The development of advanced data processing methods and the introduction of the latest technologies into various areas of life are driving the creation of new artificial intelligence-powered systems. The Polish Economic Institute, a public economic think tank, has examined the application of such systems in healthcare. Artificial Intelligence is Already Present in Medicine The spectrum of applications for AI-based technologies is wide. It is used, for example, in wearable health monitoring devices, in medical imaging diagnostics, and in the analysis of laboratory data. According to estimates by Deloitte and MedTech Europe, the comprehensive use of AI in medicine will help reduce expenditures. With new solutions, savings of 170.9 billion to 212.4 billion euros annually are possible across the European Union. An example of implementing AI-based systems in Poland is the eRADS system being developed by the Information Processing Centre. The system automates diagnostics using deep neural networks. It then generates automatic and standardized reports from radiological examinations. This accelerates the work of doctors and radiologists. Previously, humans created such descriptions themselves, which not only extended diagnostic time but also limited its capabilities. The use of algorithms enables the processing of vast amounts of data in a relatively short time. The result is a faster and more accurate diagnosis, or a reduction in the need for repeat diagnostic tests. For instance, thanks to an algorithm developed by a team from the Silesian Centre for Heart Diseases in Zabrze, unnecessary coronary angiography can be avoided in 70% of patients referred for the procedure. Doctor's Decision or Algorithm's? However, the attitude of decision-makers – in this case, doctors or diagnosticians – towards the results of an algorithm can be problematic. Numerous studies and observations suggest that people's trust in algorithms during decision-making processes is ambiguous. In a decision-making situation where individuals have a choice between their own judgment and the suggestion of an algorithm's results, they are more likely to stick to their own assessment. This 'algorithm aversion' was described by researchers from the University of Pennsylvania. Their research indicates that people are particularly skeptical of predictions presented by algorithms, even though these algorithmic predictions are more accurate than human ones. Algorithm Transparency is Key These objections are compounded by natural inclinations to prefer natural factors (humans) over artificial ones (algorithms). Statistical decision support can appear artificial, mechanical, and lifeless. Regarding the application of AI in medicine, an important aspect is also the traditional perception of medical decisions as an art. This image does not align with the perceived coldness and sterility of an algorithm. However, according to researchers, algorithm aversion can be reduced if the user has insight into and the ability to modify the algorithm. This requires a transparent algorithm design, which remains a challenge for its creators. The aforementioned eRADS system allows the doctor to view the model and verify the results of an examination performed by the algorithm.