1. Introduction: The AI Literacy Gap You Haven’t Heard About
The current conversation around Generative AI is dominated by a dizzying mixture of breathless hype and high-stakes anxiety. For many educators and parents, tools like ChatGPT feel like a tidal wave; there is a persistent fear of being “left behind” if we do not immediately master the latest “prompt engineering” hack.
Because of this pressure, AI literacy is usually sold as a narrow set of technical skills: how to prompt, how to code, or how to automate. However, emerging research suggests we are missing the most vital part of the equation. True literacy isn’t just about operating the machine; it is about Critical AI Literacy. This is a sociotechnical understanding that looks beyond the interface to examine the social, cultural, and ethical dimensions of AI systems. It is about moving from being a passive recipient of technology to an active, critical participant—someone who can see the “invisible” infrastructures shaping our digital world.
2. Your AI Is Thirsty (And Power-Hungry)
The “cloud” is a masterclass in discursive closure—a metaphor designed to hide the very real, very heavy physical infrastructure underneath. We often think of AI as an ephemeral spirit living in a digital vacuum, but its physical footprint is massive and increasingly unsustainable. One of the most surprising takeaways of critical AI literacy is the literal environmental cost of our digital assistants.
Data from recent studies highlights a staggering scale: ChatGPT consumes approximately 500,000 kilowatt-hours of electricity daily. To put that in perspective, the average U.S. household consumes only about 29 kilowatt-hours per day. This means the daily energy required to run a single AI platform is equivalent to the power usage of over 17,000 homes.
But it isn’t just about power; your AI is also “thirsty.” Data centers require millions of gallons of water for cooling to prevent system meltdowns. As researchers Akgün et al. note:
“Facilitating discussions about the fundamental mismatch between AI technologies and environmental sustainability can illuminate how the growing use of AI and data centers consumes immense amounts of water and electricity.”
In STEM education, this “phenomenon-based” learning is essential. It connects abstract technology to the very real problems of resource scarcity and climate justice, teaching students that every “generate” click has a physical consequence.
3. The Trap of “Technological Determinism”
Educators often feel a sense of “discursive closure”—a phenomenon where dominant ways of talking about technology limit our ability to imagine alternative futures. We are often told that AI integration is inevitable and that we must adapt or face “elimination.” This creates a false dilemma: you are either an “early adopter” or you are “outdated.”
This mindset strips away human agency by suggesting that technology drives social change on its own, rather than being a tool we choose to shape. To reclaim our agency, we must recognize the four forms of discursive closure that limit our viewpoint:
- Disqualification: Discounting the input of educators or students because they are perceived to lack “technical expertise.”
- Technological Determinism: The belief that technology is the sole driver of educational change, which ignores actual pedagogical needs.
- Trajectorism: The feeling that technological progress follows a linear, inevitable path that we are powerless to change, making us passive recipients rather than active designers.
- False Dilemmas: Presenting complex choices as “either-or” scenarios, such as choosing between student privacy or using “helpful” AI tools.
4. Community-Centered Design: Why Librarians and Youth are the Real AI Experts
The most effective AI programs don’t try to “fix” underserved communities; instead, they adopt an Asset-Perspective. This means leveraging the “funds of knowledge”—the unique cultural strengths, stories, and lived experiences—that students already possess.
In recent participatory design projects involving middle schoolers and public librarians, students were positioned as the primary experts. Rather than just learning technical facts, these students used AI to solve community-specific issues. For example, in projects like SimWorlds (which uses character simulations to explore fictional worlds) or the AI ABE project (which creates chatbots of historical figures), students learned to critique technology while using it. In other cases, students designed AI-powered smartwatches to detect and report bullying—a direct response to a felt need in their school community.
Critically, these students also learned the most important lesson of AI literacy: when to unplug. In the AI ABE project, when tech access was a barrier, students suggested a student-led “skit” instead. Recognizing that a human performance can provide the same interactive value as a chatbot is a sign of a truly high-functioning critical mindset.
Research emphasizes that we must move toward:
“…positioning them as builders, designers, doers, and critics of AI knowledge and tools. This helps students see themselves as creators… regarding the use and design of AI in their communities.”
5. Literacy is a Map, Not a Finish Line
Critical AI literacy isn’t a checklist you complete; it is a “dynamic practice of negotiating contradictions.” Educators are now using the Cartographies of Critical AI Literacy framework to navigate this landscape. This framework views literacy as a series of overlapping maps rather than a linear progression.
Notably, “Resistance” is not a sign of being “difficult.” In this framework, questioning a tool’s bias or involving unions is a sign of high literacy.
| The Four Cartographies | Practical Examples from Research |
| Struggle | Engaging with the AI ABE project while worrying about data privacy or internet connectivity costs. |
| Possibility | Using Powerful Prompts (curated lists) to co-create project-based learning goals with students. |
| Resistance | Developing the AI Planner but explicitly involving teacher unions to protect professional agency and prevent job displacement. |
| Action | Building the AI Grader to identify students being “left behind,” or designing SimWorlds character simulations. |
6. Datafication: The Hidden Risks of Personalized Learning
While we are looking at what AI costs the planet, we must also look at what it costs the individual student: their data privacy. We are often promised that AI will “personalize” learning, but without critical literacy, personalization can become a trap of datafication. This happens through three specific risks:
- Reductionism: Reducing complex human learning into simple, quantifiable data points like test scores.
- Abstraction: Removing student data from its original context, which leads to decisions that ignore the student’s actual home environment.
- Individualization: Profiling students into limited “target profiles” that trap them in rigid categories based on past data, essentially reinforcing old biases under a new “high-tech” label.
Critical AI literacy acts as a safeguard. It allows us to recognize when digital infrastructures are becoming exploitative rather than supportive.
7. Conclusion: From Passive Users to Active Architects
The shift we need is fundamental: we must move from teaching with AI to teaching about AI. This isn’t just a technical upgrade; it is a “sociopolitical practice” where we learn to “read the world.”
When we adopt an Asset-Perspective—valuing the wisdom of our communities over the efficiency of our algorithms—we stop being passive users. We understand AI as a sociotechnical system: one that consumes water, reflects human bias, and impacts labor. Only then can we start being the active architects of our future.
In an automated world, how will you use your “humanizing mindset” to ensure technology serves the community, rather than the other way around?
