“…However, theoretical results state that it is impossible to satisfy different fairness notions at the same time (Chouldechova, 2017;Kleinberg et al, 2017). Not only fairness notions are in tension among each other (Alves et al, 2023), but also with other quality requirements of AI systems, such as predictive accuracy (Menon & Williamson, 2018), calibration (Pleiss et al, 2017), impact (Jorgensen et al, 2023), and privacy (Cummings et al, 2019), for which Pareto optimality should be considered (Wei & Niethammer, 2022). Moreover, the choice of a fairness metric requires to take into account several contrasting objectives: stakeholders' utility, human value alignment (Friedler et al, 2021), people's actual perception of fairness (Saha et al, 2020;Srivastava et al, 2019), and legal and normative constraints (Xenidis, 2020;Kroll et al, 2017).…”