Ask HN: Any Creative AI Uses That Work with Kids?
Quick example: My sibling took the kids to a dinosaur expo. The oldest was scared and spent the day in a quiet sensory corner. The younger ones ran around roaring at life-sized dinos (they sent this in the family group chat with a “I have totally different kids” message along with some photos and videos of the day).
That day, I had been playing with SUNO a lot and tossed the whole thing into GPT. I asked for a cheeky nursery-style song about how “dinosaurs are your friends” (with each kid’s name), then ran it through SUNO.
The result? The oldest (who wanted nothing to do with the dinos) now asks for “the Dino Song Uncle made” every morning and wants to go back to the expo.
It got me thinking: if you can spin up a quick, personalized story or song on the fly, you can comfort, teach, or defuse a meltdown in minutes. But… would kids start gaming it just to get another song? (Probably, lol)
Anyone else tried clever AI workflows with kids? Songs, stories, games, emotional coaching? What worked, what flopped?
I have zero background in early childhood education/development or any sort of psycho-education knowledge, but I obviously can see how this type of workflow would have to come with serious guard rails and probably several studies and trials if it was to be used in any formal education system. Like I am not recommending anyone who works with children to start experimenting with this stuff without the right approvals / consulting the appropriate specialists.
I have done voice ai chat. You can have the ai play twenty questions or "what am I thinking of"
You can have the AI create adventure stories with interaction.
"you as a dinosaur hear a rustling in the bushes, do you A, go check the bushes or B, ignore it."
Then it takes the kid on a self-constructed story adventure.
Ive generally not had success with "Two truths and a lie". Telephone game didnt work because the chat ai just blurts it out loudly.
Early childhood development and storybooks is all over the place. The most successful books for kids are typically created by well-informed individuals or artist/writer combos who are better at the creative side than the education side but they're still pretty good at both.
The work of Julia Donaldson stands out as a leader in the field. She's responsible for the Gruffalo, The Snail and the Whale, Tiddler, and loads more. She's also got a series called Songbirds that are far more basic in their prose, but designed for early reading education. It's in those books that you'll see more method to the madness, with themed consonant and vowel pairings to hammer home elemental lessons for the youngest minds.
AI is chewing up a vast corpus of data that may contain snippets, tracts, or entire children's books, as well as studies and advice on child education. Its output will be a best match guess at what you prompt it, but it is unlikely to be a focussed educational tool like the Songbirds series is. Simply put: AI's strength will be a quick creative story crafted to a child's personal interests that can compete with even the most popular storybooks, but it might lack the nuance of stage-specific educational content.
That being said prompts can go a long way if you're prepared to do your own research and structure your instructions to narrow the output enough.
Another thing to be careful about is the pragmatic information in source data that will lack the weightings the AI needs to use appropriately. Things like dialect. If you use a storybook to teach vocabulary and pronunciation you can run into spelling and rhyming issues and more. We're British and read books from US relatives to our toddler and a book about Maine might try rhyming "on" with "dawn". In the UK that makes mum and dad sound like a rural country type from the South East, which neither of us are. The AI will be aware of phonetics, but is unlikely to weight it highly enough to produce your accent-specific rhymes, especially if they're not standard.
Another trap is to actually craft your kid's content to their interests, at least early on. Creative content is better opening their experience up as broadly as possible. But that's easy to mitigate. If they like dinosaurs, but their book on counting features mermaids, then use AI to make a dino-themed counting book. Just don't make everything dino-themed.
Extra aside - bedtime stories purposefully feature themes around sleep, contain more 'shhh' sounds, and repetitive sequences. Quite literally designed to send kids to sleep. But that muddies the AI waters when you want an output that engages the reader for educational purposes.