AI Is Problematic. Why Does PAIRR Use It?

On this page: 

Introduction · What Is AI, Anyway? · Environmental Impacts · Labor Concerns · “Data” and Bias · Privacy, IP & Voice · Acknowledgements · References

Note on the Text: The writers of this statement from PAIRR don’t feel we have resolved these questions once and for all; ongoing discussion and deliberation are crucial. We hope this document spurs further conversation among both students and faculty. We invite you to make a copy of the Google Doc version of this statement for shared commenting in your classroom or learning community.

Introduction

Artificial intelligence (AI) is rapidly changing how people write, learn, communicate, and work. Whether we choose to use AI or not, it is already shaping education, workplaces, and society. Because of this reality, our goal with PAIRR is to help you become a critical, ethical, and informed user of AI. For the PAIRR project, we believe AI feedback can potentially help students revise their writing, because AI has access to strategies drawn from writing teachers. However, like many of you reading this, we have serious concerns about AI. We worry about environmental impacts, job loss, data bias, and intellectual property rights. In fact, in the field of Writing Studies, some people advocate for not using AI at all.

In this statement, we’ll explain why, despite the challenges and concerns surrounding generative AI, we are encouraging you to learn about and work with AI responsibly through PAIRR. PAIRR is a writing support and AI literacy program developed by writing instructors and researchers. Because many students use AI in ways that harm learning, we wanted to guide you toward uses that can potentially support your learning. PAIRR aims to help you:

• Learn more about today’s AI, including ethical concerns and risks
• Practice a more skeptical and empowered way to interact with AI
• Use AI to support your writing and revision skills
• Strengthen confidence in your own voice, ideas, and judgement

If you participate in PAIRR, you will use MyEssayFeedback (MEF), a not-for-profit AI tool designed by educators, with data-privacy protections. After peer review, you will use MEF to get feedback on your writing. Many students have found PAIRR helpful for developing confidence in how to use AI ethically. As one student put it, “I feel a lot more comfortable using [AI] in an ethical way to support my writing.”

What Is AI, Anyway?

Artificial Intelligence (AI) is a problematic term that we are pretty much stuck with. Kate Crawford argues that AI is neither truly artificial nor intelligent because it depends on human expertise for its development and use. Others have proposed terms such as “collective intelligence” and “collective labor”. “AI” in this statement refers primarily to chatbots, agents, and other applications powered by Large Language Models (LLMs)—systems that learn patterns from human-produced text to generate plausible, human-like sophisticated language.

Environmental Impacts

We are especially concerned about AI’s potential for environmental damage. Big tech companies’ race to be first in AI has meant a “tsunami” of data center development. You may have heard that data centers require vast amounts of energy and water. They also require large amounts of land and mined resources, and generate e-waste (Aczel et al. 2026; WVU Libraries, 2026). Furthermore, as with climate change, AI harms often fall hardest on the people who benefit the least.

Data centers aren’t new. For decades, they have been powering much of the internet, including streaming services like Netflix, social media platforms, online gaming, tailored advertising, and crypto mining. Since 2020, however, commercial pressure to scale AI has dramatically accelerated data center development. In 2025, AI accounted for 20% of data center electricity use, but by 2030, 40-50% of electricity going to data centers will likely be used for AI. The International Energy Agency reports that by the end of this decade, the United States is “set to consume more electricity for data centres than for the production of aluminium, steel, cement, chemicals and all other energy-intensive goods combined.”

Without oversight and stewardship, AI could lead to catastrophic environmental harm. A recent United Nations University report describes the alarming threats AI poses to water, land and climate change. Yet the researchers do not argue against AI development; rather, they call for urgent action to ensure the technology is developed “within planetary limits.” They “outline a framework for a ‘responsible AI ecosystem’” built on industry transparency, efficiency by design, equity, lifecycle responsibility, global cooperation and sustainable use.”

Likewise, Sasha Luccioni, a computer scientist and leading voice in environmentally responsible AI, argues that policy makers and developers must prioritize sustainability in regulation and product design. For example, she and co-authors argue for “green policy mechanisms” –akin to FDA nutrition labels– that would require companies to disclose their energy consumption. On her Sustainable AI Group website, Luccioni argues, “AI can be sustainable. But it’s not on that path yet.”

Some analysts argue that, despite its energy costs, AI could have a positive environmental impact by optimizing energy efficiency across a range of sectors, like oil and gas, electricity systems, and transportation. For example, according to the Grantham Research Institute on Climate Change and the Environment, by 2035, advances in AI “could reduce emissions by 3.2 to 5.4 billion tonnes of carbon-dioxide-equivalent annually.” These reductions, they assert, “‘would outweigh increases from global power consumption of data centres and AI”. However, such benefits are uncertain and could be outweighed by increased fossil fuel production and emissions. Crucially, as Jon Ippolito points out, we don’t need large generative AI models–or new data centers–to improve power efficiency. Smaller AI models trained for specific purposes have driven most of the successful power grid optimization (personal communication, 2026).

