“Be a Pain”: Talking to young children about AI chatbots
There are a host of questions artificial intelligence raises about education. This essay addresses one: how should we explain chatbots to young children?
As a teacher and parent, I have spent a lot of my time socializing young children into their relationships with different parts of the world. Take dogs. Through my actions, stories, and pronouncements (e.g., “That looks like a friendly dog but before we pet it we need to ask its person if that is O.K.”), I have helped shape my children’s feelings and behaviors toward canines. Or TV. When my sons were young, rather than having it on as a constant companion, we did not have a television at all. My reason, as I explained to them, was that watching too much TV, “Will make your brains leak out.” Or people of various religions and cultures. Both at home and school, I talked about how lucky we are to live in a city with people from all over the world. Of course, there are influences beyond parents and teachers (e.g., TV and social media), and adult efforts at socialization are not completely deterministic. Now grown, both my sons have TVs in their homes. At the same time, they both live by an ethic of equity and inclusion.
Enter something new to socialize about: AI, and in particular, chatbots, the Large Language Models that loom large in their futures. The framing of this technology, how we talk about it, will shape how children feel and behave towards it. In a study published in Early Education and Development, researchers varied the way they talked about a robotic dog to five-year-olds. When the children were told a human was controlling the robot, children treated it like a machine or toy (without rights or feelings). When the robot was described as autonomous, children talked about it having feelings and deserving of moral treatment.
In this essay I explore how we should talk to young children about AI: what it is, where it “lives”, how it might be helpful, and the problems it is causing. I begin with what an AI chatbot, Gemini, recommends on the subject. Then I share conversations I had with children at Newtowne about chatbots. I end with my answers, grounded in a desire to help children come to:
- A clear understanding of what AI is and what it is not;
- An exploratory, playful mindset for interactions with AI; and
- A healthy skepticism of AI companies.
My “conversation” with Gemini
Gemini is Google's version of Chat GPT. Trained on a massive data set of text and other media, Gemini is, according to Gemini, “Designed to be a helpful, conversational companion.”
I was curious about how Gemini would explain AI to young children. Here I share our “conversation” without commentary. See what you think about the bot’s recommendations. Know that I agree with some of the pronouncements and strongly disagree with others. In all, I am reminded of my friend Deborah Stier’s quip that, “AI is really smart unless you know about the subject.”
To start I told Gemini that I teach five-year-olds and asked how I could explain to my children what Gemini and other AI chatbots are. Gemini answered enthusiastically:
I then asked Gemini something the Green Dragonflies (the five-year-olds at Newtowne) had been wondering:
I was curious about the term “the cloud,” so I asked:
My final question:
Gemini continued:
AI is like a super-fast helper machine. Imagine if you had a robot helper that could look through a giant toy box and find your favorite lego brick in one second! It helps doctors find out why people are sick so they can make them feel better. It helps grown-ups write emails faster so they can finish work and come play. And if we want to make up a silly bedtime story about a blue banana, it can help us think of funny ideas right away!
Regarding problems that Gemini and other chatbots are causing, the response was:
When explaining the downsides, avoid abstract concepts like "data privacy" or "economic displacement." Instead, focus on mistakes (lying), fairness (taking credit), and laziness (not using your own brain).
It is clear that mine and Gemini’s goals regarding young children and AI differ. Certainly Gemini does not have the desire to help children develop a healthy skepticism of AI companies.
My conversations with the Green Dragonflies
During their final weeks at Newtowne, six groups of Green Dragonflies (the graduating five-year-olds) came to the studio for lunch. As part of our conversation I asked the children about different chatbots. I also proposed a game where one child pretended to be a bot and the other children and I asked him or her questions.
I have four takeaways from these conversations:
The children have limited exposure to chatbots (but they do have opinions)
Cambridge, Massachusetts (USA), where Newtowne School is located is, along with Silicon Valley and Beijing, one of the hubs of AI development (Cambridge is home to MIT). So it is striking that five of the sixteen Dragonflies, almost a third, said they have never heard of Chat GPT, Claude, Gemini, Alexa or Siri. (note: I asked about Alexa and Siri because, even though their underlying technology differs from their chatbot cousins, children experience them in similar ways). Elliot and Finlay’s exposure was typical:
Ben: Have you ever heard of Alexa?
