Event Details

Title: Causal Equal Protection as Algorithmic Fairness 
Abstract: As decisions are automatized, computer scientists, philosophers and legal scholars have been exploring plausible formal criteria of algorithmic fairness. One of the most discussed is classification parity, roughly defined as equality in the rates of errors across protected groups. The talk revisits recent objections against this principle in light of causal analyses of algorithmic fairness. A novel principle is proposed: causal equal protection. It requires that individuals not be subject to uneven risks of error because of their socially salient characteristics. The explicit use of protected characteristics, however, may be required if it equalizes these risks.
Please join the Philosophy Department for this talk at 4:40 p.m. on Thursday, Oct. 3, in the Harmon Room of the DeWitt Wallace Library. All are welcome. Refreshments will be served.