Kids don't need chatbots. Teach them this about AI instead.

The warp-speed deployment of artificial intelligence into nearly every corner of American life has created real problems for K-12 education.

Suddenly, school boards, districts, and teachers are grappling with the role AI should play in students' lives. The consternation is already showing up in classrooms as teachers use AI to plan their lessons and analyze student performance, provide tutoring and emotional support, and offer access to AI-enabled search.

Meanwhile, companies like OpenAI and Anthropic compete to provide chatbots to students. AI supporters worry that students blocked from using chatbots will be left behind, while many alarmed parents support AI bans for a variety of understandable reasons.

Dr. Shuchi Grover, a computer and learning scientist, believes that education is caught in an AI hype cycle.

Her solution? Help students understand chatbots through safe, sandboxed models and interactive experiences, but don't make those interactions, especially with popular mainstream models like ChatGPT, Gemini, and Claude, the dominant way they learn about AI.

Instead, Grover suggests offering computer science instruction so students build a foundational understanding of computing and develop enduring problem-solving skills. The point is for students to view AI as a powerful new technology with many applications.

By contrast, some schools take the approach of, "'Oh let's make kids good prompters,' which really makes me cringe," Grover says.

Why not give students chatbots to learn about AI?

Grover, who researches K-12 computer science education, acknowledges that AI is indeed everywhere but calls the focus on chatbots "misguided."

After all, students encounter AI in their algorithms, self-driving cars, voice assistants, and facial recognition technology, among myriad other commercial products. That's why restricting students' learning about AI to chatbots only makes little sense.

There are also privacy, safety, and bias concerns, plus the risk of anthropomorphizing AI, Grover explains.

Companies may use student conversations with chatbots as training material. Students paired with a chatbot may unintentionally learn, courtesy of their own teacher, how to offload their cognition and creativity to AI. There is no clear evidence yet that providing chatbot access to students improves their learning or academic performance.

Grover also recognizes that AI companies are pushing student adoption of AI for their own reasons: The industry sees an opportunity to make lifelong users out of children and teens, she says.

While Grover generally opposes commercial chatbots in the classroom, she doesn't support an outright ban on them.

How students should learn about AI at school

There is, in fact, guidance for teaching students about AI. This summer, the Computer Science Teachers Association released new standards for AI use in pre-kindergarten through 12th grade.

While CSTA's standards encompass many concepts, like cybersecurity and physical computing, they integrate AI learning throughout.

Computer science standards and AI

In elementary school, CSTA believes students can start learning about AI through foundational concepts like explaining how a change in an algorithm leads to a different outcome, examining how data is used to train a machine learning model, and training a model to make a classification or prediction.

By middle school, students can analyze AI-written code for accuracy and usability on a specific project, and use an AI tool to generate outputs that help them solve a computational problem. Importantly, student use of an AI chatbot is useful, though not necessary, to complete the latter exercise. Plus, sandboxed chatbots that don't relentlessly engage with students can serve the same purpose.

A set of specialty standards illustrate how high schoolers can engage with AI in ways that truly prepare them for the future. For example, they can examine how data flows through a neural network, a machine learning model that will be critical for the next generations of AI. Or they can integrate a prebuilt AI agent into an app and optimize an AI algorithm's performance for a given problem.

There's more to AI beyond expert prompting

These skills imagine a world far more exciting for students than just being an expert AI prompter.

Instead, Jake Baskin, a former high school computer science teacher and CSTA's executive director, says we should be setting students up to succeed in a society where computation is deeply embedded in everyday life.

"I think we will do a grave disservice to students if we don't prepare them with the foundational understanding of how this technology works and how to create it, so that they'll have agency in deciding when to use or not use it in their lives," Baskin says.

The challenges of teaching students AI

Of course, there are numerous barriers to implementing computer science curriculum in schools.

Not every classroom can meet basic requirements for computer science instruction. While the CSTA standards offer "unplugged" versions of skill-building exercises, computer science classes ultimately need access to working laptops and reliable internet.

Even if students can access the necessary technology, Grover says it's imperative to give them curriculum relevant to their lives and life experiences. The AI4K12 Initiative, a resource hub jointly offered by CSTA and the Association for the Advancement of Artificial Intelligence, has many free curriculum options, but someone needs to evaluate and decide what's right for the student body in question.

Baskin says states often adopt varying standards for computer science, which are then implemented by individual school leaders. AI instruction at one school may be substantially different from curriculum at another school in a different neighborhood, depending on resources and interest.

Teacher preparation and professional training are also critical. In many instances, a passionate teacher steps up to teach computer science, even if it means learning the skills for the first time themselves.

Education doesn't have to be "caught in the middle"

Grover warns schools against giving in to the temptation of chatbots for students. Instead, she urges decision-makers to pursue computer science education with a focus on AI.

"It's market forces that is pushing these tools into classrooms," she says of commercial chatbots, noting the intense competition between companies to grow their share of users. "Education, for some reason, got caught in the middle of this. It doesn't have to be."