New research quantifies unavoidable manipulation in AI fairness audits

A new research paper explores the inherent limitations of fairness auditing in automated decision-making systems. The study quantifies the unavoidable manipulation that can occur even with finite audit resources, framing the problem as a min-max optimization between a company and a budget-constrained auditor. The findings establish lower bounds on post-audit demographic parity deviation, demonstrating that while increased auditing resources can reduce manipulation, they cannot eliminate it entirely. AI

IMPACT Highlights fundamental limitations in certifying AI fairness with finite resources, suggesting ongoing challenges in deploying equitable automated decision-making systems.

RANK_REASON The cluster contains a research paper published on arXiv detailing theoretical and empirical findings on AI fairness auditing. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• automated decision-making
• black-box fairness auditing
• Company Manipulation
• Demographic Parity
• Fairness Auditing
• lending
• Min-max optimization

• paper
• safety

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →