How to dance with the AI devil

December 2, 2024 • Featured • 2 min read

December 2, 2024 Featured 2 min read

Allowing AI in my ELA classroom

By Susan Carney

Teachers are rightly skeptical of Silicon Valley’s announcements. The tech industry’s pattern is to release its innovations into the world, and let society sort out the unintended consequences. Classroom fallout from cellphones and social media included the loss of students’ self-esteem, shortened attention spans, classroom distractions, new avenues to bullying and mental health issues. When I discussed the proposed Congressional ban on TikTok with my ninth grade students last year, one female student declared, “I hope TikTok gets banned because it will save me.”

But the release of generative AI tools has set off unprecedented fear among educators, laying at the schoolhouse doors one of the biggest challenges I’ve seen in 20 years of teaching.

Teachers immediately recognized AI’s potential to help students cheat. In my discipline, English/language arts, colleagues and I worried that we would spend our weekends grading robot papers. We started banning computers in our classrooms and insisting on handwritten essays. We worried that students would lose their authentic voice and the ability to do the hard work of writing. And we worried about many existential issues—an occupational hazard for those who regularly teach dystopian novels.

My panic and fears about generative AI haven’t vanished, but I have shed some of my concerns and even found ways to use it to enhance my English/language arts classroom. I know what you are thinking. Invite AI into the classroom? Isn’t that dancing with the devil? Well, folks, he is already here. So as David Bowie once sang, “Let’s Dance.”

Learning the steps

I attended an NJEA workshop and other professional development programs to learn about generative AI designed for educators. Then I played. AI converted PDFs to PowerPoints, created quizzes in seconds and instantly produced rubrics. The products weren’t perfect and needed adjustments, but I recognized that AI could lighten my workload. AI’s weaknesses showed me that AI can assist, but not replace, teachers. Why couldn’t the same be true for students?

And in fact, it is. When I looked at AI from the student’s perspective and prompted AI to complete some of my classroom assignments, I learned that generative AI could complete the work, but not well; its essays were voiceless and had few details. No students hoping to earn good grades could simply allow AI to do their work. But in AI’s weaknesses, I found a new tool to help me teach students research, writing and critical thinking skills that I previously feared AI would destroy.

Once I knew more, I invited my students into the process of learning with me. Don’t worry about trying this. You are not giving students new ideas; they are already using AI. And do not fear that these “digital natives” will know more than you. They are prolific posters and content creators, but they often don’t understand the risks, dangers and limitations of the technologies they are using.

Together we read articles and discussed AI. Students were amused by stories of attorneys who were disciplined because they submitted AI generated briefs filled with fictional case law to judges. They enjoyed examining AI generated images of the founding fathers created while Google’s Gemini was being trained to be more inclusive. We had productive, interesting discussions of AI’s limitations, hallucinations, biases and flaws. We learned about media literacy, fact-checking and inclusivity—all important topics in the ELA classroom.

Once everyone had dispensed with the myth that AI is flawless magic, we could begin exploring its abilities to enhance our work.

Two stepping through the research process

English classrooms are at the forefront of teaching students to critically think about information, sources, accuracy and diverse viewpoints—indispensable skills in an age of misinformation. It is critical that students learn to use AI to enhance rather than fully perform their research.

My students began their research projects last year learning about generative AI’s limitations as a research tool: AI hallucinates, makes up sources and sometimes does not have current information. I demonstrated many of these flaws live to my students by asking ChatGPT targeted questions.

Using more reliable websites and library databases, my students began their work. They struggled, however, to think of keywords and quickly abandoned rich topics, reporting that there was just no information to be found. This is understandable; their content knowledge and reading were limited. Rather than giving them keywords, I let them ask ChatGPT for some keywords. As my students found and read articles, they acquired additional keywords, and their knowledge grew. AI helped them begin that journey.

Once students had read multiple articles and proven they could conduct research, they laid out their notecards on a desk and organized their ideas, creating their own outlines. Then, they asked AI to generate an outline to compare to the one they created.

One student noted that AI had raised a counterargument he hadn’t considered. As a result, he started searching for more information. Another student noted that the AI generated outline had “a more logical order” than hers so she was rethinking her paper’s organization.

