AI environmental impact lesson for grades 6-8: 45-min plan

Your students have been asking ChatGPT questions all week — for homework help, for research, for answers to the kind of questions they’d normally just let sit. According to a 2025 Scholastic Science World report, one multi-turn AI conversation uses roughly 0.5 liters of water for data center cooling, compared with approximately one-tenth that amount for a traditional Google search. That gap is the lesson. This post is the 45-minute plan to make it concrete.

This lesson is the environmental ethics extension in Track B: Going Deeper — AI Ethics and Critical Thinking — the track that moves past “AI can make mistakes” into the harder societal questions.

What makes AI different from a Google search

When a student types a question into a search engine, the server looks up an index and returns a ranked list of links. It is essentially a very fast filing-cabinet lookup. When a student types the same question into an AI chatbot, the system runs billions of matrix multiplications across a neural network to generate the response token by token. Those are not the same operation. Not even close.

There are two cost moments in any AI system’s life. The first is training — the months-long process of feeding the model enormous amounts of text data and adjusting billions of parameters until the model can generate coherent responses. A widely cited 2019 study by Strubell et al. (Emma Strubell, Ananya Ganesh, and Andrew McCallum, “Energy and Policy Considerations for Deep Learning in NLP”) estimated that training a large language model can require energy comparable to the lifetime emissions of five average American cars. You and your class had nothing to do with that cost; it happened before any of you ever typed a prompt. The second is inference — what happens every time a student asks the chatbot a question. Each inference query costs approximately 10 times the energy of a comparable Google search, according to multiple independent analyses including Goldman Sachs Research (2024). That 10x difference is what your students can actually think about on Monday.

The water piece is the number that gets 12-year-olds to pay attention. Data centers require continuous cooling, and cooling at scale requires water — either directly in cooling towers, or indirectly through the power generation feeding the facility. Current reporting, including Scholastic Science World (April 2025), puts the working estimate at roughly 0.5 to a few liters of water per multi-turn AI conversation. That is not a reason to ban AI in your classroom. It is a reason students should know what they are choosing when they choose it.

This connects directly to AI4K12 Big Idea #5: Societal Impact — the principle that AI systems can affect society in both positive and negative ways, and that those effects extend to the physical world, not just the social one.

The numbers your class can actually calculate

The energy math does not require a calculator with a lot of buttons. Here is the version that works in a grade-6 classroom on a Tuesday with 28 minutes left in the period.

The water proxy formula:

conversations × 0.5 liters = approximate water used

Walk through it together: “If our class of 30 students each had one AI conversation today, that’s 30 × 0.5 = 15 liters of water.” Then pause. “Does anyone know how big a medium fish tank is?” A standard 10-gallon tank holds about 38 liters, so 15 liters is roughly the size of a third of a fish tank. A class of 30 students doing that every school day for a year — 180 days — would account for approximately 2,700 liters, or about 70 full fish tanks worth of water (derived from the 0.5-liter base estimate above; treat as an order-of-magnitude illustration, not a precise figure).

The energy formula (if you have a science co-teacher who wants numbers):

AI queries × 0.002 kWh = approximate watt-hours used

One query uses roughly 0.002 kilowatt-hours, based on Goldman Sachs Research (2024) estimates. Your classroom lights probably run at 0.1 kWh per hour per fixture. So 50 AI queries is roughly the same electricity as running one classroom light for an hour. Students should treat these as order-of-magnitude comparisons, not precise measurements.

The pedagogical purpose here is not guilt. It is calibration. Most students who have never thought about this assume that digital = costless because digital = invisible. The moment the invisible becomes a number — 15 liters, a fish tank — students can be curious rather than defensive. They can ask: does it matter how many questions I ask? That question is the whole lesson.

This section builds directly toward CCSS.ELA-LITERACY.W.7.8: gathering relevant information from digital sources and assessing the credibility and accuracy of each source. Students are not just receiving the number — they are calculating it themselves and then asking whether it holds up.

The 45-minute lesson plan, step by step

This plan assumes one class period, no prior AI lesson required, and a standard setup with whatever devices your school has available.

Minutes 0–5 — The hook image. Show students two photographs side-by-side on the projector: a large data center (rows of servers, blue lighting, cables everywhere) and a power plant (cooling towers, steam rising). Ask the room: “What do these two buildings have in common? What does either of them have to do with your homework?” Don’t answer yet. Let the room guess for two minutes, then give the one-sentence bridge: “Every time you ask an AI a question, one of those buildings is involved.”

Minutes 5–15 — Read and calculate. Read a short nonfiction passage aloud — two paragraphs on data center energy use, at a 900-Lexile level. Students follow along on a printed copy (the calculation worksheet handles this). After the read, students work the water-proxy formula with their own class numbers. How many students? How many AI conversations did we estimate we’ve had this week? What’s the total?

