A Causal Framework to Measure and Mitigate Non-binary Treatment Discrimination

The recent publication titled 'A Causal Framework to Measure and Mitigate Non-binary Treatment Discrimination' highlights the limitations of traditional fairness studies in algorithmic decision-making, which often reduce complex processes to binary classifications. This simplification neglects the critical role of non-binary treatment decisions, such as bail conditions or loan terms, which can significantly influence outcomes like loan repayment or reoffending. The authors argue that these decisions should be central to fairness analyses. Their proposed causal framework allows for a clearer distinction between the covariates of decision subjects and the treatment decisions made by decision-makers. By applying this framework to analyze four widely used loan approval datasets, the authors reveal potential disparities in non-binary treatment decisions. This work not only underscores the necessity for incorporating treatment decisions into fairness assessments but also demonstrates that in…