It’s about time: counterfactual fairness and temporal depth

Focusing on time opens up interesting lines of inquiry for algorithmic fairness. In the framework of counterfactual fairness, we can use temporal depth of counterfactuals to reason about common fairness ideals like opportunity, merit, and responsibility. In typical fairness applications greater temporal depth generally corresponds to stronger fairness requirements. We relate counterfactual depth to other causal criteria like direct and indirect effects, and comment on long-standing debates about causation without manipulation and the use of socially constructed traits like race and gender as variables with causal effects. There are diverse and potentially conflicting criteria for algorithmic fairness. Heuristics like temporal depth can help us reason about fairness in a unified way, compare differing criteria, and make good decisions.