AI tools now help decide who gets a loan, who gets hired and how long someone stays in jail. But these systems learn from data built by the same institutions that have harmed Black people for generations. These five essays show how that happens.
1. Machine Bias
Courts across the country use software to predict who will commit another crime. ProPublica checked the scores of more than 7,000 people arrested in Broward County, Florida. Black defendants were almost twice as likely as white defendants to be wrongly labeled high risk.
2. When the Robot Doesn’t See Dark Skin
Facial analysis software couldn’t detect Buolamwini’s face until she put on a white mask. She went on to test systems from IBM, Microsoft and Face++. They got darker-skinned women wrong up to 35% of the time and lighter-skinned men less than 1% of the time.
Read it at The New York Times →
3. Google Has a Striking History of Bias Against Black Girls
Noble searched “black girls” on Google, looking for things her stepdaughter and nieces might enjoy. The top results were pornography. Her essay traces how a search engine that runs on advertising decides what the public sees about Black women.
4. How Our Data Encodes Systematic Racism
Raji audits AI systems for bias. Researchers are often told that data doesn’t lie. In her experience, datasets carry the racism of the institutions that produced them. She asks the field to take responsibility for what goes into its models.
Read it at MIT Technology Review →
5. Humans Are Biased. Generative AI Is Even Worse
Bloomberg asked an image generator to make more than 5,000 pictures of people at work. Lighter-skinned faces filled the high-paying jobs. More than 80% of the images for “inmate” showed people with darker skin, even though people of color make up less than half of the US prison population.
- Written by
- Team ARD
- First published
- in Anti-Racism Daily
- Topics
- Technology
- Tags
- AIReading list