fairness

All AI systems that process social or demographic data pertaining to features of human subjects must be designed to meet a minimum threshold of discriminatory non harm.

This entails that the datasets they use be equitable; that their model architectures only include reasonable features, processes, and analytical structures; that they do not have inequitable impact; and that they are implemented in an unbiased way.

Designers and implementers of AI systems must be able (1) to explain to affected stakeholders in everyday language how and why a model performed the way it did in a specific context and (2) to justify the ethical permissibility, the discriminatory non harm, and the public trustworthiness both of its outcome and of the processes behind its design and use.