Leveraging clinical epidemiology concepts to strengthen machine learning fairness evaluations

Key Points

• The research aims to explore how clinical epidemiology can inform and enhance fairness evaluations in machine learning.
• Analyzed parallels between machine learning fairness and clinical epidemiology.
• Identified fairness criteria and root causes of unfairness.
• Discussed considerations related to multiple testing.
• Demonstrated that clinical epidemiology principles can enhance fairness assessment methodologies in machine learning.
• Uncovered potential improvements in the articulation of fairness criteria.
• Highlighted the importance of methodological rigor in fairness evaluations.

Abstract