← Back to Academy
Module 1 · Sometimes AI Gets It Wrong — Introduction | AESOP AI Academy Module 3
Color
Intro
Module Test
Lesson 1

AI Makes Mistakes

Even the smartest tools get things wrong sometimes.

Mia wanted to bake her grandmother's favorite cookies. She typed the recipe into an AI helper and asked: "How many cups of sugar do I need?"

The AI answered instantly: "You need three tablespoons of sugar." But Mia's grandmother always used one cup. The cookies came out flat and barely sweet.

Later, Mia realized the AI had mixed up tablespoons and cups. It sounded completely sure of itself — but it was wrong. It didn't know it had made a mistake.

Why AI Makes Mistakes

AI is a tool — like a calculator or a map app. Tools are helpful, but they can give wrong answers. AI gives answers based on patterns it learned, and those patterns are not always right.

  • AI can confuse similar words or measurements.
  • AI can give an outdated answer if things have changed.
  • AI can misunderstand what you asked.
How Much Should You Trust It?

Trusting AI is about balance. Here is a simple rule: the more important the decision, the more you should double-check.

Key Idea

AI can be really useful, but you still have to think for yourself. Always ask: "Does this answer make sense to me?"

Quiz 1

AI Makes Mistakes

3 questions — free, untracked, retake anytime.

Mia asked the AI for a recipe measurement, what did the AI actually do wrong?

✓ Correct — ✓ Right! The AI mixed up the measurements. It sounded sure of itself but was still wrong.
❌ Not quite. The AI did answer — but it mixed up tablespoons and cups.

of these is the best example of when you should double-check an AI answer?

✓ Correct — ✓ Yes! Medical decisions are very important — always double-check with a real doctor or pharmacist.
❌ The more important the decision, the more you should verify. Medical questions are the most important to double-check.

makes mistakes because:

✓ Correct — ✓ Exactly. AI learns from patterns — and sometimes those patterns lead to wrong answers. It's not lying; it just doesn't know it's wrong.
❌ AI isn't trying to trick you — it answers based on patterns it learned, and those patterns can be mistaken.
Lab 1

Think It Through

Talk with an AI guide about when you'd trust an AI answer — and when you wouldn't.

Lab 1 — Trust Check

Your AI guide will ask you about times when you might trust an AI — and times when you shouldn't. Answer in your own words.

  1. The AI will ask you a question about trusting AI answers.
  2. Share your thinking — there's no single right answer.
  3. The AI will help you think a bit deeper.
Think about Mia's story. When is it okay to trust AI? When should you always check?
⭐ AI GuideLab 1
Lesson 2

When AI Doesn't Know

Sometimes AI makes up an answer — and sounds very confident.

Tomas was doing a school project. He asked an AI: "Can you name a book about ocean animals by Dr. Sarah Bloom?"

The AI replied: "Sure! Dr. Sarah Bloom wrote The Deep Blue Encyclopedia in 2018. It has wonderful chapters on dolphins and deep-sea fish."

Tomas was excited — until he searched the library. The book didn't exist. Neither did Dr. Sarah Bloom. The AI had invented the whole thing.

What Is Hallucination?

When AI confidently invents something that isn't real, it's called a hallucination. The AI doesn't know it's wrong — it just produces the next likely-sounding word.

  • Hallucinations happen most with specific facts: names, dates, book titles, quotes.
  • They can sound very confident and convincing.
  • You can usually catch them by checking another source.
Remember

When AI gives you a specific fact — like a book title, a person's name, or a date — always check it somewhere else before trusting it.

Quiz 2

When AI Doesn't Know

3 questions — free, untracked, retake anytime.

happened when Tomas used the AI for his school project?

✓ Correct — ✓ Right! The AI made up a book title and an author. It sounded totally real but was completely false.
❌ The AI invented the book and the author — they never existed at all.

is an AI "hallucination"?

✓ Correct — ✓ Yes! A hallucination is when AI confidently invents facts, names, or details that aren't true.
❌ A hallucination is when AI states something false as if it were completely true — with full confidence.

the best thing to do if AI gives you a specific fact, like a book title or a name?

✓ Correct — ✓ Always verify specific facts from a second source. AI can sound confident while being completely wrong.
❌ The safest move is to check specific facts somewhere else before trusting them.
Lab 2

Fact Checker

Talk with your AI guide about what happened to Tomas and what you would do.

Lab 2 — Hallucination Lab

Your AI guide will ask you about Tomas's story and what you'd do if AI gave you a made-up fact.

  1. The AI asks you a question about the story.
  2. Share your thoughts.
  3. The AI will follow up with one more question.
Think about: have you ever trusted something that turned out to be wrong? What did you do?
⭐ AI GuideLab 2
Lesson 3

Whose Fault Is It?

When AI gets something wrong, who is responsible?

The Reyes family used a navigation app on a road trip. The app told them to turn onto a mountain road. What the app didn't know: that road had been closed for months after a landslide.

The family ended up stuck for two hours. Afterward, they wondered: whose fault was it? The app? The company that made it? Or themselves for following it without checking?

Who Is Responsible?

