Effort-aware Fairness: Incorporating a Philosophy-informed, Human-centered Notion of Effort into Algorithmic Fairness Metrics

Effort matters more than current status in fairness judgments: A human subjects experiment found that people prioritize the temporal trajectory (how someone's circumstances have changed over time) over aggregate values when evaluating fairness, suggesting current AI fairness metrics that only look at final outcomes miss an important dimension of how humans understand justice.

Traditional fairness metrics may penalize people trying to improve: Standard metrics like demographic parity don't account for effort spent to overcome disadvantages, potentially leading to unfair algorithmic decisions against individuals who have worked hard to improve their circumstances but remain disadvantaged due to systemic factors beyond their control.

Philosophy can inform better AI fairness design: The paper grounds its approach in philosophical concepts of "Force" and temporal trajectories rather than relying solely on statistical definitions, demonstrating that incorporating humanistic perspectives can lead to more ethically grounded fairness frameworks.

Effort-aware fairness has practical applications: The researchers developed concrete methods to measure effort-aware fairness in real-world contexts like criminal justice and personal finance, suggesting this approach could be implemented by AI auditors to identify and correct unfair decisions in high-stakes domains.

Current AI auditing practices may be incomplete: By showing that effort and trajectory are crucial to human fairness judgments, this work suggests that organizations auditing AI systems for bias should expand beyond demographic parity and similar metrics to consider how individuals' efforts and circumstances have evolved over time.