Week 13 — Prompt Engineering: Asking Better Questions¶

At a glance


Date	Friday, November 13, 2026
Unit	2 — AI Unmasked
Big idea	Better prompts give an AI system clearer context, constraints, and success criteria.
Goal	Students improve weak prompts and observe how clearer instructions change AI output.
Student accounts needed	None. Teacher-led demo only.
Prep time	5 minutes

Why this lesson looks the way it does¶

Prompt engineering can sound like secret magic words. For students, it should feel like good communication: be specific, give context, define the audience, state constraints, and ask for a checkable output.

Because students do not have accounts, this is teacher-led on the projector with pair discussion doing the heavy lifting. That is fine. The learning target is not "everyone chats with a bot." It is "everyone can improve an instruction and evaluate the result."

Your professional lens is especially useful here. In business settings, prompts are rarely isolated clever sentences. They are part of a workflow: define the task, provide source material, constrain the output, review it, and iterate.

CSTA 2026 standards¶

Concept area	Connection
Algorithms & Design	Students iteratively refine instructions to produce better outputs.
Programming	Students compare prompting with giving precise instructions to a computer.
Data & Analysis	Students evaluate outputs against criteria and revise inputs based on evidence.
Systems & Security	Students keep AI use teacher-driven and discuss not entering private information.
Computing & Society	Students practice responsible use: AI output is draft material, not automatic truth.

"I can…" objectives¶

By the end of today, students can say:

I can improve a vague prompt by adding context and constraints.
I can ask an AI system for a specific format or audience.
I can compare two outputs and explain which one better meets the goal.
I can avoid putting private information into an AI prompt.

Teacher prep — the 5-minute version¶

Open your preferred free AI chatbot on the teacher computer, projector only. If none is available, use the unplugged transcript version below.
Write on the board:

Last Friday: language models predict likely next tokens.
Today: better input usually gives better output.
Write the prompt recipe:

Task + context + audience + constraints + format
Prepare this weak prompt:

"Tell me about robots."

Minute-by-minute run sheet¶

0:00–0:04 — Anchor¶

"Last Friday we built sentences one word at a time and saw that language models generate likely text. Seven days later, the practical question is: if the model is predicting from our input, how do we give it better input?"

0:04–0:12 — Mini-lesson: prompt recipe¶

Show the weak prompt:

"Tell me about robots."

Ask:

"What is missing? Who is this for? How long should it be? What kind of robots? What counts as a good answer?"

Introduce the recipe:

"A strong prompt usually has five parts: the task, the context, the audience, the constraints, and the format. Not every prompt needs all five, but if the output is weak, one of these is often missing."

Build a better prompt with the class:

"Write a 120-word explanation of how classroom robots follow instructions for 5th–6th graders. Use one analogy, avoid scary language, and end with one question students could discuss."

0:12–0:36 — Paired work: prompt repair shop¶

Pairs do not need AI access. They repair prompts on paper or in a local document.

Put these weak prompts on the board:

"Explain AI."
"Make a study plan."
"Write a story."
"Help me with my game idea."
"Tell me if this is true."

Pairs choose one and rewrite it using the recipe. Require:

audience
length or format
at least two constraints
one way to judge success

Every 6–7 minutes, test one revised prompt on the projector. Before running it, ask:

"What do we expect this prompt to improve?"

After the output:

"Did the prompt control the output the way we wanted? What would we change next?"

Privacy boundary

Say this explicitly: "Do not put your full name, address, password, private family information, or someone else's personal story into an AI prompt. If the task needs private details, anonymize or do not use AI."

Ask three pairs to read only their improved prompt, not a long output.

"Listen for the recipe: task, context, audience, constraints, format. Which part made the biggest difference?"

Professional connection:

"In real work, the prompt is only one part. The human still owns the goal, the source material, the review, and the final decision."

0:42–0:45 — Teaser + prompt¶

"Next Friday we will look at one of the most important failure modes: hallucinations. That is when an AI gives an answer that sounds confident but is wrong."

Tiered challenges¶

🟢 Starter (grades 5–6): Add audience and format to a weak prompt.
🟡 Builder (grades 7–8): Add constraints that prevent a too-long, too-vague, or off-topic answer.
🔴 Extender (grades 9–10): Write a prompt that includes evaluation criteria, then propose a second follow-up prompt to improve the first output.

"Which part of the prompt recipe improved the answer the most today?"

🤔 Think About This¶

Notice a time this week when better instructions would have saved time.

It could be with a person, a game, a search box, or a computer. Ask: what detail was missing from the original instruction?

🔌 Unplugged Plan B — if the internet is down¶

Use the teacher as the "language model."

Time	Activity
0:00–0:04	Same anchor: last week next-token prediction, today better inputs.
0:04–0:12	Teach the recipe with "Tell me about robots."
0:12–0:28	Prompt repair shop. Pairs rewrite weak prompts on paper.
0:28–0:38	Teacher role-play: answer one vague prompt vaguely, then answer the improved prompt better. Keep it playful but honest.
0:38–0:42	Pairs add a follow-up prompt that would improve the answer.
0:42–0:45	Same teaser and Think About This.

Common problems & fixes¶

Problem	Fix
Students try to find "magic words"	Reframe: "This is clear communication, not a spell."
Prompt outputs are too long	Add length and format constraints before rerunning.
Students want individual AI access	Keep boundary: teacher-led only; no accounts needed.
Students include personal details	Stop and model anonymizing: "a student," "a hobby," "a city," not private facts.
AI tool is blocked	Use the unplugged teacher-as-model version. The skill is prompt improvement.

Vocabulary¶

Prompt — the instruction or input given to an AI system
Context — background information the system needs for the task
Constraint — a limit or rule for the output
Audience — the people the output is meant for
Iteration — improving something through repeated attempts

Reflection — fill in after class¶

Which prompt recipe part was easiest for students?
Which was missing most often?
Did privacy guidance land clearly?
Save one strong student prompt for Week 14's verification lesson.

Next week: Week 14 — Hallucinations: When AI Is Confidently Wrong