Making Up People: Categories, Classification, and Clustering

Topic / Focus

How categories of people get made – by societies, by institutions, and by algorithms. We start with identity and social construction (Thorn, Hacking), then look at how machine learning uses existing categories, reinforces them, and rebuilds them. We introduce clustering: a kind of unsupervised learning that groups people by similarity without being given labels.

Guiding Questions

• If a property is real (like height), what makes it socially significant?
• How do categories change the people placed in them – and how do people change categories?
• How is grouping by similarity different from predicting a label someone else supplied?
• When an algorithm finds groups “without labels,” what has already been decided before it runs?

In This Class

• Slides

Societal / Ethical Questions

• Who decides what a category means, and what happens to the people in it?
• Should people be able to see, challenge, or leave the categories a system assigns them?
• If removing a category (like gender or zip code) doesn’t stop a system from rebuilding it, what would?
• When a system both discovers and shapes our interests, who is responsible for the result?