Machine Learning

At the start of the second trimester, the sixth graders were introduced to machine learning by completing the AI for Oceans lesson in Code.org.  Students learned that machine learning is a type of artificial intelligence. They discovered how training data is used to enable a machine learning model to classify new data. In this lesson students provide training data to classify objects as "sea creature" or "not sea creature" to help remove trash from the ocean. The lesson also helps make students aware of biases in training data that may cause inaccurate results.

After completing the Code.org lesson, the sixth graders were asked to define in their own words what machine learning is. Here are some of their responses:
Faustina: "Machine learning is giving lots of data to a computer, that then is collected and analyzed by the computer that helps it make predictions."
Marco: "It is a machine that does not need coding. It learns from you teaching it new things over and over."
Isabella: "Machine learning is a type of artificial intelligence that makes its own decisions without programming. Machine learning needs to be trained in order to identify certain things like, images, music, files, etc. Machine learning or AI is like a human being being trained, if they make mistakes they are to be trained again and learn from the mistakes. Machine learning is the present-world AI and will continue to grow as time goes on."
Ryann: "Machine learning is the things we teach to machines. That's how artificial intelligence is formed into them. We feed information into these devices teaching them what they need to know instead of having to program everything into it."
Jeremy: "Machine learning is when you teach the AI by giving it examples on what a certain thing should look like. In the coding assignment we did we gave the AI examples of fishes and because we gave it more and more data it understands what a fish looks like. It is when you give the AI more examples and give it plenty of data without bias so it would function correctly."
Keanne: "I learned that AI needs to learn by collecting vast amounts of data. I also learned that if given wrong or bias information, the AI could make a mistake."

After the introductory lesson in Code.org, the students started getting a hands-on experience for training machine learning systems by using the website Machine Learning for Kids. The first project they did was Make Me Happy. In this project students taught a computer to recognize compliments and insults. They created a Scratch program where a character smiles if you say something nice and frowns or cries when you say something mean to it. 

First step was to train. In the image below, students' training data included recognizing text as "kind" or "mean" things to say. Students were able to test what the computer has learned. If they typed a nice message, the computer should classify it as "kind." If they type a mean message, the computer should classify it as "mean." If they were not happy with how the computer recognizes the messages, they go back and continue training by providing more text examples of "kind" and "mean" things.

When students think they have trained the computer properly, they started working on the Scratch Project. Watch the video below to see a sample student work.