Contraception-related topics in chat dialogues between healthcare students and generative AI patients: a natural...

Rapidly growing technology has enabled real-time digital services in healthcare, and new opportunities for sexual and reproductive health care. As assessing patient needs and communication effectively on digital platforms can be challenging, healthcare students must practise written communication, such as chat dialogues. Despite its increasing use in education, the way in which generative AI can enhance chat interactions between healthcare providers and patients remains poorly understood. The aim of this cross-sectional study was to explore contraception topics in chat dialogues between healthcare students and AI patients during AI simulations.

The AI application simulated a written chat dialogue between student and AI patient, using the CurreChat interface, to enable students to practise clinical skills and communication in digital health service chat dialogues. Purposive sampling was used to collect the data from fifth-year medical students (n = 24) and graduating midwifery students (n = 20) in higher education institutions of medicine and midwifery. Data were collected in August and October 2024. The data consisted of chat dialogues between healthcare students and generative AI patients. Natural language processing (NLP) and automated text analysis examined the contraception topics in the dialogues. The analysis software was based on pre-taught (self-supervised learning), industry-specific language models that detect meanings and their semantics in given texts.

The most significant result was that the students discussed essential aspects of contraception in the dialogues with the AI patients. Several topics in the students' part of the dialogues were similar to those in MeSH terminology and to work-related topics. The students' dialogues covered essential topics such as contraindications (114 times), contraceptive methods (93 times), and smoking (80 times), aligning with the Current Care Guideline.

Generative AI chat simulations can enhance the education of healthcare professionals globally in contraception issues by improving educational outcomes. To fully utilise the advantages of AI chat interactions, effective prompting is essential. NLP was an appropriate method for analysing the conversations and could be utilised more in future research across diverse healthcare settings.