Measuring fairness preferences is important for artificial intelligence in health care

The debate about fairness of artificial intelligence (AI) in health care is gaining momentum. At present, the focus of the debate is on identifying unfair outcomes resulting from biased algorithmic decision making. Bias can arise unintentionally from either real-world implementation or technical reasons such as biased training data or erroneous assumptions underlying algorithm design.1,2 Although this biased algorithmic decision making can be the source of unfairness for some algorithmic outcomes, it is not universally applicable for every outcome that is regarded unfair.