With A.I., Let’s Keep Students’ Learning Paramount

Eight guiding principles when choosing to use or not use AI in the classroom

The rise of ChatGPT-4 and the use of Artificial Intelligence in the modern classroom is astounding. We’re a little gobsmacked wondering at the evolution of its capacity and use in the years ahead. Today, we’re only scratching the surface of its positive effects on learning and teaching, especially for its practicality and ability to inspire and transcend limitations. This is a transformative time to be a teacher, and I’m here for all of it.

With all the excitement, however, let’s make sure we don’t lose our footing. We are steeped in learning’s call, rallying all efforts around students’ long-term retention of what we present via instructional experiences. In the pursuit of productivity, teaching digital literacy, and empowering students with tools that reflect their current and future lives, we may have tripped on a cornerstone when it comes to AI.

Let’s consider the nature of learning, for example: Is it to gain content knowledge or to use it to create, apply, connect, and improve things in the world? A combination of both? Something else?

Perception, memory, decision-making, language, and cognition all happen with systemic and not-so-systemic (i.e. unpredictable) interactions of millions of neural connections among separately activated portions of our brains. These “circuits” and chance relationships at the moment of input and subsequent processing are constantly evolving based on three elements: 1) a middle/high school student’s life and classroom experiences, 2) current physiological growth, and 3) the significant impact of the endocrine (hormones) system, including the effects of growth in each of the pituitary gland, thyroid, hypothalamus, pineal body, adrenal glands, pancreas, ovaries, and testes.

These neural, physiological, and endocrinal interactions are supremely important in young adolescents whose brains are not yet fully developed, as they remain particularly malleable to frequent activity in each of these areas. Ten- to 15-year-olds experience intense growth in the pre-frontal cortex, the area most associated with executive function, and in brain efficiency, the neural pruning of the brain, which eliminates extraneous neurons and synapses. Their brains have a stunning capacity to adapt and respond to new experiences and learning. In fact, they are unusually open to learning challenges and creativity, functionally “craving” robust, active engagement with the world, and yes, with topics we teach in middle school.

As teachers, let’s grab this opportunity while it lasts! Young adolescents are at the forefront of their own learning, not sitting inertly, waiting for it to be packaged and delivered to their intellectual doorsteps. In middle schools, we don’t, “short circuit” any of this. Here, learning requires vigorous, frequent, and active engagement, not passive observation, and it happens inside students’ skulls, not outside. If someone or something else does the gathering, processing, and learning, then simply presents it to students for them to stack temporarily in their heads and retrieve for a test with limited applications to students’ actual lives and development, they don’t invest in the content, nor do they find it meaningful enough to retain.

This premise is at odds with statements made by educators like Red River College instructor, George Seimens, in Manitoba, Canada, writing for The International Journal of Instructional Technology & Distance Learning,

“Many of the processes previously handled by learning theories (especially in cognitive information processing) can now be off-loaded to, or supported by, technology…Know-how and know-what is being supplemented with know-where (the understanding of where to find knowledge needed)…Learning (defined as actionable knowledge) can reside outside of ourselves (within an organization or a database), is focused on connecting specialized information sets, and the connections that enable us to learn more are more important than our current state of knowing.”  – From, “Connectivism: A Learning Theory for the Digital Age,” Vol 2. No. 1, January 2005

We can’t keep up with all established and evolving knowledge, of course. Students will have to be taught how to access and sort the exponentially increasing knowledge bases that can’t possibly be stored, let alone updated regularly, in our heads. So, key ingredients to student success include knowing how to keep on learning, yes, but also how to connect to larger communities and sources of knowledge, and remaining open to dynamic changes that happen at these intersections.

For middle level teachers, however, our focus is connecting what’s happening in students’ minds right here in our classrooms and throughout the school year with elements outside of the immediate brain. We judge the effectiveness of teaching and learning by capacities and content in students’ minds, not just how they can avail themselves of what’s found outside their minds. In Make it Stick: The Science of Successful Learning (Belknap Press, 2014), Brown, McDaniel, and Roedinger state that learning is, “Acquiring knowledge and skills and having them readily available from memory so you can make sense of future problems and opportunities.”

