The Hidden Biases Built Into Every AI System | Dr. Joy Buolamwini

April 16, 2026

AI Summary

5 min read

In 2017, as a graduate student at MIT, Dr. Joy Buolamwini was building an art project—a mirror that would let her digitally shape-shift into someone like Serena Williams. The computer vision software meant to detect her face simply could not see it. When she held a white Halloween mask to the camera, the software immediately recognized it as a human face. That moment, which she calls encountering the "coded gaze," launched a research career that has reshaped how the tech industry understands bias in artificial intelligence.

How Bias Gets Into AI Systems

Buolamwini explains that AI systems are essentially pattern recognition technologies trained on data. The problem begins with what she calls "pale male data sets"—collections of images that are supposed to represent all of humanity but might be 70% male-labeled faces and 80% lighter-skinned faces. The mechanism behind this skew is revealing: many early data sets were built using publicly available photos of elected officials, who globally are disproportionately male and lighter-skinned. Even more insidious, the face detection systems used to scrape those images in the first place were themselves biased—they systematically excluded darker-skinned faces, creating a self-reinforcing cycle of exclusion. Buolamwini calls this a "power shadow," where existing societal inequalities become embedded in the infrastructure that train

Continue reading the full summary in the app — free to try.

Read Full Summary →

Free • No credit card required

Listen to Audio Summary Open in App

Never miss an episode of I've Got Questions with Sinead Bovell

Get every new episode summarized in your inbox — free, ~5 minutes to read.

No spam. Unsubscribe anytime.

What you'll learn

1 (02:03) **The Coded Gaze Origin Story** - Dr. Buolamwini describes discovering facial recognition bias while building an art project at MIT
2 (06:48) **How Bias Enters AI Systems** - Explains the root cause of algorithmic bias through data and training
3 (09:25) **AI Scoring and the Surveillance State** - Walks through the invisible ways AI shapes everyday life
4 (11:50) **False Arrest Case Study: Angela Lips** - Concrete example of AI surveillance harming an innocent person
5 (13:30) **Data Deletion and the "Opt-Out Club"** - Practical advice on resisting biometric data collection
6 (15:40) **Shackles of Convenience** - Critique of the "friction-free" narrative in airports and beyond
7 (23:37) **Kill Us Slowly vs. Existential Risk** - Distinguishes acute harms from systemic, accumulating damage

+ Full timestamped outline available in the app

Show Notes

Dr. Joy Buolamwini is a computer scientist, AI researcher, and founder of the Algorithmic Justice League. Her work has exposed how bias in artificial intelligence systems can reinforce inequality at scale and why the risks of AI go far beyond the headlines.

In this conversation, Dr. Buolamwini and I explore how AI systems inherit human bias and why these biases are not just technical flaws, but deeper issues rooted in the data used to train AI.

We also talk about the invisible ways AI is already shaping our lives, from algorithmic scoring and data brokers to biometric surveillance systems that most people don’t even realize they are part of.

Finally, we explore what she means when she says AI could “kill us slowly.” Not through one dramatic event, but through long-term, systemic harm that limits opportunities, reinforces inequality, and quietly reshapes society over time.

Grab a copy of Dr. Joy Buolamwini's book here.

Timestamps:

00:00 – Can AI “Kill Us Slowly”? 02:03 – The Origin of the “Coded Gaze” 06:56 – How Bias Gets Embedded in AI Systems 09:25 – The AI Systems Quietly Shaping Your Life 13:30 – Why Biometrics and Surveillance Matter 16:09 – The Risks of Facial Recognition at Airports 23:15 – Existential Risk vs. Systemic Harm 24:29 – What Joy Means by “AI Could Kill Us Slowly” 29:02 – The Corporate Surveillance State 30:49 – When AI Gets It Wrong in the Real World 34:07 – Bias in Healthcare and Wearables 39:41 – How Bias Shows Up in Hiring and Opportunity 46:36 – The Hidden Risks of AI Photo Trends 48:26 – What Parents Should Know About Posting Kids Online 52:18 – The Future of “Faceless” Resistance 55:40 – What You Can Do Right Now 01:01:16 – “Data Isn’t Destiny” 01:06:37 – What Tech Builders Need to Do 01:12:14 – How to Get Involved 01:13:17 – Poetry, Resistance, and Closing Reflections