It may sound crazy, but it was a dog named Zelda who cracked the code on teaching AI to five-year-olds.
It all began when Samantha Westerlind, an elementary technology teacher in Cherokee County, Georgia, was contemplating how to best help her young students learn what AI was and how it learns. Concurrently, she and her daughter were also trying to teach their mutt, Zelda, how to sit.
“We just kept giving her the same command, over and over again—sit, sit, sit,” said Westerlind. “And suddenly I realized that training a dog is exactly the same as training an AI, because you give an AI model consistent data and consistent information. It’s just like asking a dog to sit, sit, sit. Eventually it will understand and learn it.”
Good Dog, Good Data: How a Therapy Dog Became an AI Teacher
The experience gave Westerlind an idea. There were therapy dogs at her school, and the counselor who brought them in had mentioned how difficult it was to work with the kindergarten students because they often forgot to practice gentle touch. Could Westerlind use AI to solve this problem?
Westerlind began working with her fourth graders, using micro:bit circuit boards (small programmable devices with an accelerometer that can detect movement along XYZ axes) to build an AI model that could distinguish between good petting (slow, gentle) and bad petting (fast, rough). The students collected movement data, trained the model themselves (making sure they entered enough data to have a 92% accuracy rate), and coded the circuit boards to display a happy or unhappy face depending on the petting style detected.
The kindergarteners and therapy dogs were brought in, and the fourth graders became the teachers and explained the process. The kindergartners wore the micro:bits on their wrists, like watches, and witnessed the happy/unhappy faces in real time. They then adjusted their behavior completely (slowing down if they saw a sad face), with no adult intervention needed.
Westerlind was stunned by the way the kindergarteners self-regulated because the AI gave them real-time, objective feedback. “This session allowed both grade levels to master complex concepts,” she said. “The 4th graders learned about data sets and model accuracy, while the kindergarteners learned that AI ‘knows’ things because we provide it with information over and over again. It moved the technology out of the abstract and into a hands-on experience that improved both digital literacy and student empathy.”
One Teacher. 1,300 Students. A Completely New Approach to AI Education.
Westerlind is doing something rare: teaching elementary students not just to use AI, but to truly understand it. Most recently, she served 1,300 K-5 students, developing cross-curricular programs in coding, robotics, 3D design, virtual reality, and AI. This year she was recognized with a Discovery Education Award for her groundbreaking work bringing ethical, hands-on AI literacy to early elementary students.
“I’ve been an educator for 16 years, and I love what I do,” she said. “I love bringing the world into a classroom in many different forms.”
At right: Samantha Westerlind is celebrated as a 2026 Discovery Education Award winner.
Here are Westerlind’s 8 Insights about AI at the Elementary Level
1. Kids are already using AI, without understanding it.
Westerlind noted that students already understood digital citizenship and knew not to share personal information with strangers online. But the conversational, friendly nature of AI chatbots bypassed all existing knowledge, because it didn’t feel like they were talking to a stranger. It felt like talking to a friend.
“Because AI is so cool, they let their guard down and started giving these AI bots their private information,” she said. Westerlind urges more education, since digital citizenship education hasn’t caught up with AI yet, and young children are the most vulnerable to that gap.
2. Elementary students are more capable than we assume.
A recurring theme is that educators and parents underestimate young children. AI can be overwhelming and confusing to adults, but Westerlind feels that age-appropriate concepts, grounded in the real, physical world, are essential for elementary learners.
3. AI education is being gatekept for older grades. That’s a mistake.
Most district policies and curriculum frameworks start AI education at 6th grade, but Westerlind believes it needs to start much earlier. “It’s so important for the younger grade levels, kindergarten through 5th grade, to have that foundation and understanding,” she said.
4. AI literacy must be built on empathy and ethics, not just functionality.
One of Westerlind’s most compelling arguments is that AI education should start with empathy. This means understanding consequences and caring about outcomes, which leads to more responsible use. The reason the dog-petting project worked well was because kids had a real stake in getting it right and not hurting the animal.
5. There’s a “gatekeeping” problem: teachers say “don’t use it,” while using it themselves.
There’s a contradiction at the heart of current AI policy: schools restrict student use while quietly adopting teacher-facing AI tools. Students notice the double standard, and it breeds either resentment or workarounds.
6. Understanding AI dispels fear.
Westerlind understands that the rapid emergence of AI is frightening to parents. She also feels passionately that understanding AI is the key to dispelling fears, which is why she’s so committed to helping young students better use it responsibly. A great first step in making this happen? Having schools host parent series, so they have firsthand awareness.
7. AI education shouldn’t start with a chatbot prompt. It should start with a question.
According to Westerlind, the dominant model in education right now is tool-first: here is a chatbot, here is how you prompt it, here is what it can do for you. “A student can type in, ‘Make me a unicorn riding a surfboard,’ and it appears, like magic,” she said. But Westerlind insists that curiosity should come first. “It shouldn’t be ‘I inputted a prompt and look, I got something from it.’ It should be, especially for the younger grades, the foundation and understanding knowledge that AI can be so much more.”
