Evaluating Fairness in the Presence of Spatial Autocorrelation

In spite of considerable practical importance, current algorithmic fairness literature lacks in technical methods to account for underlying geographic dependency while evaluating or mitigating fairness issues for spatial data. We initiate the study of spatial fairness in this paper, taking the first step towards formalizing this line of quantitative methods. Fairness considerations for spatial data often get confounded by the underlying spatial autocorrelation. We propose hypothesis testing methodology to detect the presence and strength of this effect, then mitigate it using a spatial filtering-based approach -- in order to enable application of existing bias detection metrics. We evaluate our proposed methodology through numerical experiments on real and synthetic datasets, demonstrating that in presence of several types of confounding effects due to the underlying spatial structure our testing methods perform well in maintaining low type-II errors and nominal type-I errors.