Teach with AI Footprint Lab

Prepare the lesson

For an empirical extension, use the measurement guide and electricity lab with recorded local readings.

1. Choose an activity below and download its student and teacher editions. Downloaded HTML files open offline and print to PDF.
2. In Activities, load the reference case, then open Calculator: 25 requests/day, 1 Wh, Reference sensitivity, grid 390, no water or media.
3. Assign the worksheet and the Activities page through your usual classroom system. Students can use fictional scenarios; personal usage is unnecessary.
4. Check access to linked sources on school devices. For offline work, use the worked example and clearly mark original-source checks as incomplete.

No student login, name, email, analytics or advertising is requested by this application. Settings appear in URLs and browser history. Review hosting logs and school deployment policies separately.

Choose the right lesson

Suggested full-period pacing: 5 minutes framing, 10 minutes source reading, 15 minutes calculation and sensitivity, 10 minutes interpretation, and 5–10 minutes discussion and exit ticket.

Worked calculation and answer key

1. Annual requests: 25 × 365 = 9,125.
2. Reference energy: 9,125 × 1 Wh × 1 = 9,125 Wh = 9.125 kWh/year.
3. Reference carbon: 9.125 kWh × 390 gCO₂e/kWh = 3,558.75 g = 3.55875 kg CO₂e/year.
4. At 700 gCO₂e/kWh: energy stays 9.125 kWh; carbon becomes 6,387.5 g = 6.3875 kg CO₂e/year.
5. At 10 mL/request: 10 × 9,125 = 91,250 mL = 91.25 L/year. This is an authored sensitivity case.
6. At Long (2.5×), with grid restored to 390: energy becomes 22.8125 kWh and carbon 8.896875 kg CO₂e. Water stays 91.25 L because this model applies no response multiplier to water.

Sample student response

In this authored case, raising the grid factor changes annual emissions from about 3.56 to 6.39 kg CO₂e. Electricity stays 9.125 kWh. We would need the serving location and electricity mix to choose a real grid factor.

Supported headline: “Selected grid assumptions change this scenario’s carbon estimate.”

Unsupported headline: “Every AI user emits 6.39 kg of carbon each year.” It overgeneralizes the authored inputs and omits usage, time and boundary conditions.

Common misconceptions to discuss

• “Not estimated” does not mean zero.
• A median cannot generally be substituted for a mean to estimate a population total. Repeating a median here defines a hypothetical case.
• Separate energy cases are not a statistical confidence interval.
• Water volume alone cannot establish local ecological consequences.
• A smaller personal estimate does not establish a small industry total.
• Not disclosed is missing information, not a company’s zero impact.

Assess reasoning, not a preferred opinion

Score each criterion 0–2: 2 = complete and accurate; 1 = partly demonstrated; 0 = absent or incorrect. Total: 8 points. Accept different conclusions when evidence and priorities are explicit.

Evidence

Identifies source and exact locator; distinguishes reported values from authored inputs.

Arithmetic

Shows equations, correct units and reproducible conversions.

Boundaries

Names system, geography or unknown location, exclusions and incompatible comparisons.

Reasoning

Supports a qualified claim; separates uncertainty, missing information and value judgments.

Support options: provide the units and equation scaffold, allow a calculator, pair source reading with discussion, and accept an oral explanation. Extension: have students audit the original source before using a reported value.

Standards connections and their limits

The introductory lessons practice quantitative reasoning, source evaluation, and argument from evidence. They do not by themselves satisfy an entire NGSS performance expectation or claim official endorsement.

The advanced extension supports HS-ESS3-2 by comparing resource-management alternatives and cost-benefit ratios, and supports HS-ETS1-3 through explicit criteria and trade-offs. Teachers should check their current APES course framework before assigning specific practice codes.

Advanced decision case

Hypothetical classroom case; all numbers below are authored, not facility data. A computing facility uses 100,000 kWh and 50,000 L of direct water per year. Option A costs an additional $1,000/year and saves 20,000 kWh/year, with no change to water. Option B costs an additional $500/year and saves 10,000 kWh/year, but adds 20,000 L/year of direct water consumption. Assume equal service quality, the same grid, and no other operational changes. Capital, embodied impacts and local watershed conditions are unknown. Calculate each option’s annual energy and water, cost per kWh saved and water change. Recommend an option using stated priorities, and identify missing evidence.

Worked answer: A uses 80,000 kWh and 50,000 L/year; B uses 90,000 kWh and 70,000 L/year. Both cost $0.05 per kWh saved. A saves more energy without additional direct water; B has the smaller annual expense. Accept either recommendation when students explain priorities and missing context.

Classroom pilot protocol · not yet completed

Target: 3–5 teachers across different classrooms, including students using assistive technology. Do not publish identifiable student work without the school’s appropriate process.

1. Record grade band, activity, time used and device type. Count teacher interventions needed to find the activity and export a report.
2. Before and after the lesson, ask students to identify an assumption, distinguish water volume from impact, and explain one excluded factor.
3. Observe whether a student can reproduce a result and open its source without help. Collect anonymous counts and examples of misconceptions.
4. Test keyboard-only operation, screen reader announcements, zoom, narrow screens and school Chromebooks. Record the browser/device and exact failed action.
5. Proposed release targets: teachers prepare within five minutes; at least 80% of participating students reproduce a result and identify its main limits. These are design targets, not measured outcomes.
6. Publish aggregate findings, limitations, changes made, and unresolved problems before changing the pilot status.

Send observations through the public issue form; omit student names and identifying information.

Independent review packet · reviewers not appointed

Energy/LCA and hydrology reviewers should use the downloadable factor ledger and evidence registry in Sources & method. For each claim, record the exact source version and page/table, original value and unit, conversion, system boundary, geography, and accept/revise/remove decision.

Reviewers should also examine median-to-annual scaling, media assumptions, grid accounting, water consumption versus withdrawal, comparison boundaries, and the interpretation of missing corporate disclosures. Publish reviewer expertise, affiliations/conflicts, date, review scope, disagreements and the response to each finding. Do not label the whole site “validated” based on a narrow review.