poems take existing texts and refashion them, reorder them, and present them” as collage-like assemblages. Found poems are “often made from newspaper articles, street signs, graffiti, speeches, letters, or even other poems” (Academy of American Poets). As poetry expands into the digital realm, AI language tools could serve as source material for creating found poems, as a repository of information that can be leveraged for creative expression. My approach coincides with Wu’s (2023) activity “Dada, Machine Writing, and the Found;” while Wu’s assignment invites students to experiment with composing found art using machine-generated code, the found poem activity that I present below encourages students’ close attunement to the craft of writing. By immersing students in composing poetry using AI, I invite creative experimentation in ways that reanimate the human dimension of the composing process.

In the following, I share a description of the found poem activity followed by teacher reflection and analysis of example student-written found poems.

Activity Description: Crafting a Found Poem Using AI Tools

In this activity, you will craft a found poem using an AI tool such as ChatGPT. In a found poem, you assemble, recombine, and refashion words and phrases from other sources into a collage-like poem.

You can follow these steps to create your found poem:

Select one or more topics of your choice or ask ChatGPT (or another AI tool of your choice) to brainstorm ideas for topics. You could choose similar topics or very different topics as a way of inviting interesting juxtapositions or tensions in your poem.
Prompt ChatGPT to write a text(s) on one or more topics. The AI-generated text(s) does not need to be a poem; the text could be an essay, a short story, a speech, a recipe, etc.
Select 10-15 words or short phrases from the AI-generated text(s) (phrases should consist of no more than 3-4 consecutive words).
If you would like, you may print out the text. Using scissors, cut out the highlighted words and phrases that you have selected.
Rearrange the words and phrases to create your found poem. You could move the words and phrases around to try out different combinations. You do not need to use every word/phrase from the list of 10-15. You may add your own words, phrases, or lines if you wish (please clearly indicate which words are your own and which are from the AI-generated text(s) by using italics, highlighting, etc.).

As you compose the found poem, you are welcome to experiment with new or surprising meanings, including meanings that might conflict with or go beyond the meaning of the AI-generated text. You can “speak back” to the AI text by posing questions or sharing your own thoughts on the topic in a call and response format. You are also encouraged to play with design elements including font, color, spacing, and layout.

Additional guidelines:

You may create either a paper-based or digital collage; you may create a digital collage using an application such as Photoshop.
Include both the original text(s) generated by ChatGPT and your found poem.
Following the found poem, include a brief paragraph-length reflection describing your choices. You may consider questions such as:

- What inspired you to choose this topic(s)?
- How did you go about composing the poem?
- Why did you make the creative choices that you made?
- How was the experience of composing a poem using AI tools?

You will have an opportunity to share your found poems with your peers in the class. Together as a class, we will comparatively analyze the AI-generated source texts and the found poems and consider the ways the found poems may have altered or transformed the AI-generated texts.

Reflection/Analysis

In this found poem activity, I inspire students to experiment with language and form — with the forms that language can take in a poem. As students are invited to create either a paper-based or digital collage, the activity simultaneously sparks a delving into the future and a return to the human element — the material handicraft reminiscent of the Arts and Crafts movement. Even as processes become automated and styles standardized, this activity nurtures students’ careful attunement to the materiality of language, to its shape and texture as words unfold on a page. To compose is to assemble disparate elements into a greater whole, to fashion together individual strands into a synthesis. In crafting a found poem, students can learn to appreciate the way words can take on new meanings when recombined as elements are wrested from their contexts, refashioned, cast into new relations, and transformed through creative juxtaposition. The process of creating a collage-like found poem could serve as an exercise in the rhetorical process, including invention and arrangement, and could be interpreted as remix or remediation (Palmeri, 2012). A parallel can be drawn between AI texts and found poems: as in a found poem composed of newspaper clippings, an AI language tool is composed of what has already been written, carrying vestiges of previous texts reassembled into a new composition. In this sense, a found poem, like an AI-generated text, could be imagined as one instantiation of an evolving, unfolding material process. To compose a found poem using AI is to imbue yet another layer of generativity upon generative AI.

Examples of Students’ Found Poems Crafted Using AI

I introduced this found poem activity in my first-year writing classes during the fall semester. Below, I share two examples of found poems written by my first-year writing students.

