OVERVIEW

Activity Overview:

In this lesson, students will learn what supervised machine learning is and how it operates. Students will then work with version one and two of Google’s Teachable machine to create classification models. They will analyze how biased data changes the output of the algorithm and look at ways in which to avoid algorithmic bias. Students will begin to consider how bias in data is one example of an ethical issue in AI.

Meta description

• Subject Area: Computer Science, Technology, Engineering, Advisement or general enrichment classes
• Grade Level : 6-8, 9-12
• Computer Science Domains:
  • Algorithms and Programming
  • Impacts of Computing
• Computer Science Principles:
  • Fostering an Inclusive Computing Culture
  • Developing and Using Abstractions
  • Creating Computational Artifacts
• Materials:
  • Website
• Considerations:
  • Educators will need to understand the basic concepts of AI and machine learning. They should practice building the cat/dog classification model before doing it with the class so that the demonstration goes smoothly and so that they can answer any student questions once the students start working on their own models (either the musical instrument model or building their own model with their own data sets).

Lesson Plan

Overview

In this lesson, students will learn what supervised machine learning is and how it operates. Students will then work with version one and two of Google’s Teachable machine to create classification models. They will analyze how biased data changes the output of the algorithm and look at ways in which to avoid algorithmic bias. Students will begin to consider how bias in data is one example of an ethical issue in AI.

ASSESSMENT PRE/POST-TEST

Was your classification model successful? What did you learn about the type of data your model needed to perform successfully? Was your model biased, if so how?

OBJECTIVES

Define and explain supervised machine learning. Know what a classification problem is and how AI models work with it. Understand potential issues with classification in the supervised machine learning context. Understand how the quantity of training data affects the accuracy and robustness of a supervised machine learning model.

CATCH/HOOK

To review the artificial intelligence data/learning/prediction sequence, students will participate in an interactive AI Bingo activity where they will ask various classmates to identify types of AI tools they have encountered, predict the data input and the outcome (prediction) for each type of AI. Once a student gets five AI tools in a row, they will have a bingo and the class will move into the lesson.

ACTIVITY INSTRUCTIONS

1. After the students have participated in AI Bingo, review the concepts of machine learning, data set, and prediction after the activity. Tell the students that today you are going to learn more about supervised machine learning and go over the Introduction to Supervised Machine Learning slide deck through slide 23. Be sure to use version 1 of the Teachable Machine for the student activity (https://teachablemachine.withgoogle.com/v1). Debrief with students utilizing the questions from MIT AI Ethics Education Curriculum (linked above).
2. Let students know that in order to experience how a machine learns to classify, they are going to work with a partner to build a model that will classify dogs and cats. Work through the Dog/Cat classifier section of the MIT AI Ethics Education Curriculum, making sure to ask the questions dealing with AI bias in data.
3. Depending on the ability level of the students and the time available, do either the Musical Instrument Classification activity from the MIT curriculum or the Teachable Machine Project from the DAILy Workshop.

Supplements

Any items in this section are the property & under the license of their respective owners.

REVIEW

Review the definition of a classification problem, how a machine learning model can solve it, and touch on bias in data sets.

STANDARDS