Fairness of Exposure in Forensic Face Rankings

In information forensics, (police) agents are usually presented with a ranking of suspects similar to a certain face probe whose identity should be determined. Used for estimating the relevance score of possible suspects, deep face models have been proven to lead to undesirable discriminatory outcomes for certain demographic groups. Despite other non-personalised person rankings being actively investigated, forensic face rankings still represent an underexplored, yet important and peculiar, domain. In this ongoing project, we propose a framework consisting of six state-of-the-art face models and a public data set to quantify (disparate) exposure of demographic groups in forensic face rankings. Our results show that biases in this domain are not negligible and urgently call for ad hoc fairness notions and mitigation.