Joy Buolamwini: Fighting Algorithmic Bias in AI and Facial Recognition

The Coded Gaze: Personal Discovery

• 💡 Joy Buolamwini, a post-graduate student at MIT, discovered that advanced facial recognition software failed to detect her face unless she wore a white mask.
• 📌 This experience, along with a similar one in undergrad with a robot, led her to coin the term "The Coded Gaze", highlighting how technology can render certain individuals invisible.
• 🧠 She recognized a pattern of systemic failure behind these isolated incidents, leading her to deeper research into algorithmic bias.

Groundbreaking Research: Gender Shades

• 🔬 In 2018, Buolamwini co-authored the "Gender Shades" study with Timnit Gebru, scientifically testing commercial gender classification systems from major tech companies.
• 📊 The study revealed a dramatic performance disparity: systems worked almost perfectly for light-skinned men (error rate <1%) but failed significantly for darker-skinned women (error rate up to 35%).
• ⚠️ This research proved that AI systems learn biases from non-diverse training data, leading to a "lazy math" that optimizes for patterns most present in the data, often male and light-skinned.

Industry Impact and Activism

• 🚀 The "Gender Shades" publication had a seismic impact on the industry, forcing companies like IBM and Microsoft to acknowledge and commit to improving their algorithms.
• ✅ The intense debate led to a moratorium on selling facial recognition technology to police by Amazon and other companies in 2020.
• 🗣️ Buolamwini's work highlighted how human biases are absorbed, coded, and amplified by algorithms, impacting crucial decisions from job interviews to judicial sentences.

The Algorithmic Justice League (AJL)

• 🛠️ Buolamwini founded the Algorithmic Justice League (AJL), using research, art, and political activism to fight for AI justice and frame it as a civil rights issue.
• 🎭 Through art, like her spoken word poem "AI Woman", she viscerally demonstrates how AI systems fail to recognize iconic Black women, connecting "silicon bias" to societal bias.
• 🎯 AJL has achieved concrete victories, such as successfully campaigning against the IRS's mandatory facial recognition system for account access (DumpID campaign).
• 💡 Projects like Voicing Eras (voice recognition bias) and CRASH (training citizens to identify algorithmic biases) demonstrate AJL's expanded scope beyond facial recognition.

Towards Algorithmic Equity

• 🌟 Buolamwini, described as a "Poet of Code" and the "conscience of the AI revolution," advocates for compassion through computation and ensuring technology serves everyone.
• ⚖️ Her work has contributed to discussions about an "AI Bill of Rights" in the US Congress, proposing principles like the right to know when an algorithm is used and the right to contest its decisions.
• 🌱 The ultimate challenge is to move beyond removing bias to actively designing systems for equity and justice from conception, creating a new "social algorithm."