"How to Prevent Discriminatory Outcomes in Machine Learning", Mar 2018

...Designed and used well, machine learning systems can help to eliminate the kind of human bias in decision-making that society has been working hard to stamp out. However, it is also possible for machine learning systems to reinforce systemic bias and discrimination and prevent dignity assurance...

...Discriminatory outcomes not only violate human rights, they also undermine public trust in machine learning...While ML has implications for many human rights, not least the right to privacy, we focus on discrimination because of the growing evidence of its salience to a wide range of private-sector entities globally...This white paper makes the case that businesses need to integrate principles of nondiscrimination and empathy into their human rights due diligence...

...Drawing on existing work, we propose four central principles to combat bias in machine learning and uphold human rights and dignity: Active Inclusion...Fairness...Right to Understanding...Access to Remedy...We recommend three steps for companies: 1. Identifying human rights risks linked to business operations...2. Taking effective action to prevent and mitigate risks...3. Being transparent about efforts to identify, prevent, and mitigate human rights risks...