Our individual choices about AI use can make a difference. Jegham et al. explain that “as AI becomes cheaper and faster, total usage expands, intensifying environmental strain.” According to a United Nations report, by 2025, 700 million people used ChatGPT for 18 billion interactions weekly. But different kinds of interactions have very different impacts. How much we use AI and how we use it matter. Creating a video with AI, for example, uses far more energy than a text response. Jon Ippolito’s “What uses more” app allows us to compare estimated energy use across technologies. The results may surprise you–according to his estimate, a ten-person Zoom meeting uses about seven times as much energy as generating a paragraph of AI text.¹ To estimate your total energy and water use, try Joel Gladd’s Your Digital Life. However, Ippolito and Luccioni also emphasize that AI companies’ lack of transparency makes it difficult to estimate AI’s energy impacts precisely. Nevertheless, these calculations can provide perspective on the energy required by a range of technologies, and help us make decisions about our use. And in a hopeful sign, the recent E.U. A.I. Act’s “transparency obligations” are already impacting the AI industry.

While using AI in college may help you study for a chemistry test, you might choose to skip AI for simple things like recipes, shopping lists, and generation of unnecessary images or videos. Unfortunately, it doesn’t always seem like we have a choice about using AI. As you’ve probably noticed, AI tools are embedded in many of our daily digital interactions, like Google searches. You can opt out by including “-ai” at the end of your search instructions. Luccioni points out that you can also choose search engines with no-AI options such as DuckDuckGo and Startpage. When you do use AI, consider using greener prompting practices and smaller language models, which require less energy (UNESCO, 2025; Economist, 2025). Many chatbots let you choose which underlying model to use. Choosing a model marked “fast” or “efficient” will often mean the model uses less energy. There are also search engines that offset their carbon footprint by planting trees, like Ecosia. As consumers, we can help drive markets by choosing sustainable and energy-transparent products, such as electric vehicles and solar-powered houses.

But consumer choices are not enough. Just as markets and consumer choices alone can’t ensure clean air, water, safe food, or drugs, they won’t ensure sustainable AI. Governments must regulate industries, and as voters, we can push for greener policies. This has worked in the past; for example, tougher emission standards in the European Union (called the “Brussels Effect”) and in California have influenced automakers and other nations’ environmental standards.

For our part, we feel that the relatively modest energy costs of PAIRR text-only interaction are worthwhile given the great need for both writing support and AI literacy. One of the key pillars of the UNESCO AI Competency Framework for Students is “encouraging environmentally sustainable AI”. We hope that PAIRR plays a role in encouraging students to consider AI impacts alongside other digital impacts. We also hope this discussion stimulates you to learn more and to get involved in democratic efforts to maximize AI’s environmental benefits and minimize its harms.

Labor Concerns

Will AI take our jobs?

Many are worried that AI will lead to mass unemployment. Anthropic CEO Dario Amodei recently suggested we could lose 50% of all white-collar jobs. In July 2026, 200 economists and tech leaders, including 15 Nobel Prize Laureates, made a statement warning that AI “could bring risks, including large-scale job displacement.” Students are wondering about their futures. As instructors, we wonder about this too. Will schools and colleges try to replace teachers with AI?

We just don’t know how this will play out. According to a March 2026 Brookings Institution report, “the important questions about AI’s effects on the labor market are still unanswered.” While some recent studies suggest AI is leading to job loss for entry-level jobs right now, others dispute this finding. Today’s patterns in the labor market might not predict tomorrow’s. No study can tell us what future AI tools will be able to do, or how humans will choose to regulate and shape the labor market.

The value of writing skills in AI-integrated workplaces

Uncertainty about future job markets makes one thing clear: educators should help prepare students for AI in the workplace. But how? If equity gaps in writing skills and AI literacy widen, less privileged students might feel unprepared to handle AI. PAIRR aims to support students to improve their writing skills, and learn to approach AI skeptically in the writing process, so that we can evaluate and improve on AI output and maintain our own agency. Because we use words to interact with AI, tech leaders emphasize that future workers need reading, writing, and speaking skills (especially editing skills, as The Economist recently noted) and critical thinking, complex problem solving and adaptability. PAIRR also emphasizes best practices in writing instruction like drafting, peer review, revision, and critical reflection on feedback from peers and AI–to help students develop these urgently needed skills.

Data workers’ rights

Another AI labor concern is the invisible “ghost work,” such as data labeling, which is often done by underpaid people who don’t get much credit or have much control over their working conditions. If you’re ever impressed that a chatbot or AI agent doesn’t spew outright hate speech, you have human ghost workers to thank. Journalists like Madhumita Murgia continue to draw attention to the experiences of these workers. We believe that wages and conditions can improve for data work, as it did for factory work, if data workers organize, and if more privileged tech workers unite with and support less privileged ones. See the AI Resist List for projects you can support or consider as models.

