Machine Learning for Social Good
The aim of this lesson is to help students build interdisciplinary connections and apply machine learning to new subjects. In the main activity, students work in groups to prepare and present ideas for machine learning models that will help benefit the social good. This lesson will build upon Lesson 2 and encourage students to invent their own ideas for machine learning models. Students do not actually build the model, just present the idea.
Learn about real examples of machine learning for social good
Prepare and present ideas of machine learning models to help people's lives
Understand to think about input data, model and predection independently
Interactive whiteboard (something to watch class video)
Pen/pencil & paper
Spare paper for students to brainstorm ideas
Print out "Invent a model" template (1-per group)
Lesson 3 Slides
Slides: Introduction (5 minutes)
* If you have not already, access the Lesson 3 slides using the button at the top of the page or in the lesson materials.
LESSON 2 RECAP
In the introduction, ask students to recap on the "Spot the Shark" lesson two learnings. Recall that we used machine learning image classification model to predict whether an image was a shark or a dolphin. Relate to the real world and mention that there are "real life" examples where machine learning can be used for social good to help combat the extinction of animals in the wild.
Video: Wildlife Insights (5-7 minutes)
Watch "Wildlife Insights" video to show students real-life examples of using machine learning to help protected against the extinction of wild animals. Wildlife insights is a project run in partnership with Conservation International and Google that uses machine learning to classify camera trap images in the wild. With the machine learning method, researchers can now know what animals are at risk of becoming extinct, where are animals moving to, are there new species of animals that are identified, etc?
Wildlife Insights - Saving Biodiversity with Tech and AI
Discussion: Wildlife Insights Findings (5 minutes)
After the video is completed ask students in the class to identify the three parts of the Wildlife Insights machine learning model. The answers should be something similar to:
Input data: Camera trap images of animals
Model: To help monitor animals in the wild to help combat extinction of biodiversity
Prediction: (1) Detect if an animal is in the photograph, (2) Detect what species of animal is in the photograph
Main Activity: Ideas for Social Good (20 minutes)
Group Activity: Inventing a Machine Learning Idea for Social Good
For the main activity, split students into groups of 4-5 students. Each group will be given a machine-learning-template.pdf sheet which contains headings for input data, model and prediction. The template for the sheet can be found in the Lesson 3 materials. Students should invent their own ideas for a machine learning model that solves a problem - enviornment, saving animals, helping find your friend a better birthday present, discover new planets in the sky. The aim of the main activity is to get students to be creative as possible and to get students to think about the types of training data they would need to build their machine learning model. Example of the template with instructions are as follows:
If groups are struggling with idea, here are some "real life" suggestions for applying machine learning different discplines that could be interesting:
Healthcare: Using computer vision to avoid touching your face during Covid-19 -- real life example here LINK
Physics: Use computer vision to identify new stars and planets in the sky -- did you know AI helped take the first-ever picture of a black hole? LINK
Final Activity: Presenting Ideas (10 minutes)
After each group has completed their machine learning for social good template, the final 10 minutes of the lesson are dedicated for groups to present their ideas to the wider class. Each group member should try and contribute to the presentation - one person talk about input data, another about the model, another about prediction, etc.
If time permits, feel free to get creative with the presentations - perhaps get students to create a 1-slide presentation using Google Sheets or Powerpoint. The emphasis is to get students thinking about the three different stages of a model and how to present these ideas to the wider class.