Hold this thought for a moment and consider another angle here: Look at the labels at the top of some teachers’ gradebook columns: Book Reports, Labs, Writings, Homework, Learning Logs, Participation, Quizzes, Tests, Projects, On-line Modules. These are all tasks and activities. None of these are reports of learning. They are formats used to express learning, but they are not statements of learning themselves unless the specific formats are our actual curriculum. This is for a gradebook, though, which is supposed to report students’ learning regarding academic standards. These teachers confuse the completion of tasks and activities with reports of what students have learned.

Whether or not students do book reports, projects, or on-line modules is completely irrelevant. It’s the learning that students demonstrate that matters, not the vehicles used to convey it. Over the decades, we’ve blurred the lines, thinking that students who’ve completed the projects, homework, and in-class activities learned a lot while those who did very little learned little. Doing things, however, is not the same as learning things.

Enter Artificial Intelligence. Wow, we exclaim when using CHATGPT-4 and similar AI tools, look at how much faster the task is! Look at the time it saves! Look how many of these I am able to create! (And just as great a concern) Look how I needn’t focus on my own creativity – AI did it for me!

AI reigns supreme in a culture in which we emphasize completing tasks and compliance over students’ meaningful learning and versatile applications of content. It makes things easier and faster. This doesn’t mean that using AI cannot be meaningful for students and teachers, of course, but untempered emphasis on generative AI applications may lead us astray. The goal is that students learn and mature, carrying both facets forward. AI might help, but the “heavy lifting” of learning is done by students themselves.

In his recent article, “5 Things About Using AI for Writing That I Wish Enthusiasts Would Remember” (April 15, 2024), college instructor and Tech & Learning‘s senior staff writer, Erik Ofgang, laments,

“[L]ike a character in a Philip K. Dick novel, I regularly have the unsettling experience of suspecting human-generated writing was actually machine-written… Until it happens to you, it can be difficult to fully grasp how unsettling and demoralizing it is to come across AI-generated work in your classes. When you suspect a paper is AI-generated but can’t prove it, you have to spend time grading inauthentic work and pretending this feedback matters to the student. AI submissions can also infect a whole batch of papers, making you wonder unfairly about ones that were actually generated by humans. It’s a new kind of stress that just wasn’t part of the job a year ago.”

Here’s the thing, though, writing is thinking, and whoever is doing the thinking is doing the learning. Does the use of generative AI limit students’ generative learning? Ofgang continues,

“I know that writing about the world helps me understand it. And that when I consider a topic or a question or an argument and take the time to solidify my thoughts on paper, I understand that subject better and more deeply. I know writing doesn’t come as easily to everyone and different people process the world differently, except our students deserve a chance to develop this ability that can serve them in so many ways beyond the scope of what seems to be a short writing assignment. In other words, this conversation around AI and writing isn’t about writing at all, it’s about thinking and making sure we continue to facilitate the type of thinking writing supports.” 

Later, Ofgang reports that currently, “96% of [college] students use ChatGPT for school work, 69% percent use it for help with writing, and 29% percent use it to generate entire papers.” Hmm.

John Warner, author of the forthcoming book, More Than Words: How to Think About Writing in the Age of AI (2025), posted recently, “Generative AI reveals the lack of genuine meaning underneath lots of academic “activities.” It’s a tool for calling out what I’ve called “academic cosplay,” the doing of academic-like work that is not actually rooted in the principles and values we claim for academia.”

Warner is referring to the ceaseless pressure at the university level to publish regardless of originality or value as well as implement activities that express a pseudo-semblance of effective instruction, but the idea of “academic cosplay” resonates with middle and high school teachers, too. For example, are we succumbing to superficial technology use without questioning its actual effectiveness? Are we allowing AI to subsume our own effective instruction without critical examination of its provenance, developmental appropriateness, or outcome?

