18114 Fairness in AI – how to design ‘unbiased’ algorithms?

Hammer 14 Fairness in AI -how to design 'unbiased' algorithms?In healthcare, artificial intelligence (AI) technologies are increasingly being used in clinical practice, biomedical research, public health, and healthcare administration.Examples of applications include AI imaging technologies in radiology or cardiology, decision support in emergency medicine and surgery, technologies for home care, advances in drug discovery and personalized medicine, or facilitation of clinical trials.Despite the great opportunities offered by these AI applications, they pose various risks due to errors, misuse, unintentional bias, lack of transparency, invasion of privacy, or gaps in accountability [1].As the healthcare sector has a high human impact, it is classified as high risk according to the categories defined in the European Union's AI-Act [2].Therefore, specific measures are required by law to mitigate such risks.This chapter focuses on the question of what biases occur in AI technologies for healthcare and what measures can be taken to avoid such biases.