Equal Opportunity in Online Classification with Partial Feedback
This work addresses fairness concerns in real-world applications such as criminal recidivism prediction and lending, where partial feedback is common, but it is incremental as it builds on existing fairness constraints like equal opportunity.
The paper tackles the problem of ensuring fairness in online classification with partial feedback, where true labels are only observed for positive classifications, by requiring algorithms to satisfy statistical fairness constraints like equal opportunity at every round. The authors provide upper and lower bounds showing the cost of this constraint is mild in terms of regret rate and present an oracle-efficient algorithm achieving the upper bound.
We study an online classification problem with partial feedback in which individuals arrive one at a time from a fixed but unknown distribution, and must be classified as positive or negative. Our algorithm only observes the true label of an individual if they are given a positive classification. This setting captures many classification problems for which fairness is a concern: for example, in criminal recidivism prediction, recidivism is only observed if the inmate is released; in lending applications, loan repayment is only observed if the loan is granted. We require that our algorithms satisfy common statistical fairness constraints (such as equalizing false positive or negative rates -- introduced as "equal opportunity" in Hardt et al. (2016)) at every round, with respect to the underlying distribution. We give upper and lower bounds characterizing the cost of this constraint in terms of the regret rate (and show that it is mild), and give an oracle efficient algorithm that achieves the upper bound.