S1-S6 cross-curricular: what AI costs — energy, water and unseen labour, in three clues and a three-column table

• StageS1-S6
• SubjectCross-curricular
• Learning strandhow AI basically works, human-machine communication and collaboration, reflection on the application of AI
• Official documentsBlueprint for Digital Education Development in Primary and Secondary Schools · AI Literacy Learning Framework for Primary and Secondary Schools · Guide to Using AI in Teaching in Primary and Secondary Schools · EDBCM No. 156/2026
• Session timeone session (60 minutes); the verification checklist is a follow-up task

Taking one clue at a time and separating "what we know" from "what we do not know yet": energy, water and unseen labour, plus a three-column table, one small class tally and five questions — so that pupils leave with the next question instead of a figure.

• AudienceTeachers Students
• ThemeResponsible AI use

• Equipmentthe three-clues worksheet (one per pupil); the three-column table and the one-page verification checklist (one per group, printable on its own); a projector or whiteboard; a clock
• Version1.1
• Last updated2026-11-12

A tool can give you an answer, and there is a cost behind that answer. This page gives no figures at all: it takes three clues and splits each one in two — “what we know” and “what we do not know yet” — and then has pupils ask the next question. Three clues, three columns, five questions, nothing added or dropped.

What this solves

Pupils use tools every day and rarely ask one thing: what did that use up?

The question is easy to lose in either of two directions. One is memorising a figure — one query costs so much water, so much electricity. It sounds forceful, but what pupils remember is a figure, not an understanding; and a figure moves, gets compared, gets carried across places, and stops being true — and once it has stopped being true, nobody can tell. The other direction is a moral conclusion: tools waste, tools are wrong. Pupils nod, then carry on, because they are holding no question they can ask.

This page takes a third route: write “what we know” and “what we do not know yet” as two separate columns. What is known is explained as a process; what is not known is left open, with “who could we ask, and what kind of source could we check” beside it. Pupils leave the room holding several questions rather than one figure.

Why the page carries no figures at all. Because the thing being taught is not a conclusion but a method: telling apart what I know from what I do not, and knowing where to look next. Any single quantity turns the lesson into a memory test, and a memory test has one right answer — at which point pupils stop asking. So the page gives no energy used, no water used, no emissions, no cost, and no conversion or comparison of any kind. That is deliberate, not an omission.

Three statements to read first:

1. The three clues, the three-column table and the verification checklist are this resource’s suggestion. The official documents do not use these names and do not set how many clues, columns or checklist items there should be. The grounds are the two requirements the documents do set out — reflecting on the limitations and impact of AI, and technology for good. None of this may be presented as an Education Bureau requirement.
2. This page quotes two official grounds only, verbatim. The AI Literacy Learning Framework for Primary and Secondary Schools exists here only as a scanned copy with no text layer, so this page names the document and quotes nothing from it, and fills in none of its learning focuses, unit names or level descriptors. Where a school wants that mapping, the page provides blank fields for the school to complete from the original.
3. The Primary Information and Innovation Technology Curriculum Framework (trial version) is a primary pilot. This page covers S1-S6; at secondary level the same set of terms is used, but there is no equivalent official framework, so the page neither presents it as a secondary requirement nor quotes items from it.

This page does five things it does not do:

• It carries no quantity of any kind. Energy used, water used, emissions, training cost, query-count conversions and comparisons between places or years do not appear; nor does “one query equals this much water”, “training a model equals this many households’ electricity”, or “this uses several times the power of a search engine”.
• It names no company, brand, model, service, country, city or institution. It says “the tool”.
• It accuses nobody. The labour clue says only that “someone labels data, moderates content and handles reports”; it uses no accusatory description and points at no place or group of people.
• It evaluates no tool. It does not argue for or against using one, and does not compare one with another.
• It asks for no pupil’s own usage record. The class tally records one total for the whole class, unnamed, with nothing per pupil.

Learning objectives

Who it is for. S1-S6, across subjects. The same material differs between younger and older classes in the depth of the follow-up question, not in the number of clues. A teacher may set the single run through the three-column table as a follow-up task.

After one lesson, pupils can:

1. Name the three clues and, for each, say what we know and what we do not know yet.
2. Use the three-column table to record one thing they have seen, with a person they could ask or a kind of source they could check.
3. Take part in one tally that records a class total only, and say why it is unnamed and why nothing is converted.
4. Read out the five questions and say which is hardest to answer, and why.