Student 1 Found Poem

The first student’s found poem appears below, accompanied by an image description (the ChatGPT texts used to create the poem appear in Appendix 1). The student’s poem juxtaposes ChatGPT-produced writing from across three genres — a recipe, a poem, a story — and repurposes words and phrases from the texts into a collage-like found poem, as shown below:

In her reflection accompanying the poem, the student writes:

The found poem I created basically expresses how a person might view life: “deep” and “curious.” Life is full of secrets and new things to learn. “Followed a narrow path” means navigating through the challenges and limitations in life. “Find purpose” is quite self-explanatory: finding your true purpose in life as a human being. “Heart raced with anticipation” portrays how a person can feel about what life holds for them in the future. “In a world of endless wonder” means being filled with constant curiosity. “So free but forgotten” emphasizes that everyone has their own life and focus.

The poem’s substantive theme mirrors the experience of reading the poem: in playful contradiction with the phrase “followed a narrow path,” the poem beckons the reader into multiple readerly pathways, inviting yet unsettling the reader’s journey through the poem. As words morph and mutate into unexpected forms, the poem exposes the malleability of language in its fleeting instantiations, illustrating the creative possibilities that emerge when texts are remixed and transposed into a new medium. In this visual collage-like poem, words transform into images, reverberating with deeper resonances.

Student 2 Found Poem

While the student’s poem above reassembles words and phrases from ChatGPT-produced texts, a second student’s poem, shared below, incorporates italicized fragments from a ChatGPT-generated poem alongside the student’s own writing, which appears in regular font (the ChatGPT text used to create the poem appears in Appendix 2).

Student’s Found Poem

I always thought it would last forever,

for it was you who made me stronger.

The vivid visions of our future,

slowly but surely became a memory, a delusion we created to stick together.

Watching you lose your care was the most painful thing to bear.

Everytime we talk things over, I see the eclipsing light begin to shine.

But the silence forms a cold empty space as hearts become withdrawn.

A garden of what was once roses, turned into thorns.

I am now grieving yet you are still near, I feel estranged in my own home.

A heart once full now feels so small, as we both take a different road.

When I am alone I think to myself, are we ever going to be repaired, or should I be prepared for the worst fate of all, I am not ready to be alone, years of safety once abode now abandoned and corrode, will it all come to an end or will we actually make amend, the thoughts rushing through my head, is all we once wished for now gone and dead, will we find our way back home, will our next go actually work, or will it cause further corruption, further delusion, further aggravation, where do I go to when I am alone, when the one I awaited is far gone, how do I remain aloof, how do I let things go? All these thoughts in my head… but it’s not about me is it?

A garden of what was once roses, turned into thorns.

In her reflection accompanying the found poem, the student writes:

The poem was inspired by the books I used to read when I was younger. I composed the poem by prompting ChatGPT several times until I got the perfect length with the most captivating words. In the beginning, the sentences are short to show the short term bonds created. The second part shows the overthinking going through the person’s head which is dismissed swifty. The experience of composing a poem using AI was interesting. I like the ease of using AI; however, I prefer my own creativity.

In contrast with the steady meter and simple harmony of the ChatGPT poem, the student’s poem spills outward, its lamentations seeping beyond the confines of the page. In wresting fragments of insight from the chaos of experience, the poem expresses a sense of vulnerability, offering a glimpse into a moment of intimacy fractured by distance. Seeking beyond the surfaces of language, the poem exposes latent depths: its fragmentary form seeks to resist closure, to trace the cascading contours of the consciousness, to tug at the threads of language, to extend thought beyond the seams of the canvas.

Reflections on Students’ Found Poems: Weaving a Tapestry of Language

To write is to weave a tapestry of language, yet to remediate is to reimagine the possibilities of form and meaning. In transfiguring the ChatGPT-produced texts into fragmentary assemblages, the students’ found poems create a striking visual effect: in an interplay of words, images, and shapes, the visual compositions echo, amplify, or distort the verbal elements. In contrast with the stable forms and steady meters of the ChatGPT texts, the students’ poems stimulate a sense of movement, capturing the fleeting intervals of thought and feeling in their gathering intensity. In conveying the cadences and cacophonies of experience, the poems elevate the everyday as life transforms into art.

