Skip to content

5. Fairness in Machine Learning Wrap-Up

Playback Speed:
Transcript

In this lesson, you chose a type of item from a list ...

And collected images that represent that item.

You reflected on the images you chose, and identified what had been excluded.

Then, you applied your discoveries about machine learning bias to the real-world issue of using models to make college marketing decisions.

In the real world, college admissions officers review students’ grades and test scores, but also extracurricular activities, volunteer work, jobs, and hobbies.

They read personal essays, letters of recommendation, and samples of schoolwork before making decisions, and they do not rely “only” on machines for processing applications.

Machine learning and human effort combine with the goal of selecting a diverse and well-rounded student body.

Now, think about other ways that machine learning could potentially be unfair.

For example, should a bank rely on models to determine who gets a loan?

Should doctors rely on computers to make diagnoses?

What are the risks and benefits of using machine learning in these scenarios?

And why is training a computer with diverse data so crucial?

Machines train with large amounts of data.

They find patterns in the data -- many of which may not be obvious to a human.

Then, they make a decision based on the data.

What is gained from having diverse data and perspectives in machine learning?

What is lost when perspectives are omitted, excluded, or overlooked?

Keep thinking about the benefits and drawbacks of machine learning as you encounter it in the world around you, and have fun discussing these topics with your class!

Next