Minutes 15–30 — Small group discussion. Groups of three or four. Each group gets the same three discussion questions (see the next section). Their job is not to reach consensus — it is to hear three different opinions before the share-out.

Minutes 30–40 — Class share-out and exit ticket. One spokesperson per group reports the most interesting disagreement in their group. Not the answer — the disagreement. Then the exit ticket, written on the back of the worksheet: “One thing I will do differently the next time I use AI.” Collect on the way out.

Minutes 40–45 — Standards crosswalk wrap-up. One sentence from you: “What we just did connects to this.” Put the standard codes on the board. Students copy them. This takes three minutes and it means that when their parent asks “what did you learn today,” the student has a concrete answer.

The full printable for this lesson — calculation worksheet, nonfiction reading passage, and grading rubric — is available as the Environmental Impact of AI Lesson on TPT. Open the PDF, print, teach Monday.

The three discussion questions that land with 12-year-olds

These three questions work because none of them have a single correct answer. Cold-call is safe here. Every student has a defensible position, which means the quietest student in the back row has exactly as much standing as the student who always speaks first.

Question 1 — “If every student in our school asked AI 5 questions per day for a year — how much water is that?” This is the math question dressed as a discussion question. It gets students doing estimation with real stakes: a school of 600 students × 5 questions × 0.5 liters × 180 school days works out to roughly 270,000 liters — comparable to a small backyard swimming pool (order-of-magnitude estimate; exact figures vary by model and data center). Students react to that number. Let them.

Question 2 — “Should AI companies have to tell us how much energy their product uses — the way food labels tell us calories?” This is where the lesson earns its place in a social studies or ELA argument-writing unit. It brings in transparency, consumer rights, and corporate accountability without requiring the teacher to take a side. It also connects directly to AI4K12 Big Idea #5: Societal Impact — the principle that the people who design, regulate, and use AI systems all play a role in shaping their broader effects.

Question 3 — “Is using AI for homework research the same environmental decision as leaving the lights on all day? Why or why not?” The ambiguity is the point. Some students will say yes — it’s a discretionary energy use and we should be mindful. Some will say no — the light is local and visible, the AI cost is distributed and invisible, which makes comparison unfair. Some will say the question itself is wrong because individual choices matter less than policy. All three positions are teachable. The goal is proportional thinking: not “AI is bad,” but “how does this choice compare to other choices I make?”

ISTE 1.7.b — Global Collaborator — frames exactly this: students using collaborative technologies to examine issues from multiple viewpoints. This discussion is the standard in action.

Standards crosswalk — what to tell your department head

For the full AI literacy curriculum map, including where this lesson fits in a unit, see the AI literacy teaching guide for grades 6–12.

Administrators need anchor codes, not descriptions. Here are the four this lesson covers, by the codes themselves.

AI4K12 Big Idea #5 — Societal Impact. AI systems have consequences that extend beyond the immediate user, including environmental consequences. This lesson makes that concrete through calculation and discussion.

ISTE 1.7.b — Global Collaborator. Students use collaborative technologies to examine issues from multiple viewpoints. The three-question discussion structure is the direct application.

CCSS.ELA-LITERACY.W.7.8 — Gathering and assessing information. Students gather relevant information, assess source credibility, and use data accurately. The calculation worksheet exercises this standard against real figures.

NGSS MS-ESS3-3 — Human impact on the environment. For science colleagues or interdisciplinary teams: students apply scientific principles to design a method for monitoring and minimizing a human impact on the environment. This is a bonus alignment — the lesson works fully in ELA and social studies without it, but if a science teacher wants to co-teach one period on infrastructure and energy systems, every piece of this lesson supports that extension.

When you print this crosswalk for a department meeting, the conversation about whether this is “real curriculum” or “an AI add-on” usually ends at the CCSS anchor code. Principals recognize W.7.8.

Standards source: ISTE Student Standards. ISTE is a registered trademark of the International Society for Technology in Education. These resources are not affiliated with or endorsed by ISTE.

Get the full printable — and the companion writing unit

The lesson PDF: The Environmental Impact of AI Lesson for Grades 6–8 includes the student calculation worksheet, the nonfiction reading passage at grade-band Lexile, and the grading rubric. Print on a Monday morning before homeroom. The lesson requires no prior AI unit — it is self-contained.

The argument writing extension: If you want to take this into a three-day argument writing unit, the AI Environmental Impact Argument Writing Unit for Grades 6–8 extends the discussion questions above into a structured position essay. Students write on whether AI companies should be legally required to disclose their environmental costs — a question with no textbook answer, which means every student has to construct an argument rather than find one.

If you want to try the site’s free resources before committing to either purchase, the free resources page has a hallucination worksheet and a bias activity that both pair naturally with this lesson.

And if your students are ready to push further into AI critique after this lesson, the AI bias station activity for grade 8 is the natural next step — five stations, 12-year-olds auditing real AI outputs for cultural and source bias, same 45-minute structure.

---

This post was drafted with AI assistance and human-finalized.