Oversight can help

We argue that the uncertainty about AI’s impacts on jobs, and serious labor concerns, point to the need for informed democratic oversight of AI. As Murgia puts it, “AI can do harm when people don’t have a voice.” We need to focus on how human institutions can shape AI’s labor impacts. One way is through government policies to establish incentives and guardrails, as economists have argued. Another way is to ensure that professionals in each field guide decisions about how AI is used and what it is allowed to replace, as Frank Pasquale argues in New Laws of Robotics: Defending Human Expertise in the Age of AI.

We strongly believe that writing teachers should guide how students use AI, so they can learn how to use AI responsibly and ethically. An important goal for PAIRR is to help students better understand AI’s capabilities and risks, so they can participate in democratic and professional efforts to steer the future of work.

“Data” and Bias

Another serious concern about Large Language Models is their built-in bias. AI bias is stubbornly persistent. To understand why, we need to think about the so-called “data” AI is trained on, and where that “data” comes from.

“Data” sounds like a neutral term that might make you think of numbers. The word “data” doesn’t make us wonder: Who generated this knowledge? Who wrote the words? Human knowledge captured as “data” is often tangled up with human flaws. Consider medical research: clinical studies that primarily used male subjects, a common practice for decades, produced valid scientific findings. But these findings often do not apply to biologically female individuals; the research design and resulting data reflected the gender biases of the medical researchers. To uncover the problems with thinking about “data” as neutral, we need to dig a little–because biases can be hard to see until they are pointed out.

To start with, the publicly available “data” AI was trained on–including about writing standards–is biased in favor of the beliefs, values and languages of dominant groups. Why? Most AI models were trained on the same datasets from the World Wide Web—especially Common Crawl, which was created as a free repository of information. Most of this “data” comes from people in wealthy countries who have the time, money, access and resources to be well-represented online. Languages and communities represented mostly in speech or non-digitized writing, or lacking online access, are often misrepresented or left out altogether. To make matters worse, the World Wide Web hosts a lot of toxic writing–advocating racist, misogynist, homophobic violence, for example. As mentioned in the section on AI and Jobs, AI companies have hired and underpaid human “ghost workers” to help train chatbots not to generate toxic speech. But AI bias cannot simply be removed through training. Why?

Since AI reflects patterns from its “data,” it will continue to default toward stereotypical perspectives. Imagine asking 100 people to describe “a respectable family” at dinner. You’d get a lot of different answers, right? Some answers might represent your own family. But if you ask AI today to generate an image of a “respectable family at the dinner table,” it defaults to a white, heterosexual nuclear family: a man at the head of the table, with a wife and just two kids (“Gemini Flash 3.5, Gamberg, July 11, 2026”). A single image like this might not seem like a big deal. But when these patterns repeat across countless AI-generated images and texts, they reinforce, amplify and even normalize ethnic, gender, class and other forms of bias and stereotypes.

Researchers seem to agree that technical approaches to reducing bias will never fully succeed. Consider an analysis of 5100 AI-generated images of human faces:

From Bloomberg’s analysis of 5,100 AI-generated images (Nicoletti & Bass, 2023).

After gasping, you might take a breath and think—that was so 2023! Well, a UN analysis of 133 AI models in 2026 found that 44% of them continued to exhibit gender bias, and 26% exhibited both racial and gender bias. AI-powered facial recognition technology has also led to the wrongful arrest of innocent people, particularly people of color.

AI bias can sometimes be harder to identify in AI-generated text. Data that contains linguistic biases can profoundly shape how we communicate, learn, and understand the world. An example of AI’s “textual bias” comes from PAIRR researcher and linguist Sophia Minnillo, who, in 2024, asked Microsoft Copilot when to use “ain’t” in a sentence. The response? “Ain’t’ … is best reserved for creating dialogue for uneducated or careless speakers.” Since “ain’t” is associated with non-standard English in many instances on the Web, AI categorized it as “uneducated or careless.”

Yet, consider the use of “ain’t” by U.S. President Barack Obama. In “Remarks by the President on Preparing for College,” he condemned the high cost of college. “That ain’t right,” he said, to a crowd of applauding students. “Not only is it not right, but it also ain’t right.” Obama’s voice is his own. He has the confidence to use “ain’t” however he wants–including to college students.

Would AI encourage each of us to do the same? Or, amplifying the biases it was trained on, would it actually encourage us–ever so politely–to build a wall between our academic writing and our authentic voices, making us doubt our own judgement, so that we all sound bland, impersonal, even robotic? This is a serious risk to consider, and it’s one that PAIRR tackles head-on.

If you ask a chatbot in 2026 when to use “ain’t,” the chatbot might give some positive examples as well as guidance on when to avoid “ain’t”. This might sound reassuring, but fixing individual bias problems does not address the fundamental linguistic bias built into the models and the algorithms on which they are based. AI output will always exhibit as much bias as its developers and its users feed it and accept.