My students achieved the same research, media literacy and writing goals as in the past, but AI augmented their ideas and assisted their planning.

The “write” moves and techniques

With no reliable AI detectors, AI’s essay writing skills are English teachers’ biggest fear. Battles with students and parents about AI-generated work are simply unwinnable. Many colleagues have abandoned computers and insisted on handwritten essays produced in class. They rightly want authentic work from their students because they value the art of writing.

I share their values, but timed in-class essays written on paper don’t yield the best writing from all learners. Good writing is a process that involves multiple drafts and editing, both of which students are not willing to do when writing by hand under time pressure.

So, I invited AI into the process of teaching writing.

I gave students topics and had them prompt AI to generate a short essay on that topic. Then we had a class prompt-a-thon, where student teams prompted AI to give them improved essays. Working in groups, they not only learned to critique writing but also learned basic prompt engineering and the skill of iteration.

This activity was an eye-opener. Students’ prompts surfaced misconceptions about the “rules of writing.” Sometimes students couldn’t articulate the goals of the essay. They also needed guidance about the ethical use of AI in writing and proper citation for it.

One student concluded: “AI didn’t really write a strong essay. It would be faster for me to just write it.”

Teaching tone with AI can also be fun. Students asked AI to write an email to a teacher about an unfair test. They started with an angry tone, then changed to a deferential tone. (This year I’m hoping to try out last summer’s hot TikTok words—demure and mindful, anybody?) As students analyzed the emails, they noted which changes and choices adjusted the tone. We asked: was it successful? How can we improve the output to better reflect the tone? This is writing and rhetorical analysis all in one lesson.

Interpretive dance

AI’s large language models have been trained on many great novels and novel summaries. Ask it about most books assigned in your classrooms. It offers interpretations and summarizes key ideas, much like the Spark Notes my students access online.

The new reality of AI forced me to reevaluate my reading assignments. When students enter my classroom, they are not taking multiple choice reading quizzes—which have never been the best measure of who actually read the book.

Try an AI reading check like a “Character Chat” to test their reading, imagination and creativity. When prompting AI, we ask it to adopt a persona; in this activity we applied that by prompting AI to answer our questions as a character in the book. I reviewed students’ prompts and instructions, looking for evidence of reading. Then students asked AI to answer their questions in the persona of that character. When done, their groups evaluated how well AI captured the character’s personality. This is a creative character analysis.

Students can also demonstrate reading with generative AI art tools. I asked them to use an AI platform, DALL-E, to generate a scene from the book, a book cover or a character. They were assessed on the details, symbols and mood they described in their prompts. Then we did a gallery walk and students discussed their classmates’ AI-generated products.

In a hopeful moment, I found an artistic student sketching her ideas, creating an exemplary rendering of a scene in the book and quietly resisting the idea that art can be made by machine. And guess whose work the class liked best?

I also addressed existential concerns about machines doing our thinking about novels. I asked Copilot, Microsoft’s AI tool, to make a short multiple-choice quiz about a novella assigned to my students. I deliberately asked it to address a part of the story that the author left ambiguous. As expected, AI produced a confident, “one right answer” quiz with an answer key.

I kept a straight face as the students attempted the quiz in groups. They battled over the “correct” answer to that one question, demonstrating that they saw multiple interpretations for that scene. After allowing them to struggle a bit, I explained that the quiz was AI generated. I acknowledged the nuances of the text that made it impossible to answer that quiz question. Then I asked them, “What if I had surrendered all of my thinking about this to AI?”

Our conversation that day reinforced my point that technology, which cannot reason or think, cannot replace human interpretations of literature.

I hope you’ll dance

AI has presented new challenges to teachers, and, as with most technology, it was thrust upon us with no warning, no training and no guardrails. But it is here, and we now have new content to teach: AI Literacy and Prompting Skills.

I still teach old school annotation, close reading and writing. Core foundational skills remain important. But I have found ways to open the classroom door to discuss and use generative AI.

With fuller knowledge, our classes can make AI the devil we know rather than the devil we don’t—and even make him a decent dance partner.

Susan Carney is an English/language arts teacher at River Dell High School in Oradell.