When AI causes a problem, responsibility can be shared:

  • The person who used it — chose to follow the AI's advice.
  • The company that built it — is responsible for keeping it updated and accurate.
  • The AI itself — is just a tool; it can't be blamed the way a person can.
Key Idea

AI tools can make mistakes. The people who build them and the people who use them both have a responsibility to think carefully.

Quiz 3

Whose Fault Is It?

3 questions — free, untracked, retake anytime.

the Reyes family story, who might share responsibility for what happened?

✓ Correct — ✓ Responsibility is often shared. The company should have updated the maps; the family could have double-checked a big route change.
❌ In most AI situations, responsibility is shared between the people who built it, the people who use it, and sometimes others too.

an AI tool be "blamed" for making a mistake the way a person can?

✓ Correct — ✓ AI is a tool — it can't be blamed like a person. The people who create and use AI are responsible for what it does.
❌ AI tools can't be morally responsible — that responsibility belongs to the humans who build and use them.

is it important to understand who is responsible when AI makes a mistake?

✓ Correct — ✓ When we understand responsibility, builders work harder to make AI accurate and users think more carefully before trusting it.
❌ Understanding responsibility helps both builders and users make better, more careful decisions about AI.
Lab 3

Who's Responsible?

Share your thoughts about the Reyes family story with your AI guide.

Lab 3 — Responsibility Lab

Your AI guide will ask you who you think is most responsible when AI makes a mistake.

  1. The AI opens with a question about the story.
  2. Share your opinion.
  3. The AI will ask you to think about it from another angle.
There's no perfect answer here. What matters is that you think it through.
⭐ AI GuideLab 3
Lesson 4

Bias In, Bias Out

What happens when the information AI learned from has a blind spot?

Amara noticed something odd: whenever she typed her name on her phone, autocorrect kept trying to change it. But her friend Emily's name never got changed.

Her older sister explained: "The autocorrect learned from a lot of writing. If most of that writing used certain kinds of names, it learned to expect those names and treats others as 'wrong.'"

"So it's not being mean," Amara said slowly. "It just doesn't know names like mine because of what it was taught?"

"Exactly," her sister said. "Bias in, bias out."

What Is Bias in AI?

AI learns from information that humans created. If that information shows some groups differently than others, AI will learn and repeat those patterns.

  • If most training photos of doctors show men, AI may assume doctors are male.
  • If spell-check learned mostly certain names, it may flag other names as wrong.
  • This isn't AI being mean — it's AI repeating what it was taught.
Big Idea

The information AI learns from matters a lot. When that information leaves people out, the AI will too.

Quiz 4

Bias In, Bias Out

3 questions — free, untracked, retake anytime.

did Amara's autocorrect keep trying to change her name?

✓ Correct — ✓ Right! The AI learned from a limited set of names and treated unfamiliar ones as errors. That's bias.
❌ The AI wasn't broken or mean — it learned from text that mostly used different kinds of names, so it didn't recognize hers.

does "bias in, bias out" mean?

✓ Correct — ✓ Exactly! If the training information has gaps or unfairness, the AI repeats it. The output can only be as good as what went in.
❌ "Bias in, bias out" means the AI repeats the patterns it learned — including unfair or incomplete ones.

is responsible for making sure AI learns from fair, complete information?

✓ Correct — ✓ The people who design and build AI are responsible for the information it learns from and the patterns it picks up.
❌ The builders of AI have a special responsibility to make sure the training information is fair and includes everyone.
Lab 4

Thinking About Bias

Talk through Amara's story with your AI guide.

Lab 4 — Bias Lab

Your AI guide will ask you to think about Amara's experience and where you might have noticed AI treating things differently.

  1. The AI opens with a question about Amara's story.
  2. Share your thoughts and experience.
  3. The AI will ask you what could be done to fix it.
Have you ever noticed a phone, app, or computer treat you or something about you differently than expected?
⭐ AI GuideLab 4

Module 3 Test

5 questions covering all 4 lessons. Free, untracked, retake anytime.

AI gave her the wrong measurement. What does this show?

✓ Correct — ✓ AI makes mistakes — good thinkers always ask if the answer makes sense.
❌ AI can still be useful, but it isn't always right. We need to think critically.

AI invented a book that didn't exist. This is called:

✓ Correct — ✓ Hallucination: AI confidently invents facts that aren't real. Always check specific facts from another source.
❌ An AI hallucination is when it confidently invents details — names, dates, book titles — that don't exist.

AI gives you a specific fact like a name or date, what should you do?

✓ Correct — ✓ Always verify specific facts. AI can sound confident while being completely wrong.
❌ The safest approach is to verify specific facts somewhere else first.

the Reyes family got stuck, responsibility belonged to:

✓ Correct — ✓ Responsibility is shared. Builders must keep AI updated; users must think before trusting it completely.
❌ Responsibility is shared between those who build AI tools and those who use them.

autocorrect kept changing her name because:

✓ Correct — ✓ Bias in, bias out. The AI repeated the patterns it learned — which didn't include many names like Amara's.
❌ The AI wasn't broken or mean — it simply learned from text that didn't include names like Amara's.