Your Data Privacy, Your Intellectual Property, Your Voice

Another problem with AI’s approach to training “data” is a big one: “datafication.” This can be defined as “the quantification of human life through digital information, very often for economic value.” Transforming vast amounts of human writing, images, computer code, etc. into data–often without permission, for profit–as Big Tech has done with free repositories like Common Crawl, is an instance of “datafication.” Data is big money for AI.

After exploiting publicly available “data,” tech companies now sell AI to the world—asking for money, or more of our data, in exchange. OpenAI is the subject of numerous lawsuits for its use of copyrighted material. In a $1.5 billion settlement, Anthropic paid thousands of authors for using their copyrighted material in training the chatbot Claude.

Increasing students’ AI literacy, which includes learning about such ethical problems in AI’s development, is an important part of PAIRR. One of the best explanations of those problems is the paper “On the Dangers of Stochastic Parrots: Can Language Models Be Too Big?” (2021), which has been cited more than 15,000 times! The authors urge researchers to document and credit the data used to build them, and to weigh the environmental and financial costs of ever-larger models. Big Tech has continued to “cut corners to harvest [training] data” whenever it can get away with it. PAIRR, however, protects your data privacy and never uses your writing as training data.

Data privacy refers to your right to control access to information related to your identity and any sensitive personal information. When you write something original, the result is your intellectual property (IP)—which can sometimes have monetary value! For instance, if you publish an article or a book, you or the publisher owns the copyright. If someone wants to reproduce it for commercial use, normally they have to pay a fee.

Framing human work as “data” minimizes the tremendous time, creativity, labor and resources necessary for humans to (learn to) write, develop and express their ideas in writing (or do math, write computer code, draw, compose music, etc.). It’s helpful to think of “training data,” for the Large Language Models behind AI, as “human expertise captured in writing.” Chatbots and AI agents can simulate human-like text because they were trained on human writing. They depend on powerful computer chips (GPUs, or Graphics Processing Units), complex probability models and deep machine learning techniques inspired by the human brain’s neural networks. But AI doesn’t have a brain–you do. Writing a challenging assignment is a workout for your brain. And the output of your brain–including your writing–rightfully belongs to you. PAIRR can help you learn how to avoid letting AI use your IP without your permission.

Here’s how PAIRR aims to protect your data privacy and IP, preserve and develop your authentic voice, identity and culture in your writing, and strengthen your creative, critical thinking:

1. PAIRR does not use or profit from your IP, and will never allow it to be used as training “data.” The PAIRR project currently uses MyEssayFeedback (MEF), a not-for-profit tool developed by educators. However, you can also use the PAIRR prompt and protocol with any AI platform, and we provide guidance on protecting your privacy outside of MEF. You can also use the PAIRR protocol outside of this class through Playlab, a nonprofit with strong ethical guardrails and policies, with our PAIRR AI Feedback bot.
2. Participating in PAIRR can help you learn to recognize, understand and resist bias in AI-generated material. You can choose to push back, and build your own bias filters. This starts with learning about linguistic justice and equity in your class.
3. We believe that your writing is meaningful. To support your writing development, we have built and tested prompts for feedback in PAIRR that guide AI to respect and protect your voice and your perspectives as much as possible. These feedback prompts shape feedback on your drafts for specific assignments, provided by your instructor. They are designed to help you get your ideas and reasoning across, make your own voice clearer and more persuasive, and help you understand your audience, purpose, and your own writing process better.
4. With PAIRR, you will always critically reflect on peer and AI feedback—not simply accept it—in light of your own purpose, audience and developing voice. This gives you practice critically assessing AI output, a useful skill for any workplace that uses AI.

Thank you for reading this statement, for considering participating in the PAIRR project. We welcome your feedback and questions.

Acknowledgements

Thank you to Jon Ippolito for his extensive feedback on the environmental concerns section. We appreciate your time and insight! Thank you also to the students in Marit Macarthur’s Business Writing course at University of California, Davis who reviewed the draft in summer 2026.

¹ Exact energy calculations are difficult, primarily because tech companies are not transparent in the information they provide. This estimate does not include the energy consumed in training a model or the water used. ↩

References

Aczel, M., Chamanara, S., Matin, M., Farsi, A., Marwala, T., & Madani, K. (2026). Environmental cost of AI’s energy use: Carbon, water and land footprints. United Nations University Institute for Water, Environment and Health. https://doi.org/10.53328/INR26RMA002

AI Resist List. (n.d.). Advocate for the rights of workers. Retrieved July 3, 2026, from https://airesistlist.org/#pillar-labor

Anthropic copyright class action settlement. (n.d.). https://www.anthropiccopyrightsettlement.com/

Baack, S. (2024). A critical analysis of the largest source for generative AI training data: Common Crawl. In Proceedings of the 2024 ACM Conference on Fairness, Accountability, and Transparency (pp. 2199–2208). Association for Computing Machinery. https://doi.org/10.1145/3630106.3659033