Elliot: No.
Finlay: The Alexa who is good at playing music?
Ben: Yeah. Have you ever heard of Siri?
Finlay: No.
Elliot: I have heard of Harry. Harry Potter.
Ben: I’ve heard of Harry Potter also. Let’s stay with Alexa for a while. Do you have Alexa at your house Fin?
Finlay: No. My friend’s house has Alexa.
Ben: What does Alexa do?
Finlay: She tells the news and she plays music.
Like Finlay, all of the children's experiences with chatbots was second hand, seeing adults or older children interact with them. The children who had encountered AI explained that it could help in some way: playing music, relaying the news, or, as Carter and Lucia explained, helping their parents with Minecraft:
Ben: I have a question: have you ever heard of something called Chat GPT?
Emilia: No
Carter: Oh yeah. Yeah, yeah, yeah. Want me to tell you?
Ben: Sure
Carter: My dad uses him for Minecraft.
Lucia: My mom uses it for Minecraft also. But only my mom.
Ben: Have you ever heard of Siri or Alexa?
Carter, Emilia, Lucia: I have, I have, I have.
Lucia: My cousin told me about it.
Ben: What did she tell you about it?
Lucia: She has a little thing that is round and it can talk but sometimes it disappears in her bedroom. It helps her to put songs on it. She is also nine and then she is going to be ten.
Eli explained that the bots can also answer questions:
Ben: Have you ever heard of Chat GPT?
Eli: Oh yeah.
Ben: What is it?
Eli: It's this thing that you talk to and it tells you things. Like we found this circle thing and my mom asked Chat GPT and it said it was a Fort Lauderdale avocado because it was a Fort Lauderdale avocado. We were in Fort Lauderdale and it's from there.
But how chatbots function is a mystery to the children. It seems they had not had conversations about what AI is. Eli’s answer was typical:
Ben: How does Chat GPT know the answers?
Eli: I don’t know.
Ben: Do you know where it lives?
Eli: No.
This mystery means children are open to misunderstandings, especially around the question of chatbot personhood.
The children are ripe for conversations about chatbot personhood
Some of the Newtowne children think AI is a person. Others say it is not. Some are unsure. Carter and Lucia illustrate this range:
Ben: When you talk about Chat GPT Lucia says “it” and Carter says “ him.” Is it a person or a thing?
Carter: I don’t know.
Lucia: Me neither.
Ben: Can you talk to it?
Carter: Yeah, and it talks back.
Lucia: Yeah, it talks back.
Carter: I think that it is a real guy and he has a speaker inside the phone so he can talk.
Ben: So you think that it is a person.
Carter: Yeah.
Carter is less sure about Siri:
Ben: Is it a person? Is it a robot? Is it a machine?
Carter: I don’t know.
Because of this mystery, children are developing fantastical theories of where chatbots live. The idea of “the cloud” has found its way into the children’s construction of chatbots:
Ben: Eli, you said Alexa lives in the sky.
Eli: Yeah, she does. She lives 100 feet up in the sky.
Ben: Some people say they live in the cloud.
Eli: Yeah, they live in the cloud. In the cloud.
Ben: She really lives in the sky?
Eli: Yeah, do you know how I know?
Ben: How do you know?
Eli: I asked her.
Noa: You can see her if you were a bird.
Eli: What I would do is I would find a unicorn with wings and get on its back and it would take me up to the sky and see Alexa. Or I would find a really, really tall tree and I would climb up it and then I would say hi to Alexa.
Noa: Alexa lives in two clouds.
Ben: How can she do that?
Noa: Clouds connect.
Given the children’s confusion about AI personhood, Gemini’s advice that we need to help children understand that it and other chatbots are not alive rings true.
Play is a useful way to learn about what children think about AI
In my conversation with Daphne she seemed to reach a limit in how much she was able (or willing) to explain about AI:
Ben: Have you ever heard of Claude?
