Week 2

DoodleIt: A Novel Tool and Approach for Teaching How CNNs Perform Image Recognition.

This research paper introduced middle school students to image recognition using a previously trained convolutional neural network (CNN). I liked that this project was set up to allow students to peel back the mystic of how CNN’s recognize images by assigning weights to the 6 possible trained drawings.

It was surprising to me that of the 4 students they monitored for this project, two of them had a frustrating time using the program. The researchers noted that in the paper that it could possibly be because students didn’t know to fill the canvas completely with a picture, but I do wonder if there was a disconnect when designing the program and then implementing it to the students.

This is something that I want to keep in mind when we’re designing our own program, we need to be mindful of how the program is going to be used when interacting with actual students. If the experience ends up being too frustrating for the students, then it could end up being detrimental to their AI literacy development.

A big surprising thing to me that the researchers have pointed out, is how their seems to be a general lack of tools in the education space for teaching AI concepts. This paper, and the others I have read, make it seem like there just aren’t tools/workits/lectures for students to learn. This is pretty promising for us I think, because it shouldn’t be too difficult to come up with novel ways to teach students.

Fostering Secondary School Students’ AI Literacy through Making AI-Driven Recycling Bins

This research article employed a mixed-method approach towards teaching AI Literacy to middle school students. The researchers made an effort to focus on how they could use Maker education to enhance students’ learning efforts, as they noted that current approaches towards teaching AI literacy to students revolves around utilizing simulations and games. While these methods have been helpful in improving students AI knowledge, they may not adequately prepare them for higher-level cognitive tasks.

The students engaged in learning activities to understand what AI is, experience how AI works, and its societal impacts and ethical concerns through 12 40-minute lessons over two months. Over the 12 lessons, they were introduced to the topic of “making an AI recycling bin” as the course objective. Before and after the course, students were given questionnaires to gauge their AI literacy, motivation, and collaboration.

This research was particularly interesting because it focused on a different methodology than some of the other papers, I’ve read on AI literacy. I think it’s a smart idea to focus on project-based learning (PBL) for AI literacy, as it gives a real way for students to engage with and understand the material being taught to them. The only downsides I could see of replicating a similar experiment would be time and resource constraints, as not all schools would have access to the makerspaces required to facilitate a proper PBL lesson.

A is for artificial intelligence: the impact of artificial intelligence activities on young children’s perceptions of robots.

The researchers in Williams et al. developed assessments for each of the three AI activities they conducted. They used their Popbots to teach children about three AI concepts (knowledge-based systems, supervised machine learning, and generative AI) with child-friendly activities.

The assessments for each activity were a set of multiple-choice questions designed to gauge if the student had developed an understanding of the AI concept. They based the questions off of research conducted by Wellman and Liu’s Theory of Mind Assessments (which would be worth looking into when we start to think about designing our own assessments), where they picked 3 tasks from the research to assess the students AI confidences: Knowledge Access, Content False Belief, and Explicit Fals Belief.

Children also completed a questionnaire about their perceptions of robots and AI on either a tablet or paper (maybe we should also offer the choice of having either/or option in our research?). The questionnaire was formatted based on a previous study where two on-screen characters offered differing opinions about robots. The child then chose which opinion they agreed with more, or if their opinion fell somewhere between the two.

The researchers’ procedure was conducted with the children completing the Perception of Robots Questionnaire and then the Theory of Mind Assessment. After interacting with the Popbot activities, the children were then given the AI assessment again to measure their understanding of the topics, followed by the Perception of Robots assessment again. In addition to the quantitative data that was collected from the research, the researchers also recorded observations while the children completed the activities.

Three AI Concepts

ML Classification of Legos: We could show students how to classify and sort different Lego bricks based on shape and color. Teaching a computer using ML should be doable, and it can teach students the 1st and 3rd big ideas (Computers perceive the world using sensors and computers can learn from data). I’m not sure how difficult it would be to do this in one session, but it could be fun for the students to interact and build the classifying model themselves.

Creating AI Art in the Classroom: We could have students come up with prompts for a generative AI and have them think about what the computer might make based on their input. I would think that this project would be easy to implement as we could use existing AI art generative services that exist (DALL-E, etc.). This project would teach students about Big ideas 2, 3, and 5, particularly idea 5 would be important as we could have a lesson on copyright ethics after the lesson.

Building Small Autonomous Cars: Have the students develop small autonomous cars to go around a predetermined track. This would cover Big ideas 1, 2, 3, and possibly 5. We could teach about the ethics of how autonomous systems affect our real world after the lesson, and get feedback from the students on their preconceptions of how autonomous systems operate.

The big goal behind the 3 of these projects is to have the students actually engage and participate in creating the AI system. I think this maker based education approach is a good direction to take this project as it has the potential for the students to relate to the material better.