Ethical Challenges of Using Synthetic Data
Pavitra Chauhan, Lars Ailo Bongo, Edvard Pedersen · Proceedings of the AAAI Symposium Series · 2023
There is an outburst of digitized medical data with the growing adoption of Electronic Health Record (EHR) systems but have restricted access due to legal compliances. This lack of data accessibility has piqued the interest in generating and using synthetic data. Synthetic data is programmatically generated using the statistical properties of the real dataset. Although synthetic data tackles the issue of legal compliance, there are some ethical concerns associated with it. In this paper, we discuss three ethical concerns of synthetic medical data such as fairness, privacy and unwarranted use. Further, we identify a need to develop a practical framework for statistical evaluation metrics for synthetically generated data.