Week 16 — AI for Good: A Design Sprint¶

At a glance


Date	Friday, December 11, 2026
Unit	2 — AI Unmasked
Big idea	Good AI design starts with a real human problem, useful data, clear limits, and a plan for failure.
Goal	Teams create and pitch a paper prototype for an "AI for Good" idea.
Student accounts needed	None. No computers needed.
Prep time	5 minutes

Why this lesson looks the way it does¶

After several weeks of limitations, students need to build toward something constructive. This paper sprint lets them use the unit's ideas without getting bogged down in tools, accounts, or pre-holiday tech friction.

The design frame matters: do not let "AI for Good" become "add AI sparkle to anything." Require a real problem, real users, data needed, failure risks, and a human-in-the-loop decision. That keeps the ideas grounded and lets older students do genuinely deeper work while younger students still participate creatively.

Your professional expertise is a classroom asset here. You can ask the same questions product teams ask: Who is the user? What decision improves? What data is available? What happens when the model is wrong? Who is accountable?

CSTA 2026 standards¶

Concept area	Connection
Algorithms & Design	Students design an AI-enabled solution with users, inputs, outputs, and constraints.
Programming	Students describe system behavior even though they are not coding it today.
Data & Analysis	Students identify what data their idea would need and what data might be missing.
Systems & Security	Students consider privacy, reliability, and human oversight.
Computing & Society	Students evaluate beneficial uses of AI and possible unintended consequences.

"I can…" objectives¶

By the end of today, students can say:

I can design an AI idea for a real problem.
I can identify what data the AI would need.
I can name one way my idea could fail or be unfair.
I can pitch a design clearly to classmates.

Teacher prep — the 5-minute version¶

No computers. Put out paper, markers, sticky notes, or index cards.
Write on the board:

Last Friday: check media before trusting it.
Today: design AI worth trusting.
Write the sprint template, or print the ready-made poster: AI for Good sprint poster — one per team.

Problem → User → AI job → Data → Failure risk → Human check
Prepare team size: 3–4 students, mixed ages.
Optional idea starters on the board:
help a teacher spot confusing quiz questions
help students plan study time
help a cafeteria reduce food waste
help a coach notice injury risk
help a town prepare for storms

Minute-by-minute run sheet¶

0:00–0:04 — Anchor¶

"Last Friday we practiced not being fooled by media. Seven days later, we are turning the unit positive: if AI systems can learn patterns and make predictions, what would be worth building?"

0:04–0:12 — Mini-lesson: good AI starts with the problem¶

Say:

"Bad AI design starts with, 'Let's add AI.' Good AI design starts with, 'Who has a problem, and what decision could be better?'"

Point to the template:

"Your pitch must answer six questions: What problem? Who is the user? What is the AI's job? What data would it learn from or use? How could it fail? What human check keeps people responsible?"

Add:

"If your idea has no failure risk, you have not thought about it hard enough yet. Every useful system has limits."

0:12–0:36 — Paired/team work: design sprint¶

Teams make a one-page pitch poster. Require these six boxes:

Problem: What real issue are you helping?
User: Who uses it?
AI job: Classify, predict, recommend, generate, or detect?
Data: What examples or information does it need?
Failure risk: How could it be wrong, biased, or misused?
Human check: Who reviews or decides before action?

Say:

"The AI does not get to be the boss of the whole system. It has a job. A human still owns the decision."

Circulate like a product review:

"Whose problem is this, exactly?"
"What would the model be predicting?"
"What examples would be missing at first?"
"What is the safest version of this idea?"
"What would you measure to know whether it helped?"

Mixed-age roles

Give younger students ownership of user story and poster clarity. Give older students ownership of data, failure risk, and human check. Everyone presents one sentence.

Run rapid pitches: 45 seconds per team if possible.

Prompt:

"Problem, AI job, failure risk, human check. Go."

After each, ask one product-manager question:

"What data would you need first?"
"What happens if it is wrong?"
"Who should not be allowed to use this output without review?"

0:42–0:45 — Teaser + prompt¶

"Next Friday is our Hour of Code celebration before Christmas break. It will be student-choice, low pressure, and tied to the global December Hour of Code campaign."

Tiered challenges¶

🟢 Starter (grades 5–6): Clearly explain the user and the problem in one sentence.
🟡 Builder (grades 7–8): Identify the AI job: classify, predict, recommend, generate, or detect.
🔴 Extender (grades 9–10): Define a failure mode, a fairness concern, and a human oversight step that would be required before launch.

"What is your AI idea, and what human check keeps it responsible?"

🤔 Think About This¶

Notice one small problem this week where a prediction might help.

Ask: what data would someone need, and what could go wrong if the prediction was wrong?

🔌 Unplugged Plan B — if the internet is down¶

This lesson is intentionally unplugged.

Time	Activity
0:00–0:04	Same anchor: from evaluating AI to designing it.
0:04–0:12	Mini-lesson with the six-box sprint template.
0:12–0:28	Teams complete the poster.
0:28–0:34	Teams trade posters and add one sticky-note question: data, failure, or human check.
0:34–0:42	Rapid pitches with one question per team.
0:42–0:45	Same Hour of Code teaser and Think About This.

Common problems & fixes¶

Problem	Fix
Teams start with a tool instead of a problem	Ask: "Who is helped, and what is hard for them right now?"
Idea gives AI too much authority	Require a human check before any important action.
Students ignore data	Ask: "What examples would the system learn from? Who or what might be missing?"
Older students dominate	Assign roles: user advocate, data lead, risk lead, presenter.
Ideas get too huge	Shrink to a first version that could be tested in one classroom or one team.

Vocabulary¶

User — the person a system is designed to help
Prediction — an informed guess about what may happen or what category something belongs to
Human in the loop — a person reviews or decides instead of letting AI act alone
Prototype — a rough version used to explain or test an idea
Failure mode — a way a system can go wrong

Reflection — fill in after class¶

Which pitch showed the best understanding of data?
Which group handled failure risk thoughtfully?
Did mixed-age roles work?
Save strong posters or ideas to reference during spring capstone planning.

Next week: Week 17 — Hour of Code Celebration