Baker-Bell, A. (2020). Linguistic justice: Black language, literacy, identity, and pedagogy. Routledge. https://doi.org/10.4324/9781315147383

BakerHostetler. (n.d.). Case tracker: Artificial intelligence, copyrights and class actions. Retrieved August 21, 2026, from https://www.bakerlaw.com/services/artificial-intelligence-ai/case-tracker-artificial-intelligence-copyrights-and-class-actions/

Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the dangers of stochastic parrots: Can language models be too big? In Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (pp. 610–623). Association for Computing Machinery. https://doi.org/10.1145/3442188.3445922

Borenstein, S. (2026, June 24). AI is an energy and water hog, here’s what you can do to counter that. AP News. https://apnews.com/article/ai-energy-water-climate-change-pollution-environment-77c9de6f9c6326c12d8d18e20dc99c44

Bradford, A. (2019). The Brussels effect: How the European Union rules the world. Oxford University Press. https://doi.org/10.1093/oso/9780190088583.001.0001

Brickner-Wood, B. (2026, August 19). The mark of the machine. The New Yorker. https://www.newyorker.com/culture/infinite-scroll/the-mark-of-the-machine

Brockman, G., Sutskever, I., & OpenAI. (2015, December 11). Introducing OpenAI. OpenAI. https://openai.com/index/introducing-openai/

Conference on College Composition and Communication. (1974). Students’ right to their own language [Position statement]. National Council of Teachers of English. https://cccc.ncte.org/cccc/resources/positions/srtolsummary

Conference on College Composition and Communication. (2020). This ain’t another statement! This is a DEMAND for Black Linguistic Justice! National Council of Teachers of English. https://cccc.ncte.org/cccc/demand-for-black-linguistic-justice

Crawford, K. (2021). Atlas of AI: Power, politics, and the planetary costs of artificial intelligence. Yale University Press.

DuckDuckGo. (n.d.). DuckDuckGo AI-free search. Retrieved August 21, 2026, from https://noai.duckduckgo.com/

The Economist. (2025, September 8). Faith in God-like large language models is waning. https://www.economist.com/business/2025/09/08/faith-in-god-like-large-language-models-is-waning

The Economist. (2026, July 30). AI is getting better at writing. Humans must get better at editing [Leader]. https://www.economist.com/leaders/2026/07/30/ai-is-getting-better-at-writing-humans-must-get-better-at-editing

Ecosia. (n.d.). Ecosia: The search engine that plants trees. Retrieved August 21, 2026, from https://www.ecosia.org/

Elias, J. (2022, December 13). Google execs warn of reputational risk with ChatGPT-like tool. CNBC. https://www.cnbc.com/2022/12/13/google-execs-warn-of-reputational-risk-with-chatgpt-like-tool.html

Enabled Emissions Campaign. (n.d.). Publications. Retrieved August 21, 2026, from https://www.enabledemissions.com/publications

Fernandes, M., McIntyre, M., & Sano-Franchini, J. (2024). Refusing generative AI in writing studies. https://refusal.blog/

Financial Times Visual Storytelling Team, & Murgia, M. (2023, September 12). Generative AI exists because of the transformer: This is how it works. Financial Times. https://ig.ft.com/generative-ai/

Fleisig, E., Smith, G., Bossi, M., Rustagi, I., Yin, X., & Klein, D. (2024). Linguistic bias in ChatGPT: Language models reinforce dialect discrimination. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing (pp. 13541–13564). Association for Computational Linguistics. https://aclanthology.org/2024.emnlp-main.750/

Flores, N., & Rosa, J. (2015). Undoing appropriateness: Raciolinguistic ideologies and language diversity in education. Harvard Educational Review, 85(2), 149–171. https://doi.org/10.17763/0017-8055.85.2.149

Furze, L. (2023, May 22). Teaching AI ethics: Human labour. Leon Furze. https://leonfurze.com/2023/05/22/teaching-ai-ethics-human-labour/

Gee, J. P. (1989). Literacy, discourse, and linguistics: Introduction. Journal of Education, 171(1), 5–17. https://doi.org/10.1177/002205748917100101

Gimbel, M., Kendall, J., & Kulsakdinun, R. (2026, February 19). Labor market AI exposure: What do we know? The Budget Lab at Yale. https://budgetlab.yale.edu/research/labor-market-ai-exposure-what-do-we-know

Gladd, J. (n.d.). Your digital life. Retrieved August 21, 2026, from https://your-digital-life.org/

Gmyrek, P., Berg, J., Kamiński, K., Konopczyński, F., Ładna, A., Nafradi, B., Rosłaniec, K., & Troszyński, M. (2025). Generative AI and jobs: A refined global index of occupational exposure (ILO Working Paper No. 140). International Labour Office. https://doi.org/10.54394/HETP0387

