Statistical Fairness - Turing Commons

An online platform to support open dialogue and reflection about the responsible design, development, and deployment of data-driven technologies.

Statistical Fairness ¶ The Computational Allure of Statistical Fairness ¶ Humans have tried to quantify fairness for centuries. The Law of Hammurabi, for example, states that if a man "destroy's the eye of a man's slave or beaks a bone of a man's slave, he shall pay one-half his price". 1 We rightfully recoil at the idea of such a law being espoused and adhered to in modern society. However, the idea that we can quantify fairness is still very much alive, and still not without its problems. Photograph of the basalt stele with the Law of Hammurabi inscribed (By Mbzt - Own work, CC BY 3.0, https

saved by

• Mari M

related reading

• [arxiv] an algorithmic framework for fairness elicitationarxiv.org
• Relative notions of fairnessfairmlbook.org
• A statistical approach to model evaluations \ Anthropicanthropic.com
• Measuring Fairnesspair.withgoogle.com
• AI and prediction of criminal justice risk: revisiting the COMPAS dataset, experiments, and the problem of algorithmic bias and fairnesslink.springer.com
• [1911.03020] A Human-in-the-loop Framework to Construct Context-aware Mathematical Notions of Outcome Fairnessarxiv.org
• Constitutional AI: Harmlessness from AI Feedbackarxiv.org
• Self-Fulfilling Misalignment Data Might Be Poisoning Our AI Modelsturntrout.com
• Frontiers | On Consequentialism and Fairnessfrontiersin.org
• Algorithmic Accountabilitydatasociety.net
• Frontiers | On Consequentialism and Fairnessfrontiersin.org
• [1907.10323] Fairness in Reinforcement Learningarxiv.org