Machine Learning with Younger Students: 1st Graders as teachers of AI!

Artificial intelligence is such a huge topic right now. It is having an impact on so much of our world, both good and bad. Whether, why, and how it could or should be integrated into education is part of that discussion. Working at a progressive education institution, a tenet of our mission is to educate students for the world in which they live and to be active participants in their education. I have been trying to find ways to integrate AI with my younger students in ways that are developmentally appropriate and in ways that they can actively engage with the tool. With my first graders, developing a connection with a larger project seemed like a perfect way to build their knowledge of how machine learning and AI work.

First graders are doing a larger project weaving in research, design, iteration, and building and making. They learned about the plastic trash that ends up in waterways and oceans. Students read books and researched how trash ends up in the ocean and the large collections that have formed, including the Great Pacific Garbage Patch. First graders discussed how this plastic trash is harmful to the ocean ecosystem. Ocean animals are harmed by the plastic that is damaging their homes, and sometimes the ocean animals ingest the trash or get caught up in it,resulting in the animals getting sick or even dying.

Students then learned about some great organizations that are working to clean up the ocean, including the Ocean Cleanup organization. This group is using several different types of technology to help clean up the ocean,including AI. The Ocean Cleanup developed AI-enabled cameras to identify and track trash in the ocean. The cameras are mounted on ships traveling all over the world and are equipped with AI technology to determine whether objects are plastic or trash and not ocean animals or sea life, take a picture of the object, and mark the location where it was spotted. All of that data is collected and submitted to the scientists working for the organization to better track trash in the ocean and send ships to the right locations for cleanup. The data also helps to track the progress of reducing the plastic and trash in the ocean and share the data with governments and other groups to advocate for laws that will help reduce the plastic in our water systems.

Next, students talked about the AI project we did in kindergarten. We discussed the tools we used and how the AI tools helped us to make a mascot for each classroom and write a collaborative story including all of the ideas of our classmates. Then we discussed how an AI machine learns and what machine learning means. Students made connections with the coding they were doing in Scratch Jr, giving instructions using coding language,and how that is different from machine learning, responding to prompts and questions, interacting with a user, and learning from those interactions.

Then, using the site Machine Learning for Kids, students explored a simple way to train and test a machine. The first step was showing students how we trained a machine to determine if a picture was of a human or an animal. We showed students how we uploaded images to the train section. We talked about how we picked pictures of all different kinds of animals so the machine could learn to recognize all different animals—animals with fur or without, small and large animals. We also showed all the different images of humans. We talked about how humans all look different—older and younger, different skin tones and hair types. We wanted the machine to learn to see all different types of animals and humans. Then we tested the machine. Each student had a different picture of a human or an animal to see if it could correctly identify what the picture showed. We tracked how many times the machine was correct or incorrect. The more we tested the machine, the more accurate it became, and we talked about how the machine was learning more and more. With each class that tested, the machine got better and better at identifying the pictures correctly. We also talked about how we only trained the machine on a small number of images and that the more images we provided to the machine, the better the machine would be at learning and correctly identifying the images.

Now it was time for the students to be the teachers of the machine! Making connections with what students learned about the AI cameras used by the Ocean Cleanup project, we challenged the students to train an AI machine to correctly identify images of ocean animals and trash. The first step was gathering images to teach the machine. We talked about how the better and more accurate images we provided to the machine, the better the learning model would be. If we trained the machine on inaccurate images or poor information, the machine would not learn what it needed to learn. Students went around the school building to take images of trash they found in different places and spaces, especially after lunchtime! They focused on empty water and drink bottles, takeaway containers, and empty chip and snack wrappers. Then we sorted through the images to find the best quality images to train the machine. The students uploaded over 500 images of trash to train the machine. We also found about 150 images of ocean animals. Then students put their training model to the test. Each student showed the machine a picture of an ocean animal and one of trash to see if the machine had learned enough to identify the image correctly. The machine was able to correctly identify the trash every time but was incorrect about the ocean animal about ten times.

After reflecting on teaching the machine, students determined that if they uploaded more images of ocean animals—maybe the same as the trash, about 500—the machine would have been better able to identify the ocean animals. They also made connections with how important it is to train a machine with accurate, good information. Just like when a teacher makes sure to give students good, accurate information, that is also important when teaching a machine. If the machine gets the wrong information or poor quality information, the machine is not that smart and produces inaccurate information. The hope is that students will build on their knowledge of how AI machines work and learn and think about this as they get older and start to use AI more independently. Students should understand and think about how to use an AI source, look into and think about how the machine works and was trained, and be critical about the information that the AI responds to prompts with.

The other aspect of this project was for students to think about how artificial intelligence can be used to help. As students are moving on in their school and learning journey, we want them to develop a mindset about how technology can be a tool to help. Students should think critically about how they are using technology in their lives and how that technology is impacting the world around them. Our goal is to help them develop a mindset of seeing technology as one tool in their lives and how to use it as a tool that can help our world.

This project was a good hands on experience for students on how machine learning is trained. It also was a way to have engaging conversations with first graders about artificial intelligence and how it can be a positive tool. The project also helped students to develop an understanding that AI is only as good as humans train it to be and this will be an important pieces of information for students to continue to think about as they grow and develop a deeper understanding about AI.