Grantham Research Institute on Climate Change and the Environment. (2025, June 23). New study finds AI could reduce global emissions annually by 3.2 to 5.4 billion tonnes of carbon-dioxide-equivalent by 2035 [Press release]. London School of Economics and Political Science. https://www.lse.ac.uk/granthaminstitute/news/new-study-finds-ai-could-reduce-global-emissions-annually-by-3-2-to-5-4-billion-tonnes-of-carbon-dioxide-equivalent-by-2035/

Hao, K. (2020, December 4). We read the paper that forced Timnit Gebru out of Google. Here’s what it says. MIT Technology Review. https://www.technologyreview.com/2020/12/04/1013294/google-ai-ethics-research-paper-forced-out-timnit-gebru/

Herndon, H. [@hollyherndon]. (2018, November 29). AI is a deceptive … term. CI (collective intelligence) is more useful [Post]. X. https://x.com/hollyherndon/status/1068229047707152384

Hofmann, V., Kalluri, P. R., Jurafsky, D., & King, S. (2024). AI generates covertly racist decisions about people based on their dialect. Nature, 633(8028), 147–154. https://doi.org/10.1038/s41586-024-07856-5

International Energy Agency. (2025). Energy and AI [Executive summary]. https://www.iea.org/reports/energy-and-ai/executive-summary

Ippolito, J. (2025). What uses more? Learning With AI. https://what-uses-more.com

Jegham, N., Abdelatti, M., Koh, C. Y., Elmoubarki, L., & Hendawi, A. (2025). How hungry is AI? Benchmarking energy, water, and carbon footprint of LLM inference. arXiv. https://arxiv.org/html/2505.09598v6

Jha, A. (Host), & Hern, A. (Host). (2025, April 9). Power play: Will AI help or harm the climate? [Audio podcast episode]. In Babbage. The Economist. https://www.economist.com/podcasts/2025/04/09/will-ai-help-or-harm-the-climate

Johnson, J. (2025, February 22). Small language models (SLM): A comprehensive overview. Hugging Face. https://huggingface.co/blog/jjokah/small-language-model

Kean, E. (n.d.). MyEssayFeedback [AI writing-feedback tool]. Retrieved August 3, 2026, from https://myessayfeedback.ai/

Kilpatrick, L. [@OfficialLoganK]. (2023, December 27). Hot take: Many believe prompt engineering is a skill one must learn to be competitive in the future [Post]. X. https://x.com/OfficialLoganK/status/1740099060357374356

Klein, J. (2019, May 30). Fighting the gender stereotypes that warp biomedical research. The New York Times. https://www.nytimes.com/2019/05/30/health/gender-stereotypes-research.html

Kolko, J. (2026, March 10). Research on AI and the labor market is still in the first inning. The Brookings Institution. https://www.brookings.edu/articles/research-on-ai-and-the-labor-market-is-still-in-the-first-inning/

Kruppa, M., & Schechner, S. (2023, February 3). Google stalled release of chatbot for years. The Wall Street Journal.

Lahart, J. (2025, August 26). There is now clearer evidence AI is wrecking young Americans’ job prospects. The Wall Street Journal. https://www.wsj.com/economy/jobs/ai-entry-level-job-impact-5c687c84

Leopold, T. (2025, January 8). Future of Jobs Report 2025: The jobs of the future – and the skills you need to get them. World Economic Forum. https://www.weforum.org/stories/2025/01/future-of-jobs-report-2025-jobs-of-the-future-and-the-skills-you-need-to-get-them/

Luccioni, S. (2025, September 24). We’re doing AI all wrong. Here’s how to get it right [Video]. TED Countdown. https://www.youtube.com/watch?v=Bl-vPf_IAoA

Luccioni, S., Gamazaychikov, B., Hooker, S., Pierrard, R., Strubell, E., Jernite, Y., & Wu, C.-J. (2024). Light bulbs have energy ratings — so why can’t AI chatbots? Nature, 632(8026), 736–738. https://doi.org/10.1038/d41586-024-02680-3

MacArthur, M. (2025). Large language models and the problem of rhetorical debt. AI & Society, 40(8), 6425–6438. https://doi.org/10.1007/s00146-025-02403-w

MacArthur, M. J. (2026). Commercial generative AI is a good idea for teaching writing! Schools should insist that teachers lead the development of AI educational technologies. In C. Basgier, A. Mills, M. Olejnik, M. Rodak, & S. Sharma (Eds.), Bad ideas about AI and writing: Generative practices for teaching, learning, and communication. The WAC Clearinghouse; University Press of Colorado. https://doi.org/10.37514/PER-B.2026.2777.2.25

MacKenzie, D. (2025, November 20). The future of search: Will we still google it? London Review of Books, 47(21). https://www.lrb.co.uk/the-paper/v47/n21/donald-mackenzie/the-future-of-search

