🤖

Combatting Algorithmic Bias in Technology

Aug 27, 2024

Lecture Notes: "The Coded Gaze" by Joy Buolamwini

Introduction

  • Joy Buolamwini introduces herself as a "poet of code" on a mission to combat "the coded gaze," a term she uses to describe algorithmic bias.
  • Algorithmic bias can lead to unfairness and can spread rapidly and on a large scale, resulting in exclusionary experiences and discriminatory practices.

Personal Experiences with Algorithmic Bias

  • Joy encountered algorithmic bias while developing the Aspire Mirror project, which involved using facial recognition software.
  • The software failed to detect her face unless she wore a white mask.
  • Previous experience with social robots also highlighted this bias when the robot couldn't detect her face during a project at Georgia Tech.
  • In Hong Kong, a social robot demo failed to recognize her face, using the same facial recognition software.

How Algorithmic Bias Occurs

  • Computer vision uses machine learning and training sets to recognize faces.
  • If training sets lack diversity, the system may fail to recognize faces that deviate from the established norm.
  • The opportunity to create more diverse training sets exists to better reflect humanity.

Consequences of Algorithmic Bias

  • Facial recognition is increasingly used by police departments in the US.
  • Georgetown Law reported that half of US adults have their faces in facial recognition networks which are unregulated.
  • Misidentification in facial recognition can lead to serious consequences, unlike the lighter implications on platforms like Facebook.
  • Algorithms extend beyond facial recognition into areas like hiring, loans, insurance, college admissions, and pricing.
  • Machine learning is also used in predictive policing and judicial decisions.

Solutions to Combat Algorithmic Bias

  • Importance of "incoding" movement: creating more inclusive code.
  • Three tenets of incoding:
    • Who codes matters: Assembly of diverse teams to check biases.
    • How we code matters: Incorporating fairness in system development.
    • Why we code matters: Making social change a priority.

Initiatives and Call to Action

  • Joy launched the Algorithmic Justice League for fairness in algorithms.
  • Invitation to join the incoding movement to build inclusive technology.
  • Call to report bias, request audits, and join the conversation using #codedgaze.
  • Encouragement to participate in creating inclusive training sets, such as a "Selfies for Inclusion" campaign.

Conclusion

  • Joy concludes with a call to action for creating a world where technology benefits everyone and promotes inclusivity.
  • Ends with an invitation to join her in the fight against the coded gaze.