Some of those excited by AI may need to refocus on students’ long-term retention of content and skills and the growth that comes from such. In multiple books, videos, blogs, articles, and online posts over the years, we see recommendations for using ChatGPT and AI to do things for students that these authors claim teach critical thinking, spark creative ideas, provide tools, and help them work more efficiently. These are all positive things, of course, but the problem is the idea of AI doing things for students: AI is doing it, students are not.

It is by wrestling with these very elements, and the organic dynamic created in the mind during those wrestling moments, that students invest in their own learning and comprehend, stretch, and connect. So, yes, let’s use ChatGPT and AI for some of these things as it improves engagement and learning for students, but let’s also apply a critical lens and consider what actually moves content, skills, and their versatile applications into students’ minds for long term retention.

The list below is a sampling of recommended uses of ChatGPT-4 and AI in classrooms from several dozen published and posted resources over the past few years up through April 2024. These authors recommend we use generative AI to:

  • Draft initial versions of essays
  • Provide comparisons and analyses among pieces of literature and other topics students can use as catalysts for their early writing drafts
  • Identify themes, main ideas, and supportive details in writings and presentations
  • Generate questions to ask about a topic
  • Create scenarios for a situation under study
  • Generate conversation bots that simulate historical figures and people from historical eras or knowledge domains with whom students can converse
  • Summarize knowledge, text, videos, graphics, plays, and experiences
  • Generate creative and synergistic graphics depicting people, principles, and other elements conveying a message
  • Provide feedback to students about their work
  • Provide perspective from another point of view
  • Remix student work into a different medium
  • Create media presentations for students to deliver to classmates
  • Entertain
  • Analyze a character or historical figure’s motivations
  • Create scripts students can act out in class
  • Generate poetry, songs, stories, parodies, and jokes
  • Design better AI prompts
  • For teachers: Design developmentally appropriate instruction, design rubrics, grade student writings, coding, projects, and tests, and write IEP’s, report card comments, and emails to parents

Strikingly, all of these (except the one for teachers) are excellent experiences for students to do themselves. And for the elements listed for teachers, we are empowered and versatile as educators when developing these things for ourselves. I know we don’t have time to do all these things well, but we have to ask for ourselves and our students, what is lost when we leap past them? What changes in the learning dynamic when we ask AI to serve students with the elements already generated, sifted, and sorted? Did we simply hand students a recipe to follow, or did we cultivate the soil within them to do this on their own? Are we powerful in our teaching when handed a recipe to follow, or is it in the genesis of instructional design using deep knowledge of students and our subject content that we become truly competent and effective?

Stravinsky reminds us, “Exposure to blind alleys I have learned throughout my life as a composer chiefly through my mistakes and pursuits of false assumptions, not by my exposure to founts of wisdom and knowledge.”

Real and long-term learning is messy and organic. As middle level educators we embrace this. We know that whoever does the editing of a math problem, paragraph, code, music, presentation, and graphic does the most learning. We know that starting down one path of processing knowledge (writing, summarizing, graphing, building, connecting, for examples) often releases the subconscious, and we make meaningful associations resulting in something much improved over the original idea. Intellectual wiggling, wavering, and jumping tracks are valuable. Handing students a grocery list to memorize doesn’t lead to long-term retention of content, and we know that teaching critical thinking is much more than simply weighing the merits and biases of what AI generates and presents.