Mejias, U. A., & Couldry, N. (2019). Datafication. Internet Policy Review, 8(4). https://doi.org/10.14763/2019.4.1428

Menéndez, A. (2015, November 19). Are we different people in different languages? Literary Hub. https://lithub.com/are-we-different-people-in-different-languages/

Metz, C. (2021, February 19). A second Google A.I. researcher says the company fired her. The New York Times. https://www.nytimes.com/2021/02/19/technology/google-ethical-artificial-intelligence-team.html

Metz, C., Kang, C., Frenkel, S., Thompson, S. A., & Grant, N. (2024, April 6). How tech giants cut corners to harvest data for A.I. The New York Times. https://www.nytimes.com/2024/04/06/technology/tech-giants-harvest-data-artificial-intelligence.html

Metz, C., Weise, K., Hernandez, M., Isaac, M., & Singhvi, A. (2025, March 16). How A.I. is changing the way the world builds computers. The New York Times. https://www.nytimes.com/interactive/2025/03/16/technology/ai-data-centers.html

Murgia, M. (2024). Code dependent: Living in the shadow of AI. Henry Holt.

Nicoletti, L., & Bass, D. (2023, June 9). Humans are biased. Generative AI is even worse. Bloomberg. https://www.bloomberg.com/graphics/2023-generative-ai-bias/

Noble, S. U. (2018). Algorithms of oppression: How search engines reinforce racism. New York University Press. https://nyupress.org/9781479837243/algorithms-of-oppression/

NPR. (2022, September 9). The impact of California’s environmental regulations ripples across the U.S. https://www.npr.org/2022/09/09/1121952184/the-impact-of-californias-environmental-regulations-ripples-across-the-u-s

Obama, B. (2014, March 7). Remarks by the President on preparing for college. The White House, Office of the Press Secretary. https://obamawhitehouse.archives.gov/the-press-office/2014/03/07/remarks-president-preparing-college

O’Donnell, J., & Crownhart, C. (2025, May 20). We did the math on AI’s energy footprint. Here’s the story you haven’t heard. MIT Technology Review. https://www.technologyreview.com/2025/05/20/1116327/ai-energy-usage-climate-footprint-big-tech/

PAIRR Project. (n.d.). PAIRR AI Feedback bot [AI application]. Playlab. Retrieved August 21, 2026, from https://www.playlab.ai/project/cmrwitvmb0flble0wlah9cc7h

Park, S. (2024). AI chatbots and linguistic injustice. Journal of Universal Language, 25(1), 99–119. https://www.sejongjul.org/archive/view_article?pid=jul-25-1-99

Pasquale, F. (2020). New laws of robotics: Defending human expertise in the age of AI. Belknap Press of Harvard University Press.

Pawar, R. (2026). AI, algorithms, and linguistic bias: How technology shapes language use in the digital age. Frontiers in Social Sciences Research, 2(2), 35–44. https://fssrjournal.org/index.php/fssr/article/view/29

Playlab. (n.d.). Playlab [AI application platform]. Retrieved August 3, 2026, from https://www.playlab.ai/

Reinhardt, J., Gupta, A., Poole, R., & Avci, D. (2023). Critical language awareness: Language power techniques and English grammar. University of Arizona Libraries. https://opentextbooks.library.arizona.edu/languageawareness/

Rotman, D. (2026, May 26). A reality check on the AI jobs hysteria. MIT Technology Review. https://www.technologyreview.com/2026/05/26/1137855/a-reality-check-on-the-ai-jobs-hysteria/

Sanford, A. (2024, February 14). Artificial intelligence is putting innocent people at risk of being incarcerated. Innocence Project. https://innocenceproject.org/news/artificial-intelligence-is-putting-innocent-people-at-risk-of-being-incarcerated/

Schwartz, R., Vassilev, A., Greene, K. K., Perine, L., Burt, A., & Hall, P. (2022). Towards a standard for identifying and managing bias in artificial intelligence (NIST Special Publication 1270). National Institute of Standards and Technology. https://www.nist.gov/publications/towards-standard-identifying-and-managing-bias-artificial-intelligence

Shapiro, S. (2022). Cultivating critical language awareness in the writing classroom. Routledge. https://doi.org/10.4324/9781003171751

Simonite, T. (2020, December 3). A prominent AI ethics researcher says Google fired her. Wired. https://www.wired.com/story/prominent-ai-ethics-researcher-says-google-fired-her/

Smith, G., & Rustagi, I. (2021, March 31). When good algorithms go sexist: Why and how to advance AI gender equity. Stanford Social Innovation Review. https://doi.org/10.48558/A179-B138

Sonnenfeld, J. A., Henriques, S., Griessel, J., Alam-Nist, A., & Yu, P. (2026, May 4). The real job destruction from AI is hitting before careers can start. Yale Insights. https://insights.som.yale.edu/insights/the-real-job-destruction-from-ai-is-hitting-before-careers-can-start (Original work published in Fortune, April 29, 2026)

