Super Data Science: ML & AI Podcast with Jon Krohn avatar

Dexa/Super Data Science: ML & AI Podcast with Jon Krohn

Learn more
Published Oct 31, 2023

727: Unmasking A.I. Injustice — with Dr. Joy Buolamwini

Dr. Joy Buolamwini of the Algorithmic Justice League exposes the profound impacts of algorithmic bias in AI, highlighting the urgent need for intersectional ethics and accountability in technology to combat systemic inequalities and promote inclusive, fair data practices.
Episode Highlights
Super Data Science: ML & AI Podcast with Jon Krohn logo

Topics covered

Popular Clips

Episode Highlights

  • Gender Bias

    Gender bias in AI systems, particularly in facial recognition technologies, presents significant challenges. highlights how these systems often perform better on male-labeled faces than female ones, revealing a broader pattern of phenotypic bias 1. This bias is evident in the disparities in error rates, with darker-skinned women experiencing significantly higher error rates compared to lighter-skinned individuals 2. notes, "For pale males, it was 100% accurate on predicting gender, but with black males, it was 99% accurate."

    When you look at dark-skinned women, the error rate was around 47%.

    ---

    Such disparities underscore the need for more inclusive datasets and algorithms that account for diverse skin types and genders.

       

    Intersectionality

    Intersectional analysis is crucial in evaluating biases in AI systems. emphasizes that without an intersectional lens, evaluations miss the complete picture, potentially masking significant disparities 3. Her research, inspired by anti-discrimination law, shows that single-axis analysis fails to capture the complexities of discrimination, particularly at the intersections of race and gender 4. explains, "If you looked at demographics, so gender and phenotype, skin type, in this case, would there be a difference? Would the story vary with a sharper lens?"

    When we looked at the intersection, we would see a case where the gap could be as large as around 34% difference.

    ---

    This approach reveals the nuanced ways in which biases manifest, highlighting the importance of comprehensive testing that reflects real-world diversity.

       

    Corporate Accountability

    Corporate accountability in addressing AI biases is a mixed landscape. describes IBM's proactive stance, as they ceased selling facial recognition to law enforcement and worked on improving their models 5. In contrast, Amazon initially attempted to discredit research findings before eventually halting sales to law enforcement 6. states, "Amazon's initial approach was actually to attempt to discredit the research."

    These company commitments are a step in the right direction, but we can't rely on self-regulation or voluntary commitments for safety.

    ---

    Such responses highlight the need for external oversight and the development of an AI auditing ecosystem to ensure responsible AI deployment.

Related Episodes