We’re better than some of the current uses of AI, and we can’t short-cut robust learning. Let’s use ChatGPT-4 and AI tools in all the wonderful ways we can, but let’s keep alert, following eight guiding principles when choosing to use or not use AI in the classroom:

  1. Do not short-cut anything that is an important thought process for students to experience in order to move content to long-term memory. Never let AI think for a student what he can think for himself. Think in terms of AI as a tool used to build capacity for learning, not do the learning.
  2. Move away from emphasizing and reporting on what students do, and instead, emphasize and report on what they learn, making sure that expressions of student learning show what students know and can do themselves.
  3. Ask students to create the first draft of anything on their own without benefit of AI. They can do research and other gathering of elements via AI and other sources like books, videos, interviews, models/examples, and more in preparation for drafting, of course, but the actual draft is done by the student. Seeing what AI creates and evaluating and finessing it for our own use doesn’t have the instructional impact some think it does. We don’t learn to swim by staring at the water or just by watching and comparing examples of poor and good swimmers; we jump in the water and swim.
  4. Grade your students’ work, don’t outsource it. When we grade students’ expressions of learning ourselves, we enter into advocating relationships with the students, thinking deeply about the most helpful feedback and what their expressions mean for next steps in instruction. As a result, we are more responsive and versatile in the subsequent interactions when working with students. We don’t have this depth or dexterity at our mental fingertips when something else grades the work and simply hands us the report.
  5. Periodically, re-visit what learning really means, how it is achieved, and how it is demonstrated accurately. AI generated products may be more an expression of the kind of prompts used in their genesis, the algorithms of data culled, or access or lack of access to particular generative AI tools than they are expressions of what student know themselves. Yes, learning may mean demonstrating dexterity with AI tools, but it’s just as important to define it in terms of what students know in any one or more domains and what they can generate and connect for themselves.
  6. Be careful when chasing quick learning and time-saving efficiency, as they don’t always result in what’s effective. Learning takes a lot of work and energy, even in an age of AI. We’re so grateful for the many, many efficient tools, resources, syntheses, and instructional elements AI can provide, however, and those things make us more efficient, yes, but we’re making a mistake when our prime focus is taking short cuts to learning. An individual human’s real learning is our goal, and that takes time. Reminder: After COVID, teachers and schools were pushed to do “quick learning” and catch everyone up. We found out quickly that it is ineffective.
  7. Remain vigilant about the misinformation, biases, and stereotypes perpetuated by AI that we and students use. AMLE’s, article, “Five Ethical Considerations to Keep in Mind Before Using AI In Your Classroom,” by Marri Bayour, Jason De Hart, Joe Pizzo, and Megan Vosk is great reminder here.
  8. Be aware of the developmental needs of young adolescents, including the cognitive science ideas that lead to successful learning. How does the use of AI support what we know about middle schoolers’ minds and how they grow? Our students are deeply impressionable, and one real concern is their unusual and sometimes addictive focus on screens and social media. Middle schoolers need to experience significant interactions with one another and adults in person. They need long exposure to the natural world, exercise, and fine and performing arts that is not related to screen work. They need time and structure (and sometimes, not so much structure) to reflect, dream, and invent on their own, free from pixels and algorithms. If there is anything we can facilitate in students’ learning and growth that can be done just as effectively without using AI and screens, lean towards those activities.

This list of guidelines will undoubtedly change and grow as we converse, lean into curiosity, dream big, take constructive risks, and learn more about AI’s instructional and personal uses. In these, we model the very things we want students to do in their own learning lives. There is architecture, however, carved from thoughtful engineering principles by dedicated artisans (middle school educators and researchers) that should inform our efforts. Let’s start with the cornerstones of middle level instruction as we explore the wonderful possibilities of AI and use its tools to improve student learning over that which otherwise could be achieved without it. There’s a lot of learning to be had, a lot of future to build.

Rick Wormeli is a long-time teacher, consultant and author living in Herndon, Virginia. His book, The Collected Writings (So Far) of Rick Wormeli: Crazy, Good Stuff I Learned about Teaching Along the Way, is available from AMLE. His newest books are Fair Isn’t Always Equal: Second Edition, and Summarization in any Subject: 60 Innovative, Tech-Infused Strategies for Deeper Student Learning, 2nd edition, co-authored with Dedra Stafford. He can be reached at rick@rickwormeli.onmicrosoft.com, @rickwormeli2, and at www.rickwormeli.com.