In the previous video, you learned about the importance of using fair data when training a computer program.
You wrote a paragraph describing your discoveries and added more diverse and inclusive images representing your chosen topic.
Now, you will consider the benefits and drawbacks of colleges using machine learning models for marketing.
Marketing is the way a business or organization convinces people to buy their product or service.
For example, an online store uses marketing to create and purchase the ads you see when you use the internet, or a business in your town could write a blog post to create awareness in the community and get more people to shop there.
Many colleges market to potential students by sending them promotional materials, such as brochures about the school.
Some of these colleges use models to help them decide which students will receive their promotional materials.
Some might even use models to help them decide which students to admit.
Colleges can receive thousands of applications every year.
Using a model to evaluate students is faster and more efficient than reading through all of the applications..
But is it fair to the student -- or the school?
Based on what you have learned so far, do you see potential for bias?
High school grades and scores on standardized tests are often used as data points in college admissions.
High grades and test scores “can” help predict whether a student will be successful in college But imagine if a school developed a machine learning model that made marketing and admissions decisions and the “only” data it used was grades and test scores.
In your machine learning journal, return to the “Fairness in Machine Learning” section.
Think back to what you discovered in the previous videos.
What do you think of schools “only” using SAT scores and GPAs to determine who receives marketing materials?
Who might miss out on the opportunity to go to that college?
What valuable student characteristics or experiences would the college be omitting, excluding or overlooking?
For example, what about a student who does many hours of volunteer work?
Or a student who is dedicated to an extracurricular activity?
Or one who has family commitments, such as caring for a younger sibling?
Are those responsibilities less important than getting good grades, or do they simply show a different set of experiences?
As you reflect on the biases that a limited model would create, consider the following questions, and answer at least “two” of them in your journal: What aspects of college readiness does a model like this not account for?
What experiences could be excluded, omitted, or overlooked?
Why is it so important to use diverse data when training a machine?
And is a computer capable of making fair college marketing decisions, especially with a limited model like the one you’ve considered here?
Why or why not?
You do not need to type the questions in your journal.
They are just on the screen for you to reference as you write your responses.
Now, it’s your turn: In your machine learning journal, think about the additional Fairness in Machine Learning Reflection questions.
And record your answers to at least two questions in your machine learning journal.
- In your machine learning journal, think about the additional Fairness in Machine Learning Reflection questions.
- Record your answers to at least two questions in your machine learning journal.
This project will be shared with your teachers
Students can submit their work on this page. View their submitted work on the student progress page of My Classes.