According to Graham, reading and writing are deeply intertwined. You can’t have another one. Graham said the meaning and conveying meaning of reading and writing, both borrowed from some of the same processes in the brain. He studies the development of writing and the effectiveness of digital tools that support K-12 students’ writing.
“Reading and writing are great tools for learning,” Graham said. But when AI tools “think” for students (such as generating large and complex text), some of this learning disappears. Take the outline of the paper as an example. Graham said that an overview of the paper requires thinking about the information, making decisions about what information is included and what is excluded, and organizing that information to make an argument.
“If we get out of our mind, it’s unlikely that we learn a lot and check the depth of the material we write as much as possible,” he said.
According to Graham, revision is an important process in the development of writing and learning. “When we write, new ideas come to us… When we modify, the same thing happens,” he said. “When AI tools are used to bypass certain important steps in writing development, such as essay outlines or revisions, the “struggle” of learning is taken away.
However, according to Graham, there is a correct way to use AI as a “writing partner.” He said when you write, you make some minor tweaks. For example, you might write a sentence and wonder if you need to make a different word choice, or change the punctuation. When you use chatgpt to put forward alternative sentences to already written sentences, Graham continues saying that you have to do a “wholesale evaluation” of the material.
Using AI to assist writing can be a metacognitive practice rather than a time-saving strategy. Rather than adjusting your own writing, consider using AI to generate alternative sentences. But you still need to be able to determine the “best” sentence.
“I think one of the biggest challenges of writing at high school or at any level right now is basically time. Very little time is dedicated to writing,” Graham said.
Unfortunately, on a good study day, time is tight. Having enough time seems impossible when time has to be used to solve other problems in the classroom (such as student indifference and loss of learning).
The pandemic has changed everything
It wasn’t until the 2021-22 school year that Knight noticed changes in his students and technology. “During virtual school, it’s “really useless.” Knight continued, that doesn’t necessarily change the type of technology available to students, but rather a shift in the relationship between students and that technology.
Over the past four academic years, Knight has witnessed a decline in social regulatory skills in his students. He and his colleagues now provide students with five minutes of rest during the 90-minute block, a practice that didn’t exist before the pandemic. Despite his efforts to creatively engage his students in classroom activities, Knight often finds that they quickly blow over what he calls “the best point of social engagement.” A socially attractive learning activity can now quickly turn too much energy into a bad learning environment, Knight said.
His first interaction with new AI technologies in the classroom was negative. Some of the Knights students use Chatgpt to cheat. So Knight decided to start using AI detection software, but posed too far, and he mistakenly accused a student of cheating. The result is a damaged relationship with his students.
During the 2024-25 school year, Knight prefers paper and pencils and does not assign open-ended written response work on laptops or computers. He no longer uses AI detection software.
Schools’ continuous response to AI technology
According to Justin Reich of the MIT Teaching Systems Lab, investment in technologies such as generative AI is not necessarily the only way to improve students’ learning, or even the best way to do so. “Sometimes schools choose technology, sometimes they choose something else,” he said.
Reich said the pandemic forced all K-12 schools, regardless of their previous tech philosophy, to adopt and adapt to a wide range of technological changes at lightning speed. He continued, this accelerated movements and contact before many teachers and students were ready to move forward.
Sometimes, less preparation involves a more difficult road to success, especially during times of less connectivity and social isolation.
It is easy to spiral when considering the possibility and disruption of AI in the classroom. But remember, “In the last century, it was common for people to invent techniques to bypass students’ thinking.” At some point in time, the encyclopedia provided shortcuts for assigned students “summarizing topics based on multiple sources,” he said, “the calculator does the same thing in math classes; a recent example might be Google Translate.”
Reich used these early technological advancements that changed the learning culture when introduced to students and teachers – but are now commonly used – a reminder that “as a field, we have some understanding of processing and management techniques”.
Over the past few years, some schools have experienced high-level students using large language models or generating AI to help them answer homework or paper questions. When these numbers reach a “crisis” level, the pace of the class will accelerate when many students ask a machine to do a lot of work without knowing it.” [students] Learn something, but they just feed him the answer from chatgpt,” Reich said.
Where are we now?
The good news is that when Chatgpt 3.5 was released, the exponential growth of AI forecasts had fallen out. “It’s a good thing for the school,” he continued. Reich said in a conversation and survey of students that, in general, young people know that they (not AI) should do the job. However, most students agree that using AI when they are oppressed, getting stuck in problems or determining that the work they are getting is not valuable.
Reich and his colleagues advise teachers to encourage students to think that AI tools can help a small part of their work, rather than assisting the entire work. “So, if you’re in trouble, don’t ask the machine to do the job. Ask the machine to help you with the next step,” he said.
Ultimately, one solution is not suitable for all solutions. Reich said some schools and teachers, such as Knight, may decide that if they go back to pencils and paper, their learning environment and students are best, while other educational spaces use AI tools and discuss with the students around them.
Rather than treating AI as an efficiency tool, Graham viewed AI as a means of deep learning. “How does it help us do what we want to do without hindering learning and also benefiting a wide range of kids?” Graham said. This may sound like a daunting task, but there are some reasonable ways to implement AI as a tool that benefits both teachers and students in the classroom and promotes learning.
For example, Graham spoke with a teacher who used Chatgpt, who made a writing sample with some of the most common mistakes used by students in the class. The class looked at the “student” examples generated by AI and deepened their understanding of their own writing without awkwardly picking out individual students’ writing.
Graham said that while it has been widely understood that AI detection software is not reliable, AI has proven to be very good at providing feedback. Graham continued that this does not mean that AI is better than humans, but that AI is able to replicate feedback, which is similar to feedback generated by humans and is similar to humans’ feedback. But AI’s feedback provides us with tools we often use when giving teacher feedback, which is doubtful.