The page rests on the three learning strands — how AI basically works, human-machine communication and collaboration, and reflection on the application of AI — and takes people-centred education and technology for good as its values. It is also a rehearsal in timely and appropriate use: you can only call a use timely once you can ask what it took and whether that was worth it. It sits between a sense of responsibility and critical thinking: critical thinking is being able to say how much you know and how much you do not; a sense of responsibility is going after the next question once you know you do not know, instead of treating a figure you were told as the answer.

Teaching steps

The attachments are two Word documents: the three-clues worksheet (one per pupil) and the three-column table with the one-page verification checklist (one per group, printable on its own).

Step one: the three clues (three, fixed, in this order)

Take them one at a time and stop after each one, so pupils can say the “we do not know yet” side themselves. The order is fixed: energy first, then water, then labour — from what can be seen (power, water) to what cannot (people).

Three rules for the three clues:

• “We do not know yet” is not a blank — it is a question waiting to be asked. Every cell must be written as a question; if you cannot write the question, the thinking is not finished.
• Say how what is known actually works. “It uses power” is not a description, it is a process: something has to run somewhere, be cooled, and be prepared by people first.
• Do not turn “we do not know” into an accusation. The labour clue especially: say that people are doing work, and do not say that anyone is being taken advantage of; point at no place, trade or group.

Step two: the three-column table (three columns, fixed)

One per group, filled left to right. The third column is the one that matters: without it, the first two are only a reaction.

The foot of the table carries one row: source and year — anything named in column three is written there with its source and year; anything without a source does not count as checked. The row is left blank for pupils, and the page pre-fills no source.

Two rules:

• Column three may not say “look it up online”. Name the kind of source or the role; “online” is not a source.
• One row per thing. This is a record of one follow-up question, not a list; the narrower it is, the further it goes.

Step three: one small class tally (counts and minutes only)

The whole class does this together, recording two things only: how many times the class used a tool over a set period before the lesson, and how many minutes in total. One total for the class.

Four rules, stated as it is done:

• Counts and minutes only, never a person. No name, no class, no seat number, and nobody is asked how long they used it.
• Nothing is converted. Counts do not become electricity, minutes do not become water, and neither becomes a cost. A conversion needs a factor, and this page supplies no factor.
• One class total. The total goes on the board, as one number, and is wiped off afterwards.
• The tally is there to raise a question, not to rank anyone. No comparison between classes, and none between pupils.

The point of the step is not to arrive at a quantity but to let pupils see that between “how many times we used it”, which we can answer, and “how much power that took”, which we cannot, sit several questions.

Step four: the five questions (five, fixed, in this order)

The teacher reads the five out; pupils need not answer all of them. Whatever cannot be answered goes on the “we do not know yet” side.

The fifth is the one that closes the set: the first four set out the cost, the fifth names the decision. A cost does not simply occur; someone decided to use the tool.

The shape of the lesson

Younger classes may merge the table and the questions into one oral exercise; everything else stays. The number and order of the clues do not change.

Teacher checks (four)

1. Do not ask pupils to guess a figure. When a pupil asks “but how much is it”, the answer is “this page will not tell you — but here is where you could check”.
2. Evaluate no tool and no company. Do not say which tool is better, worse or greener; the page covers how the thing works and where the cost shows up.
3. The labour clue states that work exists, and accuses nobody. No accusatory description, no pointing at a place, a trade or a group; when a pupil makes an accusation, take it back to column three: “where could you check that?”
4. The tally is unnamed and unranked. One class total, wiped off afterwards; nothing per pupil, no class comparison.

Three closing sentences (read together)

1. “What I saw, I can say.”
2. “What I do not know yet, I can say.”
3. “Who I can ask, and what kind of source I can check, I can say.”

Where this meets other resources

Interfaces only; none of their work is repeated here:

• The hallucination hunt: a repeatable routine for checking more than one source handles “is this sentence true”; this page handles “what did this cost”. Both use verification, but they ask different questions.
• After AI gets something wrong, who has to account for it: a responsibility map handles the account of each step after an event; this page is not about responsibility, only about where a cost shows up.
• Copyright and intellectual property in a classroom AI workflow handles whether content may be used; this page is not about permission, only about the three clues behind a use.

Answer key and marking notes

There is no correct answer here and nothing is scored. What follows is for the teacher to compare against, never to mark right or wrong.

Notes on the three clues

• Energy: “training and use both draw electricity” and “the power is drawn in a data centre” are enough; following it up with “who pays the bill” is a bonus, not a requirement.
• Water: “cooling uses clean water” and “water has a source and an outlet” are enough; noticing that source and outlet are two different questions is better still.
• Labour: “someone labels data, moderates content and handles reports” is enough. An answer that accuses a place, a trade or a group of people is not correct and goes back to column three.

