AI fairness: not just a dream

AI Ethics

Joe Franklin

Associate Data Literacy and Essentials Manager, DataCamp

Fairness in AI

  • Fairness: Ensure no group is favored over another
  • Concerns race, gender, socioeconomic status, etc.

An checklist containing race, gender, and socioeconomic status.

  • AI should predict patient outcomes equitably
  • There should be no bias towards any specific group

A red cross on top of the word "bias".

AI Ethics

Why does fairness matter?

  • AI's rapid processing can result in large-scale impacts

An icon illustrating the large-scale impacts of AI.

An icon depicting vulnerability.

  • Fairness prevents negative targeting of vulnerable populations
  • Essential for responsible AI implementation, ensures equitable consideration for all
1 Icons made by noomtah & Parzival' 1997 from www.flaticon.com
AI Ethics

Promoting fairness

  • Fairness promotion is challenging but possible
  • Reduces potential bias by omitting certain variables
  • Variables include race, gender, age, socioeconomic status, sexual orientation, religion

A person trying to fathom "fairness through unawareness".

AI Ethics

Unintentional issues exist

 

  • Even with unawareness, unintentional bias can still occur

 

  • Robust strategies needed to ensure fairness

An icon showing a patient's chest x-ray.

1 Icons made by Freepik from www.flaticon.com
AI Ethics

Minimizing bias

  • The main objective of AI fairness is minimizing bias
  • The first step is acknowledging bias exists
  • Remain skeptical and vigilant of AI
  • Conduct frequent monitoring and audits for fairness
AI Ethics

Let's practice!

AI Ethics

Preparing Video For Download...