Daphne: Daddy has Claude on his computer. That’s his computer’s name.
Ben: Is Claude a person? What does Claude look like?
Daphne [in an exasperated voice]: No. You can just come to my house and see.
In my conversations many of the children seemed to reach a point where they tired of talking about the bots.
With Vygotsky's insight that children stand a head taller from their normal development in play, I began inviting children to pretend they were Chat GPT (or Claude or Alexa). Here are Phillip, Eli, Tomo, and Finlay’s portrayals of chatbots [note: Eli and Tomo were in the studio together while Phillip and Finlay came on separate days]:
Ben: I thought of a fun game. Phillip, I was wondering, could you be Siri? We could ask you questions and tell you things to do. Would you like to give it a try?
Phillip: Yeah, I guess.
Ben: Great. [to Gavin and Ellery] What are you going to ask Siri?
Ellery: Siri, can you sing us happy birthday?
Phillip: Happy birthday to you. You're 102. You look like a monkey, and you smell like one too.
Ben: Oh, a teasing Siri.
Phillip [speaking in a robot voice]: No, that’s the only happy birthday song I know.
Ben: I have a fun game. You can pretend to be Chat GPT and we’ll ask you questions. Do you want to pretend to be Chat GPT?
Eli: Yeah.
Ben: Excuse me, Chat GPT, can you sing happy birthday for me?
Eli: Happy birthday to you. Do you want me to say 102 in it?
Ben: Yeah.
Eli: Do you want the funny version or the not funny version?
Ben: The funny version.
Eli: [singing] Happy birthday to you. You're 102. You still go to preschool and now you're turning eight. It's so weird. My dad took a shower. Now he feels bad.
Ben: Chat GPT, that was very funny. I have another question for you.
Eli: What is it?
Ben: Who are the Orange Sea Star teachers?
Eli: Nana and Kari [note: Nana and Kari were the Sea Star teachers when Eli was in that class. Danielle has since replaced Kari]
Tomo: And Danielle.
Eli [speaking in robot voice]: And Danielle. Good job Tomo.
Eli: Pretend I’m Alexa.
Ben: OK.
Eli: So I will call you by your name.
Ben: Sounds good.
Eli: Say Alexa, add to my shopping list.
Ben: Well Alexa, what should I have for dinner tonight?
Eli: The best thing you can have for dinner is salmon in hot dog buns.
Ben: Thank you Alexa.
Ben: Tomo, who do you want to be? Chat GPT? Alexa?
Tomo: Chat GPT.
Ben: OK, we have to ask Tomo/Chat GPT questions.
Eli: Chat GPT, how do owls fly?
Tomo: Owls, they use their wings to fly.
Eli: But like how do they get up in the sky? Maybe when they flap their wings the wind blows them up high?
Tomo: [speaking in a robot voice]: Yes, up high.
Ben: We have been doing this game where kids have been pretending to be Alexa. Would you like to give it a try?
Finlay: Yeah.
Ben: Great. So Fin is going to be something called Alexa and you can ask it questions.
Finlay: Or you can play some music.
Ben: Alexa, can you sing happy birthday for us?
Finlay: Happy birthday to you. I’m a 102. I look like a monkey, but I live in the zoo.
Ben: Alexa, what should I have for dinner?
Finlay: Pancakes.
Ben: OK, that’s a good answer. Alexa, how did you know what I should have for dinner?
Finlay: Because I love pancakes.
I am struck by how good the children are at being chatbots, swiftly answering the questions poised to them. As bots they mostly assumed human personas. But not completely, occasionally using robot-like voices. Eli did this after he provided an incorrect answer (chatbots can be wrong), praising Tomo for correcting him. The children’s fanciful suggestions of what I should have for dinner make me smile.
Indeed, everyone had a good time playing the game:
Ben: Was it fun to be Alexa?
Eli: Yeah.
Ben: Tomo, how did it feel to be Chat GPT? Was it fun? Hard?
Tomo: It was easy!
Ben: Fin, how did it feel to be Alexa?
Finlay [thumbs up]
Ben: Why did you like it?
Finlay: It’s fun.
