The Robot Who Was Never Wrong: An AI Safety Story for Children About Confident Chatbots and Fact-Checking (Ages 7 to 11) (Paperback) (ISBN-13: 9798187408412)

Otto had been right nine times in a row. Then Otto invented a horse.

The story

At the back of THE BIG BOOK OF IS THAT TRUE is a page Maya Mistry is quietly proud of. It says OTTO across the top, with two columns, RIGHT and WRONG, and a forest of tally marks.

Lately the RIGHT column has been winning. Otto knew the capital of Peru. Otto said ducks can sleep with one eye open, which sounded exactly like the sort of thing Otto would make up, and turned out to be gloriously true.

Then the school announces its first Kettlebridge History Fair, and Mr Field writes the brief on the board: ONE AMAZING TRUE FACT.

So Maya and Dev ask Otto about the canal. And Otto tells them about Duchess, a famously strong horse who pulled 40 boats at once and was awarded a silver medal by the king.

Otto is absolutely sure. Otto says the medal is 100 percent genuine and the story is celebrated by canal historians everywhere.

There was no Duchess.

The tell

A sure voice is not a source. Otto says everything in the same certain voice, when it is right and when it is making things up, so the voice proves nothing. Only checking proves things.

Being right before does not make it right now. Nine correct answers is a coin coming up heads nine times. It tells you nothing about the next flip.

And you cannot check the machine by asking the machine. "Are you sure?" only gets you the sure voice again. That is asking the seller whether the watch is real.

The two tests

Ask again. Real facts stay put. The market is on Saturday every time you ask. Made-up answers wobble, new names and new numbers each time. If the story changes when you ask twice, it was never a fact.

Then ask where the fact lives. Every real fact has a home: a book, a photograph, a newspaper, an archive box, a person who was there. If nobody can say where a fact lives, it does not.

What your child will be able to do

Explain that an AI can sound certain and still be wrong or make things up, and name at least one way to double-check a claim.

Curriculum

Aligned to the National Curriculum for England, Computing at Key Stage 2 (collecting, analysing, evaluating and presenting data and information) and statutory Relationships Education. Maps equally to the Curriculum for Wales Digital Competence Framework (Citizenship, and Data and computational thinking), Scotland's Curriculum for Excellence (Technologies TCH 2-02a, not all information is credible), and the Northern Ireland Curriculum (Using ICT, the Evaluate strand). US grades 2 to 5, elementary, Australia and New Zealand Years 3 to 6. Not endorsed or accredited by any government body.

Reading it

Ten chapters, about 7,800 words, ages 7 to 11, with Maya's Safety Spotter Tips at the back. A wrong fact gets out, and the book is about putting the right version where the wrong one went. Wrong is ordinary. Staying wrong is the mistake.

Good for

AI for kids and chatbot literacy at the age children start asking machines their homework. ChatGPT for kids explained without panic or sales talk. Media literacy and critical thinking for kids, in a story about a school history fair. Online safety for kids and cyber safety for kids where the risk is not a stranger but a confident wrong answer nobody checked.

Kids chapter books ages 9 to 12 for a child who argues with everything. Also for any adult who has pasted a confident answer into something important.

The series

Book 4 of 6 in Safety Spotters, for ages 7 to 11. Younger readers start with Little Safety Spotters, ages 5 to 7. Older readers continue with Safety Spotters: Case Files, ages 11 to 14.

Free classroom resources and the fact-checking worksheet at safetyspotters.com

Check it before you trust it.