AI and the Confidence Trick: Sounding Sure but Being Wrong

AI almost always sounds confident, like it really knows the answer. But sounding sure is not the same as being right. Even confident answers can be wrong.

Some examples

• AI might say a fact about a planet that is totally made up.
• AI might give a date that sounds right but isn't.
• AI might say 'definitely' about something it guessed.
• Always double-check facts in a real book or trusted website.

Try it!

Ask AI a fact you already know the answer to. Did it get it right? Now try a fact you don't know — would you check it before sharing?

If you asked a friend a hard question, they might say 'hmm, I'm not sure.' AI usually doesn't do that. It guesses anyway and sounds super confident, even when it's totally lost.

Some examples

• AI might invent a fake fact rather than say 'I don't know.'
• It might guess your hometown's mayor — wrongly.
• Newer AIs are getting better at saying 'I'm not sure.'
• You can actually ASK AI 'how sure are you?' and it might say.

Try it!

Ask AI a tricky question, then ask 'how sure are you?' See if it admits doubt.

Did you know AI sounds super confident even when it's making stuff up? It writes the same way for true facts and silly mistakes. The tone is a trick — not a clue.

Some examples

• AI might say 'definitely' about something fake.
• AI rarely says 'I might be wrong'.
• Fancy words can hide a wrong answer.
• Real experts say 'I'm not sure' more than AI does.

Try it!

Ask AI a tricky question and notice how sure it sounds. Then check the answer with a grownup or book.

Here's why "AI Sounds Sure Even When It's Wrong" matters: Learning about AI is one of the most important skills you can build for the future! AI uses a confident voice no matter what — that voice is not proof — and knowing how to apply this gives you a concrete advantage.

• Learn what "confidence" means and why it's important
• Learn what "tone" means and why it's important
• Learn what "fact-checking" means and why it's important

1. 1Find out more about AI Sounds Sure Even When It's Wrong by asking an AI a question about it
2. 2Talk to a grown-up about what you learned
3. 3Write down one new thing you learned today

When AI guesses something, it also gives itself a score: 'I'm 90% sure it's a dog.' That score is called confidence. But here's the trick — a high score doesn't always mean right. AI can be confidently wrong!

Some examples

• A photo app saying it's 95% sure your hot dog is a dog.
• A spell checker very sure your name is misspelled.
• A weather AI 80% sure of rain on a sunny day.
• A music AI sure you'll love a song you actually hate.

Try it!

Next time you see an AI suggestion (autocorrect, photo tag, song pick), ask yourself: 'Could it be confidently wrong?' Spot one mistake.