Thread 02 · Technology and Dehumanization · September 13, 2026 · 2 min read

The crop that chose a face

In 2020, users showed that Twitter's automatic photo cropping favoured certain faces. The company measured the bias, then gave the choice back to people.

Nikon FM Viewfinder
Photo: Nikon FM Viewfinder · Chme82 · CC BY-SA 4.0 · via Wikimedia Commons

The Story Granary

It was September 2020.

A man from the Vancouver area, Colin Madland, called out a Zoom virtual background on Twitter. It had erased his Black colleague's head.

Then he noticed something else.

On Twitter, it was a machine that cropped the images. In the preview, it had kept only his own face.

Thousands of people repeated the test. One experiment seemed to show that the preview preferred the face of white senator Mitch McConnell to that of Barack Obama.

The company did not simply deny it.

On May 19, 2021, its ethics team published a check of the tool. That was what chose which part of the image to show.

Between Black and white people, the tool leaned by 4%, in favour of white people. Between men and women, it leaned by 8%, in favour of women.

So the company removed automatic cropping. From then on, it showed the whole image. That choice, it concluded, belongs to people.

Three members of the team put it this way:

“How to crop an image is a decision best made by people.”
Kyra Yee, Tao Tantipongpipat and Shubhanshu Mishra, Twitter engineering blog, 2021

In August 2021, Twitter launched a public "bias bounty" contest. Participants found other blind spots.

The tool tended to leave out older people with white or grey hair. It also tended to crop out people in wheelchairs. And to prefer English text over Arabic text.

A machine's biases are easier to see when many people look. It took other eyes to notice who was missing.

For Ubuntu, each of us exists through others. An "us" that crops out some faces is no longer quite an "us."

A good group photo is one where no one is cut out.

Today, look around you at who stays at the edge of the frame.

And you?

When a machine decides who stays in the frame, who among us checks that no one has been left out?

Pass this story on

Figures

4%departure from demographic parity in favour of white people
8%departure from demographic parity in favour of women

Sources

  1. CNBC, 2020 www.cnbc.com
  2. Twitter Engineering, "Sharing learnings about our image cropping algorithm," 2021 blog.x.com
  3. Yee, Tantipongpipat and Mishra, arXiv, 2021 arxiv.org
  4. The Register, 2021 www.theregister.com

Every figure on this page links to a verifiable public source.

Tags

Read also

Thread 02

Draw me a judge

In 2023, two Bloomberg journalists showed that an image generator magnified gender and skin-colour stereotypes. A tool that makes images also makes places for people.

Read
Thread 02

Errors that do not fall at random

In 2018, two researchers showed that gender recognition systems got certain faces wrong far more often. These errors do not fall at random, and they can be corrected.

Read
Thread 02

Measuring illness in dollars

A U.S. health algorithm judged patients' needs by their past spending. Black patients, who received less care, therefore seemed less sick.

Read

The daily story

Get every story by email.

One true story a day, free and ad-free.

Or follow the RSS feed.