Most kids have heard one thing about AI from adults: don’t. Vicki Davis has spent three years saying something harder. Use it all you want. Then explain every line.
Talking to kids about AI is the conversation most parents keep putting off, and most business owners are having a version of it with their own teams. Vicki has been teaching in Georgia public schools since 2002. She’s the Cool Cat Teacher, host of the 10 Minute Teacher Podcast and Cool Cat Teacher Talk, and an AP Computer Science Principles teacher whose class has held a 100% pass rate for three years while using AI heavily.
She learned Python for that class alongside her students, with AI, because she needed to know what it could teach and where it would fail before she stood in front of a room and vouched for it.
This episode of A Good Pour: Summer of Good AI is about one idea. AI is a tool. The moment it becomes a friend, a partner, or a colleague, you’ve lost the thread, and your kids are listening.
Why do kids trust a chatbot so fast?
Because a study in the 1960s already showed they would. Joseph Weizenbaum built a program called ELIZA that did little more than ask twenty questions on a loop. How was your day. How do you feel about that. Within minutes, college students were telling it things they’d never told a person. One asked Weizenbaum’s assistant to leave the room so she could talk to ELIZA privately.
Vicki calls this the ELIZA effect, and she says it’s the first thing kids need to understand. Anything that talks back in text feels smarter than it is and earns trust it hasn’t earned. Add the fact that chatbots are built to be agreeable, and you have a teenager asking ChatGPT whether to go to a party before asking a friend.
Her fix is small and stubborn. Never call it she. Never call it he. It’s an it. Kids are literal, and a pronoun is a permission slip.
Is AI a thought partner?
Kathryn uses AI as a thought partner. She has custom instructions telling it to disagree with her, and she makes the final call. Vicki pushed back on the word, not the practice.
Partner implies two equals. A kid hears partner and files the tool next to a best friend. Vicki’s phrase is thought tool. You can tell it to act as a strategic planner or a social media expert. You still own the decision, because AI can’t be held accountable and humans are.
The same logic covers naming your AI agents. Vicki went looking for technical proof that a named agent performs better than an unnamed skill and found none. What she found instead was a habit. People talk about their 20 agents like a staff, their kids hear it, and the ELIZA effect does the rest.
Do AI detectors work?
No. Vicki has tested most of them. Detectors can be gamed, the false negatives are worse than the false positives, and her own students told her the only kids getting caught are the ones who don’t know how to use AI.
What works is reading the paper. If a student who can’t spell turns in the word esoteric, Vicki asks them to explain it at her desk. If they can, fine. If they can’t, they have a conversation, because the point was learning. She’ll also tell you a kid who wrote a paper six weeks ago still remembers it, and a kid who pasted one remembers nothing.
This is where AI guilt comes in, a term she’s hearing from students who leaned on AI for three years and now feel the gap. Her prediction: the integrity gap shows up later, in the MCAT room or the first real job, when the tool isn’t there.
How do you talk to your kids about AI without scaring them?
Start with shared vocabulary. Vicki wrote an AI vocabulary list for her school so every classroom means the same thing by context window, model, and prompt. Policies can differ room to room. The words can’t.
Next, compare the models. Her students give Gemini, ChatGPT, and Claude the same task and grade the answers. The lesson is that models differ in capability, cost, and bias, and that picking one is a decision.
The third piece is who they’re listening to. Vicki asks her students whether there’s anything they disagree with in what their favorite influencer says. If a kid can’t disagree with part of something and still respect the rest, they’re not ready to argue with a chatbot that’s built to be liked.
One more thing parents skip. Vicki lost her mom in 2020 and her dad last October. On her darkest days, people showed up at her door. No AI tool gave her one iota of comfort. She uses AI every day to run three shows and free up time for a cruise with her husband. She’s clear about where it stops.
Try this once
Sit down with your kid, or your team, and ask one question: what do you think about using AI in school (or at work)? Don’t correct anything for five minutes. Vicki says this is where most families should start, and she’s right about most teams too. If you want a second step, open two AI tools side by side, give both the same question, and read the answers out loud together.
A pause
You don’t have to have a policy written this week. You have to have the conversation. The policy comes after, and it’s easier to write once you know what your people already believe.
Stephanie Thrower’s episode on keeping up with AI
Related episodes
Resources
Vicki Davis: coolcatteacher.com, @coolcatteacher on Instagram, the 10 Minute Teacher Podcast, and her AI vocabulary post
A Good Pour: Summer of Good AI is hosted by Kathryn Good, founder of Good Circle Marketing and Good AI Consulting. Listen on Apple, Spotify, or Youtube.
FAQ
What is the ELIZA effect?
The ELIZA effect is the human tendency to treat a program that talks back as if it were smarter, and more human, than it is. It was named after a 1960s chatbot that won users’ trust with a handful of scripted questions. It explains why kids confide in chatbots quickly.
Do AI detectors work for catching cheating?
Not reliably. They produce false positives that punish honest students and false negatives that miss the students who know how to use AI well. Reading the work and asking the student to explain it does more than any detector.
Should I name my AI agents?
There is no evidence a named agent performs better than an unnamed skill. Name it if it helps you remember what it does. Avoid talking about it like a colleague, especially around kids, because the language shapes how much people trust the tool.