No one has to convince me of AI’s destructive, demoralizing impact on education; as a teacher, I experience daily the countless ways artificial intelligence erodes the cognitive, creative, and persistive abilities of students. With the most understandable of intentions, we teachers implore students not to surrender their thinking to the bots, preaching ethics and honesty, even as we know that “just say no” failed at deterring youth drug abuse and teen sexual activity.
Then, usually, we go right on teaching as if AI didn’t exist, praying students won’t touch the shiny magic button that promises to solve all their problems.
The majority of students do press that button sometimes, even “good kids” who are seriously dedicated to learning but run short on time or confidence. Policing AI use is impossible. In-class writing assessments privilege quick thinking over deep thinking, and no take-home assignment, however creative or personal, is fully “AI-proof.” We can’t design our way around AI; if we truly care about developing students’ critical thinking skills, we must take the digital bull by the horns and teach them AI literacy.
AI literacy cannot just be siloed in computer science courses. My discipline, English, has been one of the biggest victims of AI’s temptations to cognitively outsource, even as the AI age has made the kind of analytic skills we teach ever more important. Ironically, reversing that trend may require humanities classes to fire up Claude and ChatGPT so we can teach students how to critique them.
To teach AI literacy, I invite my high school English students to examine what large language models have to “say” about the novels we’re reading. The students use those canned analyses as counterexamples to the creative, discursive yet defensible readings they themselves generate. They also form counter-theses to disagree with what an LLM says a novel’s themes are, supported by textual evidence they locate. They scour LLM summaries for omissions and inaccuracies. By examining AI, not as a means of producing a product, but as an additional “text” to critique, students both build their analytic faculties and, hopefully, begin to suspect AI’s “expertise.”
We also compare LLM-generated essays with human-generated ones, teasing out AI’s sophisticated-sounding yet generic prose, identifying “fluff” words that sound fancy but say nothing. Students learn to populate their own analytic writing with “funky” devices like anthimeria, synesthesia, anastrophe, and even occasional colloquialisms. In a world where everyone has access to LLMs, I want my students to discover the value of developing genuine voice to stand out.
Sometimes we “invite” LLMs to join our literary discussions. After an initial thrill, students grow frustrated with AI’s repetition, its inability to understand the essence of many of their questions, and particularly its sycophancy. When students begin groaning and asking to shut Claude up, I suspect they’ve learned something special about human interaction.
My students compare their manual note-taking process to the output of AI note-taking tools, paying attention to what gets included or excluded in summaries, what gets emphasized or downplayed. They identify how phrasing choices (e.g., “protestors” vs. “agitators”) create different meanings. As the bots seek to give us what we want, the students develop a habit of asking, “What am I not seeing?”
When we do research, the students evaluate whether LLM searches accurately report their sources, to what extent AI draws from ideologically extreme sites, and how wording a query (e.g., “is abortion safe” vs. “is abortion murder”) often leads to politically slanted results. Students learn how actual source data often differs wildly from AI’s summaries of it. They see the “oracle” has clay feet.
To better understand AI’s social and cultural context, we study the antidemocratic, anti-humanist politics of AI boosters like CEO Sam Altman, venture capitalist Peter Thiel, and philosopher Curtis Yarvin. We feed chatbots “political” questions to track their biases and censorship strategies. We explore computer scientist Joy Buolamwini’s work on algorithmic bias and test for ourselves how AI image generation requests for, say, “doctor,” return mainly white males. Students seriously rethink their impressions of AI tools as objective or simply utilitarian.
Finally, we study human cognition vs. machine “thinking,” how our brains don’t just rely on pattern recognition but also make intuitive leaps. We use Edward De Bono’s lateral thinking activities to practice the kind of divergent idea generation that LLMs are incapable of.
It may seem counterintuitive that to save students’ independent thinking from AI, we may have to engage them with AI, but if we appreciate that students learn via doing, then we cannot simply lecture them about the advantages of puzzling through challenging cognitive tasks themselves.
I have watched my students steadily lose their awe of AI, as well as their sense of it as a “time-saver,” given all the fact-checking and other auditing its flaws demand. Their experience tracks with multiple studies that demonstrate the value of AI literacy. (In one study, AI literacy, more than actual familiarity with literature, better predicted participants’ ability to detect AI authorship of text).
Research suggests that although Generation Alpha’s defaults to utilizing AI, they simultaneously mistrust it. My challenge has expanded beyond encouraging AI skepticism to helping students avoid general cynicism; a student who believes truth is inaccessible is just as imperiled as an overcredulous one.
The Brookings Institution and the American Psychological Association now call for AI literacy education at all levels of schooling, yet many teachers find any classroom use of AI to be anathema. Teaching AI literacy is not necessarily capitulation to the AI companies’ self-declared inevitability, just as teaching fire safety doesn’t mean you’re pro-arson.
I still protest data center construction, vote for candidates who challenge Big Tech, and publicly name AI’s many economic, ecological, and philosophical harms. But students need more than our moral outrage right now, however justified. They need us to help them analyze and critique the AI-infused world that has been forced upon them. They need teachers to do what we do best and what no bot can truly do: teach.
