Teaching Who Breathes It, Who Pays for It
The civics-forward page in this series. Most students have never seen a regulatory filing, and are surprised to learn that the documents deciding their electricity bill are public, readable, and open for comment.
Why this is hard to teach.
Distributional arguments are harder to teach than aggregate ones because averages feel like facts and distributions feel like opinions. They aren't. “Small on average” and “concentrated in specific counties” are both measurements, and this guide asks students to hold them at the same time.
The efficiency finding is counterintuitive, and students resist it. Optimizing for less total energy, cost, and water sounds unambiguously good. The resistance is the lesson. Let them argue for aggregate optimization before you show them what it does to the distribution.
Third, this guide is where a class most easily tips into cynicism. The venue where these decisions get made is obscure, technical, and attended almost entirely by parties with money at stake. That is true, and it is also an opening rather than a wall: the proceedings accept public comment from anyone. Teach the second half or you will produce a room that has learned only to feel powerless.
Target misconceptions.
“Efficient means fair.”
They are different objectives, and a system optimized for one trades away the other unless someone writes the other into the objective function. This generalizes far past data centers. It is true of any system optimizing on your behalf.
“The diesel generators are the health problem.”
Grid electricity outweighs backup diesel by roughly 12 to 1 in the modeled health burden. Generators are visible, locally permitted, and photographable, which is why they dominate coverage. They are also the smaller share by an order of magnitude.
“My electric bill went up, so data centers did it.”
Rates move on fuel costs, storm recovery, transmission investment, and rate-case outcomes at once. The New Jersey State Policy Lab's assessment is titled “Are Data Centers Raising Your Electric Bill? Mostly Not. Yet.” Both halves of that title are doing work.
“These decisions are made behind closed doors.”
They are made in public dockets that accept written comment from anyone. The problem is not secrecy. It is that almost nobody without a financial interest ever shows up.
Two classroom-ready activities.
Find the docket
40 min · public recordsStudents locate their state utilities commission, determine whether a large-load or rate proceeding is currently open, and read one filing from it, any filing, start to finish.
Then each student answers three questions in writing: who filed this, what are they asking for, and who would pay if they got it.
The debrief question that does the teaching: “Was this hard to find, or just unfamiliar?” Almost always the answer is unfamiliar. Students who have been told these decisions are hidden discover that they are published, searchable, and dull, which is a more actionable finding than secrecy would have been.
Materials: internet access. State utility commission sites are public and require no account. If no proceeding is open in your state, North Carolina's E-100 Sub 208 works as a substitute.
Write the objective function
30 min · no techGive groups a simple routing problem: five data centers in five regions, each with a different electricity price, carbon intensity, and water intensity. Assign each group one objective — cheapest, lowest-carbon, or lowest-water — and have them route the workload.
Then reveal the fourth column they weren't given: which regions already carry the highest per-household health burden. Have each group check where their optimizer sent the load.
Every group discovers their objective degraded something they weren't measuring. Then ask them to rewrite their objective so it doesn't, which is exactly what the researchers did, and most of the effect goes away. The lesson is not that optimization is bad. It is that an objective function is a moral document, and somebody wrote it.
Materials: a printed table of made-up but plausible regional figures. Deliberately not real data; the point is the structure, and real numbers invite arguing about the numbers.
Discussion prompts.
Ordered from easy to charged.
- 1.Who is “everyone on one utility's bill”? Are you on that list?
- 2.Why is the health burden about seven times higher in some counties than the national average?
- 3.If a routing algorithm sends work to the cheapest region, who is that decision good for?
- 4.The New Jersey analysis says “mostly not — yet.” What would have to change for “yet” to arrive?
- 5.These proceedings accept public comment and almost nobody comments. Why not, and what would change that?
- 6.Is it fair to ask a distressed county to weigh investment against health burden? Who should be making that trade, and are they the ones making it now?
Seeing whether it landed.
The docket summary. One paragraph explaining a real filing to someone who has never read one: who filed, what they want, who pays. Plain language, no jargon carried over.
The objective rewrite. Give students any optimization they encounter elsewhere, such as school scheduling, delivery routing, or hospital triage, and ask what it optimizes and who is not in the objective.
The two-truths test. Ask a student to state both “the average burden is small” and “the burden is severe in specific places” and explain why both are measurements rather than opinions.
When someone asks “is this proven?”
In teacher voice
"The health figures are modeled and scenario-dependent: the number you'll see quoted, $20.9 billion, is the high-growth 2028 case, and the low case is about $11.7 billion. Always say which. The finding that efficiency optimization amplifies inequity comes from a trace-based simulation on ten hypothetical facilities, not from auditing any real company's routing, and that's a meaningful limitation. And whether data centers are raising retail rates specifically is genuinely contested by serious people. What isn't contested is that the cost-allocation rules are being written right now, in public, by people who accept comment."
Delivering a correction to a room that cares
Every guide in this series involves telling people that a claim they hold is imprecise. The flagship educator page carries the three-beat rule — the sequence that keeps a correction about scale from being heard as indifference. Read it before you teach any of these.
The three-beat rule →Where this leads
Teaching this is a different skill than knowing it.
Teaching AI Well is the facilitator library behind CPAI's Certified Applied AI Trainer Program — ten lessons on how to teach AI honestly, including the material on this page.