In reimagining writing as an assemblage of refashioned elements, we might ask: What is lost and what is gained when words are cast into new relations, reassembled into a new poem? While it could be argued that meaning is diminished in AI-generated texts, in the students’ found poems, meaning augments, reverberating with rich resonances. Writers have characterized ChatGPT-generated writing as “flattened” (MLA-CCCC Joint Task Force on Writing and AI), yet the students’ writing reanimates the shape, texture, and dimensionality of language, illustrating the ways meaning can be transformed through creative juxtaposition and remediation. By turns playful, provocative, and poignant, the students’ poems complicate the simple harmony and neat closure of the source texts as words unravel into a dynamic multiplicity of meaning. Through an experimental poetics, the poems seek to unsettle conventional forms, to undo narrative coherence. In counterpoint with the harmonious flow of the ChatGPT texts, the students’ poems create a sense of dissonance: phrases such as “so free / but / forgotten” form discordant elements that interrupt the simple, linear progression of the ChatGPT-produced texts. In subverting the static forms of writing that might characterize AI-generated texts, the students’ poems expose the fragmentary nature of the contemporary moment, rendering experience as fractured yet dynamic, fleeting yet deep. In contrast with the ChatGPT-produced writing, which could be critiqued as impersonal, detached, or lacking in experience, consciousness, and subjectivity, the students’ poems convey a sense of immediacy, the words flowing from an individual, subjective consciousness. In this sense, the students’ poems gesture toward a return to the human element of poetry — to the poem as an expression of intimacy, as a capsule of emotion captured within fleeting intervals, preserved through the permanence of mediation on the page.

From stasis to dynamism, from fixity to fluidity, the students’ poems recast language as incessant fluctuations of form and meaning. This found poem activity thus illuminates the way writing itself is a process — a process of drawing from existing texts, curating knowledge, stitching together threads, (re)fashioning ideas, and awakening resonances. In recasting composing as a continually evolving process, the AI found poems destabilize existing patterns of language, experimenting with the ways language could reveal (or obscure) meaning. In the students’ poems, words spill outward, brimming with surprising revelations. In conceptualizing language as a tapestry, we can envision poetry as an unraveling, as each thread, untangled from the source, is exposed to new light.

Potential Pitfalls: Preventing Patchwork Plagiarism

While the student examples above illustrate the creative possibilities of crafting found poetry using AI, one potential pitfall of this approach is that the activity could promote patchwork plagiarism: students might lift ideas and language from the AI text without reassembling the text in a new way; students might also reproduce the ideas in the AI text without critically engaging with them. The activity thus raises ethical concerns regarding integrity: what counts as an “original” found poem, and what would be considered plagiarism or unethical attribution of sources? For instance, lifting more than a few consecutive words from a source could be considered plagiarism, yet the nature of a found poem invites writers to lift language from other sources and to recombine the words in new ways. In recognizing these tensions, we could take steps to prepare students for the exercise, including by engaging students in reading examples of found poems and considering the ways writers remix or recast others’ words into a piece of their own. We could also introduce strategies for crafting found poems in ways that encourage students to go beyond simply reproducing the source’s ideas. One useful tool is a word shuffler, which randomizes the sequence of a list of words.

In Figure 1 below, the first column shows a list of words as they appear in an essay by ChatGPT, while the second column shows the list shuffled into a random order. In this way, students can be encouraged to reassemble the words in the AI source text and to imagine surprising re-combinations of language and meaning.

Figure 1: Word Shuffler

A related challenge is that AI language innovations might paradoxically suppress inventive, innovative uses of language, including language choices that subvert or transcend convention. Creating found poems using AI could inspire students to become more intentional about their writing: beyond simply copying and pasting AI-generated texts as their own, students are encouraged to think critically about meaning, language, and form. As described in the prompt above, I invite students to experiment with new or surprising meanings, including meanings that might conflict with or go beyond the meaning of the AI-generated text, by posing questions or sharing their own thoughts on the topic in a call and response format. As illustrated in the examples, the students recast the AI texts in an altered light, probing into the spaces between the texts, complicating their coherence while inviting altered ways of seeing in/through a poem. By “speaking back” to an AI-generated text, students critically interrogate and creatively experiment with writing produced by large language models such as ChatGPT.

