Explainable Affirmative Action
Research Seminars: Virtual Market Design SeminarWhen organizations select from a set of applicants, they often use a priority ranking in tandem with policies designed to ensure diversity or boost the chances of applicants from certain groups. These policies add significant complexity to the selection process, and examples from around the world show that they often do not have the intended effect, or are accompanied by unintended consequences. Motivated by a desire to make these rules more transparent, the authors of this paper provide three axioms intended to capture explainability: monotonicity, lower invariance, and non-bossiness. They show that these axioms characterize a family of "outcome-based" selection rules. This family of selection rules is rich enough to incorporate many (but not all) existing reserve and quota policies. The authors illustrate their findings using examples from the allocation of affordable housing in New York City.
Personen