🎯 What you will learn
- Bias starts with the examples
- Models answer even when they should not
- Checking results and sources protects people
👋 Let’s get started!
Every example a model learns from teaches it something. If the examples are missing a group or a case, the model cannot know about it, but it may still answer confidently. Let’s see this happen! ⚠️
🪜 Step by step
1. 📚 Biased examples
If every example is a dog, the classifier will call everything a dog.
examples = [(1, "dog"), (2, "dog"), (3, "dog")]2. 🤖 Confident mistakes
The model still gives an answer, even when it has no good example.
print(guess(50))3. ✅ Check and fix
Add more and better examples, and check results against trusted sources.
examples.append((40, "cat"))🧪 Try it!
Change the code and press ▶ Run to see what happens. You can’t break anything!
Only dogs
examples = [(1, "dog"), (2, "dog"), (3, "dog")]
def guess(value):
best_gap = 100
best_label = None
for number, label in examples:
gap = abs(number - value)
if gap < best_gap:
best_gap = gap
best_label = label
return best_label
print("Guess for 50:", guess(50))
examples.append((40, "cat"))
print("After one more example:", guess(50))🎨 Unplugged activity
⚠️ Try an AI tool with a question about your own school or town. Check its answer with two trusted sources. Write down one thing it got right and one thing it got wrong.
Find the missing data
examples = [(1, "dog"), (2, "dog")]
print("Trained on:", set(label for _, label in examples))📖 Key words
Tap a card to flip it over!
- Bias
- Unfair patterns from the data ⚖️
- Hallucination
- Confident but wrong output 👻
- Source
- Where information comes from 📰
🏆 Challenge
Add an example (50, “bird”) to the list, then print guess(48). Press ▶ Run.
Challenge
examples = [(1, "dog"), (2, "dog")]
def guess(value):
best_gap = 100
best_label = None
for number, label in examples:
gap = abs(number - value)
if gap < best_gap:
best_gap = gap
best_label = label
return best_label
Knowledge Check Quiz
- Where does a model’s bias come from?
- Why can a wrong AI answer still sound sure?
- What should you do with an AI answer?