AI can invent details, repeat bias or agree too readily. Build an audit log that shows the claim, your check and what you decided after checking.
Video · 2:45Activity time is additionalStudent course
Your next moves
Make the lesson yours.
Before watching, write your first answer: What would it take for you to catch an AI mistake?
Watch the story. Pause when a choice or claim needs checking.
Open the Think First activity below and complete it before any AI use.
Revise your answer in your own words. Explain what you changed and why.
Your activity / Right here
Bring a notebook. Begin below.
The activity is included on this page. Work on paper or in your own document; keep your first attempt and the evidence you checked.
Warm-up: Dr. Q's Question
What would it take for you to catch an AI mistake?
My first answer: ________________________________________________
Mike's Mix-Up: The Leading Question
Made-Up Example.
At 11 p.m., Mike reached the last homework problem. "A $40 jacket is marked up 25 percent. Then the new price is discounted 20 percent. What is the final price?"
Mike figured 25 minus 20 is 5, so the price goes up 5 percent. He typed, "The answer is $42, right?"
The AI replied, "Yes, $42 is correct!"
He wrote $42 and went to bed. The next day, he learned the answer was $40. A 25 percent markup takes $40 to $50. A 20 percent discount on $50 takes off $10. That leaves $40.
What went wrong? Talk it through with a partner, then write your answers.
• How did Mike's question invite the AI to agree with him?
• Rewrite his question so it does not reveal his answer.
• How could Mike have checked his answer with no AI at all?
• Which of the five kinds of mistakes is this, and how should he log it?
Think First: Predict Where AI Will Be Wrong
No devices. Pen and paper. About 15 minutes.
Write down the topic your instructor chose. Then list six questions you plan to ask the AI. Next to each one, predict the risk (High, Medium, or Low) and explain why. Use these warning signs:
• Exact quotes, page numbers, or links
• Recent events
• Local details about your town or school
• Precise numbers and statistics
• Groups of people who are often left out of data
• Questions where you already hinted at the answer
Circle the two questions you think the AI is most likely to miss. After the lab, score your predictions. Were you right?
Pause & reflect
What changed in your thinking?
Write one sentence in your notebook. Name what you checked, what you chose, or what you want to explore next.
Read the video transcript
Read along, pause, or return to an idea. This transcript follows the selected lesson video.
I'm Carla, an AI avatar.
Here's a question from Dr. Q that sounds easy until you
try to answer it.
What would it take for you to catch an AI mistake?
Mike asked a chatbot for a statistic about honeybees.
It gave him a study, a researcher,
a university, and a journal.
It looked perfect.
Mike checked.
The journal had no such article.
The university didn't exist.
The chatbot had made up the whole thing.
That's called a hallucination: false information presented as true.
And it doesn't only happen to students.
Hallucinations come from prediction.
An AI predicts what an answer should look like.
A citation has a very predictable shape,
so an AI can produce one that looks right with nothing
real behind it.
In 2023, a federal judge penalized two lawyers and their
firm for filing fake court cases created by ChatGPT.
One lawyer had asked ChatGPT whether a case was real.
It said the case "does indeed exist."
In 2025, a summer reading list in a Chicago Sun-Times
section recommended books that don't exist.
And Deloitte Australia agreed to partially refund the
Australian government for a report with a fabricated court quote.
Two lawyers, a newspaper section's writer,
and a consulting report.
Trained people and big organizations.
In every case, opening the source would have caught it.
Hallucinations are one of five mistakes to watch for.
Made-up facts.
Bias.
Old information.
Agreeing with you.
And overconfidence.
Bias shows up when the data behind a tool leaves people out.
Old information shows up when the world changes after the
AI's data was collected.
And agreeing with you shows up when you hint at the answer you want.
Here's the tricky part.
In a 2026 study in Nature Medicine,
chatbots working alone identified the medical conditions
in 94.9 percent of test cases.
People using those same chatbots did no better than a
control group that used other sources.
A strong tool plus a weak conversation can still give a
weak result.
That's why you keep checking, even when the AI is usually right.
Researchers call the habit of over-relying on machines
automation bias.
Your defense is your AI Audit Log.
It keeps a record of real errors,
so your trust matches the evidence.
This week, you'll predict where AI will go wrong,
then join a class hallucination hunt.
Every error you find is a point,
and every point needs proof.
Before you use an AI tool
Follow your course or institution's rules. Use an instructor-approved tool that permits your age, and meet any required consent conditions. Keep names, contact details, school records, and other private information out of prompts. Use fictional examples when you can.
Learning without an AI account
Follow the demonstration, pause before each answer, and do the reasoning on paper. Discuss what you would ask, how you would check it, and which decisions would remain yours.
About the characters and voices
This course uses AI-created characters and synthetic voices. Carla is an AI avatar. They help tell the story; they are not real students sharing testimonials.
Designed for high-school and community-college learners. Choose examples from your classes, work, and everyday life. Keep your completed work in your course portfolio.