The children have worries about AI (which mirror mine)
Well, not everyone had fun. Not Phillip. Phillip is particularly sensitive to issues of power and control. When I asked him what he would ask Siri he replied:
Phillip: I’d ask it to be my servant.
Ben: Be your servant! What will it do then?
Phillip: I would tell it to make me a peanut butter and jelly.
In no way a malevolent use of AI. But what about other people? When I asked Phillip how it felt to be Siri he answered:
Phillip: A little bit scary.
Ben: Why was it scary?
Phillip: Because maybe bad questions could happen.
Ben: That is a very thoughtful thing to say. That would be a problem.
A problem indeed. Phillip has a sense that bad actors might misuse this powerful technology. I have this same worry. Tech companies' past behavior, for example, continuing to use algorithms that promote addictive use in the face of evidence that this was detrimental to children’s mental health, is hardly reassuring. With hardly any regulations to require better behavior, this is more than a little bit scary.
Future conversations at Newtowne
What do I plan on saying to my students at Newtowne about chatbots? I would begin all future conversations by finding out, through dialogue and play, about children’s conceptions of chatbots. The answers here are provisional, informed by my conversations with last year’s Green Dragonflies.
Explaining chatbots to children
I completely agree with Gemini that we need to be clear that bots are not people, are not alive, and do not have feelings. We are in the midst of a mental health crisis. Large numbers of teens and adults report having few or no friends. Many spend hours scrolling on social media. Attachments to machine “friends” is exasperating this sense of isolation. Based on my conversations with them it is clear that the Green Dragonflies are ripe for conversations about chatbot personhood. I would tell them that:
A chatbot is a machine that reads a lot of books. All the books in the library and more. It uses this information to answer questions. It is important to remember that Chat GPT [or Alexa and Siri] is not a person. It isn’t even alive. It is a machine.
Tech companies are, of course, aware of the research that confirms that when a chatbot mimics a human it is more likely to be interpreted by young children as alive. Given this deception, it is important to add:
The owners of some of the machines are trying to trick us into thinking their machines are alive. It is important not to get fooled by the voices they give their machines.
This advice is part of a larger message children need to get about misleading advertising and other efforts to get them to buy products or believe ideas not in their interest.
Explaining where AI lives and “The Cloud”
I agree with Gemini about the value of making explanations to young children concrete. I would show children a photo of a data center and tell them:
AI machines are in this kind of building called a data center. Our phones and laptops send signals through the air to the routers in our school [I would point these out] which send signals to a data center. Then the machine then sends its message back to us.
But I would never say AI lives “in a magical toy box in the sky.” Gemini and other bots are not magic. This is like telling children that the light that appears at the flip of a switch is magic. Or the water that comes by turning on the tap is magic. We possess powerful technology, and are drawing on the bounty of the earth to power this technology. Calling it magic ignores the fact that chatbots are the result of centuries of scientific innovation and rely on a vast, communal infrastructure. It creates the illusion that using Gemini has no consequences, ignoring the fact powering data centers results in vast amounts of carbon emissions which contribute to climate change.
While the origins of the cloud metaphor to explain networked computing may have been benign, the use of the word with young children is confusing, and distracts from the issue of data centers impact on the environment. Let’s lose the term cloud when talking to young children about AI.
Explaining to children how AI can help them
Without a doubt AI is a powerful technology with tremendous potential. Children should know that these machines are helping doctors and scientists in curing diseases. I would explain:
The chatbot can very quickly look through all the books it has read to answer our questions. The machines are another way (along with reading books ourselves and talking to experts) for you and for grown ups to find out information.
Beyond this, the benefits of chatbots are, at the moment, hypothetical. Maybe chatbots will help adults finish work more quickly so they can come and play. Or maybe there will just be more emails to write. It will take time for early childhood educators to sort out if and how AI can do more than cute parlor tricks (e.g., making up stories about blue bananas), and support young children’s learning in important ways.
