INTEGRATION OF ARTIFICIAL INTELLIGENCE INTO MEDICAL EDUCATION AND FUTURE PRACTICE: STUDENTS’ EXPERIENCES, RISKS, AND PERSPECTIVES

Abstract

The article examines the features, priority goals, and risks of integrating artificial intelligence (AI) technologies into medical education. The work examines how future doctors implement digital tools in their training and assess their role in future practice. A survey of 79 4th and 6th year students of the Ternopil National Medical University named after I. Ya. Horbachevsky, who study “Obstetrics and Gynecology”, revealed a paradox. On the one hand, AI has become a part of daily routine for 67.1 % of respondents, another 31.6 % turn to it several times a week. The main advantage is the optimization of time spent: 97.5 % of respondents confirmed significant resource savings. However, the implementation is superficial, as 100 % rely on general-purpose text language models (ChatGPT, Claude) rather than specialized medical software. Expert assistants (UpToDate, Glass AI) are integrated by only 8.9 % of respondents, and image generation programs by 6.3 %. It was found that 83.5 % of students turn to AI to search for clinical protocols; 67.1 % to prepare for the KROK/OSKI exams; 67.1 % to solve clinical problems or write case histories, and only 17.7 % to translate articles. An important aspect was the analysis of self-assessment of prompting skills on a 5-point scale: 12.7 % of students rated them 5 points, 36.7 % rated them 4 points, 46.8 % rated them 3 points, and 3.8 % rated them 2 points. Despite high confidence, 93.7 % of respondents personally encountered AI “hallucinations” – the generation of non-existent facts. Future doctors are clearly aware of the risks of degradation of clinical thinking (63.3 %, n = 50), getting used to “easy answers” (65.3 %, n = 52), and unconscious assimilation of false information (43.0 %, n = 34). The introduction of mandatory discipline in medical AI was fully supported by 23.4 % (n = 18) of respondents, actively supported with reservations by 18.2 % (n = 14), and taken in a neutral position by 45.5 % (n = 35). Based on the data obtained, it is concluded that medical students demonstrate a high intensity of AI use against the background of a critical shortage of work with profile software and a high frequency of encounters with neural network errors. The results justify the need for a strategic transformation of higher medical education – a transition from ignoring technology to the implementation of systematic training in medical prompting, data verification methodology, and digital hygiene for the safe use of AI in future practice.