Bayesian fairness is an overarching concept of fairness that explicitly takes into account the subjective uncertainty of the decision maker. This allows us to be as fair as possible, for most standard definitions of fairness, even under high uncertainty. It also enables decision making policies that explicitly take into account informational feedback effects, leading to an intertemporal notion of fairness.

In our AAAI presentation, and the accompanying paper, we present results under the notion of fairness as balance, a type of group fairness concept, where we need to protect disadvantaged groups. Feel free to check out our source code too.