Stimulating Students’ Critical Reflection on Writerly Choices and Processes

As illustrated in the examples above, to stimulate students’ critical thinking about their writerly choices and processes, I invite students to compose a reflection accompanying their found poems. In offering guidelines for the reflection, I encourage students to consider questions such as: How did you go about composing the poem? Why did you make the creative choices that you made? How was the experience of composing a poem using AI tools? I also offer students an opportunity to share their found poems with their peers in the class.

Juxtaposing the found poems with the AI-generated texts sheds light on the ways students’ found poems alter, remix, or transform the AI-generated texts in an ongoing experimentation with language. Encouraging reflection on the process of composing a found poem could further enhance students’ engagements with transforming language and meaning through remediation. Even so, I acknowledge that my interpretations of the example student-written found poems might differ from or perhaps go beyond the students’ intentions, as evidenced by their reflections. This activity thus opens spaces for cultivating students’ meta-awareness of their own writerly choices as well as for deepening students’ knowledge of the literary and aesthetic theories that undergird their compositions. The exercise could also encourage a closer attention to the composing process: for instance, in one student’s reflection shared above, the student describes the iterative process by which she prompted ChatGPT: “I composed the poem by prompting ChatGPT several times until I got the perfect length with the most captivating words.” To stimulate students’ careful attunement to the ways language shapes meaning in discourse, instructors could offer writing workshops that support students to experiment with different ways of prompting ChatGPT and reassembling words and phrases to create a found poem. The process of crafting a found poem could invite playful experiments with language; as one student writes in her reflection, “The experience was actually fun. It felt like taking random words and playing with them, while creating a meaningful poem.” At the same time, we could acknowledge the ways composing with AI could subvert the potential for meaning or restrict the creative process; as the second student expresses, “I like the ease of using AI; however, I prefer my own creativity.” In problematizing the notion that AI could ameliorate or transform the creative process, we could inspire students to interrogate the ways composing with AI could at once catalyze and constrain creativity, enrich and elide the potential for meaning. By engaging students in reflection alongside creative experimentation with AI language models, we can raise students’ critical consciousness about language as we work together to question writing norms and interrogate models that reproduce existing patterns of language. Crafting experimental poetry using AI could thus counterbalance the perhaps standardizing tendency of AI language models, opening spaces for alternative ways of thinking and writing. By composing found poetry with AI, students transform an AI language tool’s conventional style into innovative forms, fragmenting and reassembling AI-produced texts into surprising reconfigurations. In unsettling stabilized forms of writing, we can push the boundaries of creative expression through remix and remediation. Ultimately, to experiment with language is to interrogate the possibilities and limits of expression as technologies continue to evolve.

Questions to Guide Reflection and Discussion

How does the process of crafting found poetry with AI challenge traditional notions of authorship and creativity?
Discuss the ethical implications of using AI-generated texts as raw material for creating found poetry. What considerations should be taken into account?
Reflect on the potential benefits and limitations of using AI to inspire creative writing. How does this impact the learning process for students?
Consider the educational value of engaging students in critical thinking through the modification and reinterpretation of AI-generated texts.
Explore the idea of patchwork plagiarism in the context of AI and found poetry. How can educators ensure students understand and navigate these ethical boundaries?

References

Found poem. (2023, June 28). Academy of American Poets. https://poets.org/glossary/found-poem

MLA-CCCC Joint Task Force on Writing and AI Working Paper: Overview of the Issues, Statement of Principles, and Recommendations. (2023). MLA/CCCC. https://aiandwriting.hcommons.org/working-paper-1/

Palmeri, J. (2012). Remixing composition: A history of multimodal writing pedagogy. Southern Illinois University Press.

Verse by Verse. Google AI Semantic Experiences. https://sites.research.google/versebyverse/

Wu, K. (2023). “Who’s talking: Dada, machine writing, and the found: RiTa.js for visual artists and writers.” TextGenEd. https://wac.colostate.edu/repository/collections/textgened/creative-explorations/whos-talking/

Appendix 1

ChatGPT texts used to create the first example student-written found poem (words and phrases that appear in the student’s found poem are bolded)

ChatGPT Text 1: Short Story

Once upon a time, in a small town nestled deep within the enchanted forest, a young girl named Emily lived with her grandmother. The town was known for its quaint cottages and mystical creatures that roamed freely among the trees.

