In this lesson, students will redesign the YouTube recommendation algorithm to meet their needs and reduce bias. This is a culminating project that can span several days of work and spark student reflection on lessons learned from the curriculum.
This lesson introduces students to how supervised machine learning can be trained to classify complex datasets based on labeled data. Students will train their teachable machine models and learn that AI can learn from labeled data. They will also revisit the idea that training AI systems with an increasing amount of data does not necessarily mitigate bias if there’s not enough diversity in the data.
This lesson further shows students that different algorithms can have different purposes for different stakeholders and that such relationships can be visually represented using an ethical matrix. Students will create their own ethical matrices for their best PB&J sandwich algorithms.
