Students increasingly use ChatGPT and other large language models (LLMs) to do their homework assignments and assist with class projects. Many professors feel the pressure to modify their pedagogy to incorporate AI use in their courses. However, many of the colleges and universities at which they teach pay little attention to the escalating carbon emissions from burning the fossil fuels necessary to power AI. An LLM query demands nearly ten times as much electricity as a Google search query. The climate crisis concerns all students, of course, but particular ethical questions arise when using AI to teach such subjects as environmental studies, sustainability, green engineering, sustainable business, and climate change literature.
Using LLMs for classroom instruction or to complete assignments contributes to the very planetary damage that students need to learn to mitigate. Training AI models uses a tremendous amount of computing power—and therefore electricity—that typically comes from nonrenewable sources. A Goldman Sachs report anticipates that data center power demands will increase by 160 percent by 2030. Many data centers running generative AI models waste a lot of water, particularly to dissipate heat. Generating images also places high demands on the energy grid. While “the cloud” may sound clean, remote data storage has a high carbon footprint. Google’s recent sustainability report admits that the company’s greenhouse gas emissions rose by 48 percent since 2019, mostly because of data center energy consumption, and that further incorporating AI into its products will contribute to higher emissions. Still other environmental damage likely remains undisclosed.
Many professors who teach about climate change and the environment face an ethical quandary as AI becomes increasingly common. For example, Maya Vinokour, an associate professor in the Department of Russian and Slavic Studies at New York University, is concerned about “the relationship between AI and computing power, water and energy usage,” and about the vast, unsustainable, energy-consuming LLMs that contradict values of environmental stewardship. Stacey Balkan, who teaches environmental literature at Florida Atlantic University, worries that “the digital platforms that many are touting as a silver bullet in the face of imminent climate collapse are fueled in large part by coal and natural gas.”
Brent Bellamy, an English professor at Trent University in Canada, notes that, “as with many environmental impacts, the problems occur at scale.” In other words, environmental issues often become significant or noticeable only when they accumulate or intensify because of large-scale human activities; cutting down one tree might not make a big difference, but deforestation has proven devastating to planetary well-being. Using one ChatGPT query might not be damaging, but the entire energy-intensive system required to generate that query has large-scale ramifications. The University of Oslo in Norway focuses “on digitalization and sustainability, as if they were two separate tracks, but very little on sustainable use of new technology,” says Sissel Furuseth, a professor of Nordic literature. “I have voiced some energy-related skepticism about AI in my own department, but the main reservations my colleagues seem to have concern fake news and the risk of cheating.”
Some faculty members warn their students away from AI use because of its high cost to global ecosystem health. Imre Szeman, who directs the Institute for Environment, Conservation, and Sustainability at the University of Toronto Scarborough, realizes that his students might ignore his warnings about the dangers of AI and intellectual honesty, but “linking it to the future of the planet means a lot to them.” Historian Robert Johnson of National University in La Jolla, California, says, “I tell students not to use it because one of the skills they are learning is writing and thinking.” But for him, the environmental costs present their own concerns: “I think the carbon costs are kind of baked in at this point.”
With students’ rapid and widespread adoption of LLMs, many professors who teach about climate change and the environment have not had a chance to establish classroom policies. Magdalena Mączyńska, who teaches climate literature at Drexel University, says, “I feel compelled to keep abreast of this new technology, poised as it is to change everything—but I also feel compelled to resist it, for the exact reasons of environmental degradation.” Matthew Schneider-Mayerson, an associate professor of English at Rice University, admits, “I haven’t yet proposed any assignments that use LLMs, partially because of energy considerations.”
With AI creep, there is no turning back from students using LLMs; the carbon costs of AI may indeed be baked into higher education instruction. Students want to know how to use LLMs when they graduate, especially as the job market changes around AI use. To move effectively into the workforce, students need training in critical thinking and writing skills and in how to use AI to enhance those skills. Perhaps AI can even be used to help generate solutions to mitigate our high-emissions lifestyle.
Students in my nineteenth-century American literature class metaphorically sail the high seas as we study Moby-Dick. But in my Literature of Climate Change class, we read works that portray the devastating sea-level rise of these same waters. In my experience, students want to use ChatGPT as a co-intelligence to help write their papers. Somehow, it feels lazy but legitimate to use an LLM as a coach in my survey class, but it feels negligent and reckless to use an LLM in a course on climate change.
In Literature of Climate Change, we read novels, stories, and poems that thematize concerns about, and consequences of, climate change. For example, students enjoy Barbara Kingsolver’s novel Flight Behavior, about a young mom in Appalachia who comes to learn how climate change threatens Appalachia’s crops, homes, and monarch butterflies. We read Lydia Millet’s A Children’s Bible, which depicts children’s love of the natural world being threatened by neglectful parents. We read Paolo Bacigalupi’s short story “The Tamarisk Hunter” as we discuss how climate stress and desertification can threaten friendships. My prehealth students particularly connect with Toiya Finley’s “Outer Rims,” because the physicians in the story have tough choices to make as a heat-intensified disease spreads and a climate-amplified storm descends.
Jesuit universities such as mine prize teaching students about social justice. Climate change is a social justice issue in the United States and around the world because more vulnerable communities will suffer the effects of climate change first and hardest. For example, disadvantaged communities tend to live in urbanized areas that experience higher temperatures because they have fewer trees or bodies of water. Burning carbon in developed countries disproportionately harms low-income and low-emissions countries, which then have a harder time recovering from and adapting to the extreme weather. How can my students view their professor as a credible instructor if course instruction methods violate one of the university’s central messages? How can the classroom model for students the values of environmental responsibility and net-zero emissions if professors use AI?
Since my university is a Laudato Si campus that aims to fulfill Pope Francis’s call to journey toward sustainability, our climate education aims should align with the tools used to achieve those aims. We need sustainable pedagogy to teach about climate catastrophe. I wrestle with climate education initiatives that use technology that frustrates campus efforts to decarbonize.
Yet some professors of “green studies” are innovatively making the topic of AI use part of the course material itself. Lisa Ottum, an English professor at Xavier University, says that she will ask her students “to justify their use of ‘carbon-intensive writing’ for particular tasks or do something to reward or incentivize ‘lower-carbon communication.’” She also has ideas for creatively holding students accountable; she says professors “could ask students to sign something similar to an academic honesty pledge where they acknowledge the harmful environmental impacts of any AI they used—or, if it’s true that guilting people doesn’t work: give them some other chance to positively affirm their rejection of AI on the basis of its environmental hazards.” Megan Cole, who teaches English at Victor Valley College, is “building discussions, debates, and critical reading and writing exercises about ChatGPT, labor, and the environment” into a course. “The goal is to get students thinking critically about the full scope of generative AI’s ramifications, and to understand why it might not be such a great idea to use.”
No educator would require students to smoke cigarettes to learn about nicotine’s ill effects on their health, nor would they demand that students text in the car to learn about the dangers of distracted driving. Instructors must think carefully about their use of a carbon-hyper-intensive technology. If the medium is the message, then using generative AI in a course on climate change or sustainability might end up teaching the wrong lesson.
Debra J. Rosenthal is professor of English at John Carroll University.