Emily was a curious and adventurous girl, always seeking new adventures.

On this particular day, as the sun shone brightly in the sky, Emily woke up with a smile on her face. She decided to explore the forest in search of a legendary treasure that had been rumored for generations-the Jewel of Dreams.

Whispering her plan to her grandmother, Emily set off with excitement shining in her eyes. She followed a narrow dirt path that led her deeper and deeper into the heart of the forest. As she walked, the birds sang melodiously, and the leaves danced with the cool breeze.

After a couple of hours, Emily stumbled upon an ancient stone tablet covered in cryptic symbols. Her heart raced with anticipation as she deciphered the message, which revealed the location she was trying to find.

ChatGPT Text 2: Recipe

Here’s a delicious banana cake recipe for you:

Ingredients:

· 2 cups all-purpose flour

· 1 1/2 teaspoons baking soda

· 1/2 teaspoon salt

· 4 ripe bananas, mashed

· 1/2 cup unsalted butter, softened

· 1 cup granulated sugar

· 1/2 cup brown sugar

· 2 large eggs

· 1 teaspoon vanilla extract

· 1 cup buttermilk

Instructions:

Preheat your oven to 350°F (175°C). Grease and flour a 9×13-inch baking pan.

In a medium bowl, whisk together the flour, baking soda, and salt. Set aside.

In a large mixing bowl, cream together the butter, granulated sugar, and brown sugar until light and fluffy.

Add the eggs, one at a time, beating well after each addition. Stir in the vanilla extract.

Mix in the mashed bananas…

ChatGPT Text 3: Poem

In a world of endless wonder and delight,

Where time dances in the pale moonlight,

A new day dawns with promise and grace,

And I find solace in this sacred space.

The sun rises, spreading its golden hues,

Awakening the earth, banishing the blues.

Birds sing their sweet melodies, so free,

Their songs of joy touching every tree.

Oh but, how the wind whispers, bending the grass,

Whispering secrets of a forgotten past.

Leaves rustle in harmony, a gentle symphony,

Telling stories of seasons, moments of epiphany.

The flowers bloom, painting the world with hues,

Their delicate petals catching the morning dews.

A vibrant tapestry of colors, I adore,

Nature’s artwork, forever wanting more.

As day unfolds, time weaves its mysterious spell,

Each passing hour, a tale to tell.

I watch as moments slip through my hand,

Wishing to cherish them, but they expand.

Appendix 2

ChatGPT text used to create the second example student-written found poem (words and phrases that appear in the student’s found poem are bolded)

ChatGPT Poem

Watching love fade, a painful sight,

As shadows creep, eclipsing light.

A heart once full, now feels so small,

As love’s petals wilt, one by one, they fall.

The laughter shared, now echoes gone,

In silence, we stand, hearts withdrawn.

The warmth we felt, like summer’s embrace,

Now replaced by a cold, empty space.

In every glance, I see the change,

A love once vibrant, now estranged.

I search for answers, reasons why,

But love’s departure, it doesn’t comply.

I hold on tight, refusing to let go,

Hoping love’s embers will start to glow.

But sometimes, love takes a different road,

Leaving us to navigate the untrodden code.

Yet, through the pain, strength will arise,

A new chapter awaits, a different prize.

For love is not limited to just one face,

It’s a journey, evolving, in every new place.

So, I’ll cherish the moments we once shared,

Knowing that love, in time, can be repaired.

And as I heal, I’ll find love anew,

In myself, in others, in skies of blue.

Remember, my friend, you’re not alone,

In this bittersweet journey we all have known.

Hold onto hope, let your heart mend,

For love’s story never truly comes to an end.

1 A quoted phrase from an AI-generated Shakespearean sonnet

2 I have secured IRB approval for sharing students’ AI found poems (IRB-20244), and I have also secured the students’ permission to publish their writing.

About the author

name: Ruth Li

institution: Alfaisal University

Ruth Li is an assistant professor of English at Alfaisal University in Riyadh, Saudi Arabia. Her research and teaching interests include writing pedagogy and assessment, educational linguistics, discourse analysis, and the teaching of literature and creative writing. Her work appears or is forthcoming in College Composition and Communication, Assessing Writing, the Journal of Interactive Technology and Pedagogy, and the Journal of Response to Writing.

License

Icon for the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 International License