A4 AI in Practice

1. Retrieval (5 min). Revisit variables and lists in Python.
2. Worked example (10 min). Predict if an animal is a cat or a dog from height and weight, using a hand-drawn dividing line.
3. Practice (25 min). Plot a small dataset in Python and move the dividing line to see which points are classified wrongly.
4. Reflect (10 min). Define feature, label and prediction.

Do-it-yourself version for children

1. Retrieval (5 min). Features and labels.
2. Worked example (10 min). Run a provided nearest-neighbour classifier on a small dataset. (The dataset library and classifier are still to be built; see the status note above.)
3. Practice (25 min). Change the dataset, try different features, and print predictions for new items.
4. Reflect (10 min). What does the model do when it sees an item unlike any it was trained on?

The “Teach the machine” activity on the Try page uses this same nearest-neighbour idea, with 3 neighbours voting.

Do-it-yourself version for children

1. Retrieval (5 min). Why do we hold back test data?
2. Worked example (10 min). Show a model that scores 100% on training data and 60% on test data.
3. Practice (25 min). Split data, compute accuracy on both sets, and try a simpler model to reduce over-fitting.
4. Reflect (10 min). Explain over-fitting to a younger pupil in 3 sentences.

Do-it-yourself version for children

1. Retrieval (5 min). The fairness idea from A2 and A3.
2. Worked example (10 min). A pretend loan-approval or hiring dataset where one group is under-represented. Show different accuracy for different groups.
3. Practice (25 min). Count examples per group, compute accuracy per group, and propose fixes (more data, different features, human review).
4. Reflect (10 min). Write a short statement: “This model could be unfair to… because…”

Do-it-yourself version for children

1. Retrieval (5 min). Recall training and test data from Lesson 3.
2. Worked example (10 min). A tiny next-word predictor (see the code below).
3. Practice (15 min). Predict the next word by hand, then check with the program. Add more text and watch predictions change.
4. Investigate (12 min). Using teacher-provided example AI answers (printed or on screen, not live chat), mark claims as correct, unclear or invented and say how you would check each.
5. Reflect (8 min). Write three rules for checking an AI answer before trusting it.

from collections import Counter

text = "the cat sat on the mat the cat ate the fish".split()
next_words = {}
for current, nxt in zip(text, text[1:]):
    next_words.setdefault(current, Counter())[nxt] += 1

print(next_words["the"].most_common(2))
# prints: [('cat', 2), ('mat', 1)]

This program only counts which word follows which. It has no idea whether anything is true, which is the point of the lesson.

Do-it-yourself version for children

1. Starter (8 min). Pick a case card (face recognition at a shop door, AI-written homework, a hiring tool, a voice clone) and say what could go wrong.
2. Discuss (15 min). Structured debate: benefits, risks, who is affected, who is responsible. Include privacy, copyright and the right to opt out, and what guardrails (rules and filters that limit what an AI tool will do) can and cannot fix.
3. Write (20 min). A one-page reasoned argument and a personal AI code of conduct (for example, “I will say when AI helped me”).
4. Share (7 min). Read your code to a partner; badge.

Do-it-yourself version for children