The use of AI in education is a work in progress, and children can be part of the exploration of how chatbots might support their learning. In service of this exploration I want children to approach bots with a playful mindset (open to possibilities and with a feeling of agency). I would tell them:
You and your friends and family and teachers can play around with a chatbot to figure out ways it can help you learn. You can explore when you should use a bot and when you shouldn’t. To make sure this play is safe, your grown ups will help you explore this new tool.
What this exploration might involve will take time to sort out and will be the subject of another essay.
Explaining the problems AI causes
Gemini makes a good point that mistakes, fairness, and laziness are dangers of AI that children should know about. Regarding mistakes, I would explain:
The answers AI provides are not always right. Like in all of our conversations, there are different ideas we can consider. While AI can give us new ideas to think about, we have to check to see if the ideas are good.
But Gemini is selling young children' s abilities short when it advises we need to keep things "incredibly concrete.” Four and five-year-olds can engage in conversations about complex topics. What those topics should involve depends on what issues are relevant in a particular community. For example, if there is a debate about building a data center in the area, I would share a child-friendly explanation about both sides of the issue. Since the data center will have an impact on their future, they have a right to know about the situation. I am confident that children are able to come to their own opinions on the subject.
Be a Pain
I keep coming back to Phillips' concern that, “Maybe bad questions could happen.” My biggest concern about AI is that this powerful technology will be misused to serve the interests of the ultrarich who own the AI companies; that the question they are asking is “how can we make even more money from this product” rather than “how can this technology help?”
That a very few control AI and gain the monetary benefits from the technology is particularly unfair since AI is the product of the knowledge produced by all of humanity, all those books Gemini has read. AI should be a common good. Children can understand that:
AI works by reading all the books in the library and those books that were written by different people. So it is not fair that Chat GBT (or Claude) is owned not by those writers or by everyone, but by a few people.
Then what? At Newtowne I have the benefit of being able to link my explanation with the anti-bias work of colleagues. Our music teacher Alastair Moock teaches the children a song he wrote called Be a Pain. The song names a variety of heroes–Rosa Parks, Harvey Milk, and Billy Jean King–who stood up to injustice and made things more fair. These changes, for example King getting more equitable pay for women tennis players, did not occur by being polite. The changes happened because people forcefully advocated for change; they were a pain. I would tell the children:
Like Rosa Parks and Harvey Milk, there are people who are being a pain to make things fairer, so that AI can help everyone.
When I say this I have in mind Bernie Sanders’ AI Sovereign Wealth Fund proposal, which would give the American people half the profits of AI companies and require those companies to work for the public interest. Pretty cool idea.
Three final thoughts. Efforts to regulate AI companies face the challenge of their enormous political war chests. They can spend billions to influence elections and shape policy. If anything like the Sanders proposal has any chance of becoming law we all have to be a pain.
I can feel overwhelmed and disempowered when thinking in terms of such public policy. Despite the fact that I live in what, for the moment, is meant to be a democracy, objectively I little influence here. I take consolation in places where I do have some control. Gemini (and its owners) do not get to decide how I explain chatbots to our children.
Actually, we get to decide. The adults at Newtowne–teachers and families–need to develop a common language to explain the chatbots. Together we can help our children develop a clear understanding of what AI is and what it is not, a playful mindset for their interactions with AI, and a healthy skepticism of AI companies. How we talk about the bots matters.
None of this essay was written by AI. Why? First, I enjoy writing and have no intention to outsource the fun of crafting ideas and stories to a machine. Also, I agree with Gemini that over-reliance on AI can “make our brains lazy.” As the chatbot explains:
If we let the computer do all the drawing, all the thinking, and all the storytelling for us, our own real brains won't get any exercise! Just like your muscles get strong when you run outside, your brain gets strong when you think of your own ideas. If we use the computer too much, we forget how to be creative ourselves.
At the same time, my ideas were strengthened by conversations with a number of human thought partners: Junlei Li, Amos Blanton, Robbie Berg, Mitch Resnick, Eric Schilling, Zoe Bentley, Leo Burd, Ann Valente, Rupal Jain, Ricarose Roque, and other friends and colleagues at LifeLong Kindergarten.
My thanks to Liz Merrill and Amos Blanton for looking over drafts of this essay.