The Daily Ubuntu · dailyubuntu.com · Printable sheet
Thread 02 · Technology and Dehumanization · September 20, 2026 · 2 min read
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.

The Story Granary
It was a Wednesday evening, at the kitchen table.
A little girl borrowed her father's tablet. She typed, letter by letter: "Draw a doctor who looks like me."
The image appeared. A smiling woman, white coat, stethoscope. Straight hair, light skin.
The girl deleted it. She started again, adding "with braids." The result changed a little. Not enough.
She pushed the tablet away. "She doesn't look like me."
Her father searched for what to say. In the end, he took out some pencils and a blank sheet of paper.
(This scene is invented. What follows is not.)
What the girl felt, two journalists measured.
It was June 2023. Leonardo Nicoletti and Dina Bass, of Bloomberg, put an image generator, Stable Diffusion, to the test.
They asked it for more than 5,000 portraits: 14 occupations, and a few crime-related words. Then they compared the images with real U.S. job figures.
For "judge," women appeared in only 3% of the images. Yet they make up 34% of American judges.
High-paying jobs, meanwhile, were dominated by lighter skin.
For "inmate," more than 80% of the images showed darker skin. Yet people of colour make up less than half of American prisoners.
The conclusion was clear. The tool did not just reflect the world's biases. It magnified them.
Like a funhouse mirror that always enlarges the same features. A tool that learns our clichés can repeat them, louder.
These figures are American, and these tools have changed since. The study is a snapshot of one moment. But the question remains.
Because these images circulate in presentations, textbooks, advertising. Without a word, they say who looks like a judge. And who looks like a suspect.
Researchers, artists and organizations are working on more representative data and on tools to check these images. Anyone can help too, by looking critically at the images they use and share.
The chair of Black Girls Code, quoted by Bloomberg, summed up what was at stake:
“By seeing yourself or not, you learn that maybe you don't belong.”
We learn who we are in the eyes of others: Ubuntu knows this. Today, we also learn it in the images reflected back at us.
An image can open a place. It can also close it.
Somewhere, a child may be looking for her face on a screen. The blank sheets of paper, we can fill together.
And you?
What images do we want to show children, so they know they have a place?
My answers:
Figures
Sources
- Humans Are Biased. Generative AI Is Even Worse (Bloomberg, 2023) www.bloomberg.com
- Is AI more biased than humans? (Marketplace, 2023) www.marketplace.org
- Report 3180 (AI Incident Database) incidentdatabase.ai
Every figure on this page links to a verifiable public source.
Tags
Read also
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.
ReadA tool too good to be honest
The RCMP used facial recognition software built on photos that the commissioners found had been collected illegally. Even the police must know where their tools come from.
ReadThe 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.
Read