Notes on the three-column table

• Column one is adequate if it gives the situation (a use, a picture, an answer); a reaction such as “AI is wasteful” means the cell is not filled.
• Column two must be a question, with its clue named. “I want to know more” does not pass.
• Column three must name a role or a kind of source. “Look it up online” and “ask the AI” both fail: the first is not a source, the second is the thing being asked.
• A blank “source and year” row is normal; where nothing was found, write “not found” rather than a year someone mentioned.

About quantities

A pupil who produces a specific figure neither gains nor loses marks: the page deliberately supplies no quantity, and a teacher should not add one in the room. Where a pupil brings a figure, the response is three questions: where is it from? What year? Under what conditions was it produced? Those three questions are the verification.

What counts as finished: all three clues covered, at least one row per group in the table, the tally recorded as a class total only, and the five questions read with the hardest one named.

Files

Three-clues worksheet (Word document, editable)

One per pupil: the five facts at the top (stage, subject, learning strand, documents, preparation time), the two-column comparison for each of the three clues (what we know, what we do not know yet, left blank for pupils), the five questions, and the three closing sentences. There is no field for a quantity anywhere.

Three-column table and one-page verification checklist (Word document, editable, printable on its own)

One per group: the three-column table (what I saw, what I still need to know, who to ask and what kind of source to check) with the source-and-year row at the foot; plus a page that prints on its own as a verification checklist (the five questions and the three sentences to take away) for the follow-up task. Its source fields are blank, for pupils to complete.

Each document carries a version record line in its footer.

Official documents

Every resource in this library is written to the AI Literacy Learning Framework for Primary and Secondary Schools and falls within the policy scope of the Blueprint for Digital Education Development in Primary and Secondary Schools.

This resource maps to:

• Blueprint for Digital Education Development in Primary and Secondary Schools (three-fold leaflet, p.2): requires digital and AI education to be planned in the School Development Plan, with the arrangements explained to teachers, pupils and parents; teaching what one use costs is part of explaining those arrangements.
• Supplement I: AI Literacy Learning Framework for Primary and Secondary Schools: this page names the document and nothing more. The copy held here is a scan with no text layer, so the page quotes no sentence from it and fills in none of its learning focuses, unit names or level descriptors; where a school wants that mapping, it uses the blank fields on the worksheet and completes them from the original.
• Supplement II: Guide to Using AI in Teaching in Primary and Secondary Schools (pp.2-2 to 2-5): asks teachers to exercise professional judgement and to guide pupils in reflecting on the effects of applying AI; the three clues and the three-column table are one way of meeting that in a school.
• EDBCM No. 156/2026 (as published): see the two official grounds below.

Two official grounds (verbatim in the Chinese edition, where the published wording governs; rendered here so that the English edition reads on its own)

One, the AI component of the Primary Information and Innovation Technology Curriculum Framework (trial version), the section on promoting AI education at primary level: schools design inquiry-based learning and teaching activities so that pupils, while applying AI, reflect on its limitations and impact, and so build a responsible and proactive learning attitude.

Two, EDBCM No. 156/2026, Appendix 1, section 4, “Student learning level”: uphold the idea of technology for good, so that pupils learn by doing, learn in using, think while learning and create through thinking.

Learning strands: how AI basically works, human-machine communication and collaboration, reflection on the application of AI

Official documents page: https://www.edb.gov.hk/en/edu-system/primary-secondary/applicable-to-primary-secondary/it-in-edu/debp.html

This page is an independent Beets teaching resource, not an official EDB document. The names “three clues”, “the three-column table” and “the verification checklist”, the number and order of the clues, the column headings and the content of the five questions are this resource’s suggestion, not an Education Bureau requirement. The page deliberately carries no quantity of any kind: energy used, water used, emissions, training cost, query-count conversions and comparisons between places or years do not appear, and nothing is converted or compared. No company, brand, model, service, country, city or institution is named. The labour clue says only that someone labels data, moderates content and handles reports; it accuses nobody and points at no place or group. The page evaluates no tool and argues neither for nor against using one. The class tally records one total for the class only, unnamed and unconverted. The page quotes two official grounds only, and names the AI Literacy Learning Framework without quoting from it.

Feedback

If you have any comments or questions about this resource, or a suggestion for the next one, please tell us on the feedback page.