Sperber, L., MacArthur, M., Minnillo, S., Stillman, N., & Whithaus, C. (2025). Peer and AI Review + Reflection (PAIRR): A human-centered approach to formative assessment. Computers and Composition, 76, Article 102921. https://doi.org/10.1016/j.compcom.2025.102921

Startpage. (n.d.). Startpage private search engine. Retrieved August 21, 2026, from https://www.startpage.com/

Sustainable AI Group. (n.d.). Who we are. Retrieved August 21, 2026, from https://sustainableaigroup.com/#whoweare

Taylor, D. B., Metz, C., & Miller, K. (2024, October 8). Nobel physics prize awarded for pioneering A.I. research by 2 scientists. The New York Times. https://www.nytimes.com/2024/10/08/science/nobel-prize-physics.html

Tenen, D. Y. (2024). Literary theory for robots: How computers learned to write. W. W. Norton.

Thompson, D. (2025, August 27). The evidence that AI is destroying jobs for young people just got stronger. Derek Thompson [Substack newsletter]. https://www.derekthompson.org/p/the-evidence-that-ai-is-destroying

UN News. (2026, June 5). AI’s environmental costs threaten water, land and climate. United Nations. https://news.un.org/en/story/2026/06/1167658

UN Women. (2026, June 22). AI is already rewriting reality for billions of people. It is getting women wrong. https://www.unwomen.org/en/news-stories/media-advisory/2026/06/ai-is-already-rewriting-reality-for-billions-of-people-it-is-getting-women-wrong

United Nations Educational, Scientific and Cultural Organization. (2024). AI competency framework for students. https://unesdoc.unesco.org/ark:/48223/pf0000391105

United Nations Educational, Scientific and Cultural Organization. (2025, July 8). AI large language models: New report shows small changes can reduce energy use by 90% [Press release]. https://www.unesco.org/en/articles/ai-large-language-models-new-report-shows-small-changes-can-reduce-energy-use-90

University of Arizona Libraries. (n.d.). How can I protect my privacy when using ChatGPT & similar tools? Retrieved August 21, 2026, from https://ask.library.arizona.edu/faq/407973

van Dijck, J. (2014). Datafication, dataism and dataveillance: Big Data between scientific paradigm and ideology. Surveillance & Society, 12(2), 197–208. https://doi.org/10.24908/ss.v12i2.4776

VandeHei, J., & Allen, M. (2025, May 28). Behind the curtain: A white-collar bloodbath. Axios. https://www.axios.com/2025/05/28/ai-jobs-white-collar-unemployment-anthropic

Vaswani, A., Shazeer, N., Parmar, N., Uszkoreit, J., Jones, L., Gomez, A. N., Kaiser, Ł., & Polosukhin, I. (2017). Attention is all you need. In Advances in neural information processing systems (Vol. 30, pp. 5998–6008). Curran Associates. https://proceedings.neurips.cc/paper/2017/hash/3f5ee243547dee91fbd053c1c4a845aa-Abstract.html

Vieira, H. (2024, March 25). Madhumita Murgia: “AI can do harm when people don’t have a voice.” LSE Business Review. https://blogs.lse.ac.uk/businessreview/2024/03/25/madhumita-murgia-ai-can-do-harm-when-people-dont-have-a-voice/

Watson, M. (2018, May–June). Contesting standardized English. Academe, 104(3). https://www.aaup.org/academe/issues/104-1/contesting-standardized-english

We Must Act Now. (2026). We must act now: A statement on AI’s transformation of the economy. Stanford Digital Economy Lab. https://www.wemustactnow.ai/

WVU Libraries. (2026, May 28). Q&A with Dr. Dustin Edwards on ideology, policy, and protest in the AI supply chain. Ex Libris Magazine, West Virginia University. https://exlibris.lib.wvu.edu/news/2026/05/28/q-a-with-dr-dustin-edwards-on-ideology-policy-and-protest-in-the-ai-supply-chain

Young, V. A. (2010). Should writers use they own English? Iowa Journal of Cultural Studies, 12(1), 110–117. https://doi.org/10.17077/2168-569X.1095

Young, V. A., Barrett, R., Young-Rivera, Y., & Lovejoy, K. B. (2019). Other people’s English: Code-meshing, code-switching, and African American literacy. New City Community Press. https://parlorpress.com/products/other-peoples-english-code-meshing-code-switching-and-african-american-literacy

Zimmer, K. (2026, July 21). Using AI chatbots? Here are four ways to reduce the energy drain. Knowable Magazine. https://knowablemagazine.org/content/article/food-environment/2026/four-ways-to-reduce-energy-use-with-ai-chatbots

AI Use Statement

We did not use AI to come up with words or ideas for this statement. We used AI in a limited capacity for research assistance and formatting.

License

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License (CC BY-NC 4.0).
PAIRR · Peer and AI Review + Reflection · August 2026