Gender bias in medicine continues to contribute to diagnostic errors, particularly in cases where diseases manifest differently in women and men. Conditions more commonly studied or recognized in male patients, such as myocardial infarction, can go underdiagnosed or misdiagnosed in women due to variations in symptom presentation. This leads to delayed treatments, repeated consultations, and even adverse clinical outcomes. To address this challenge from an educational standpoint, we propose the development and implementation of AI virtual patients specifically designed to train healthcare students and professionals in gender-sensitive diagnostic reasoning.

Computerized simulations used for medical training, or more shortly "virtual patients", offer a safe, scalable, and cost-effective approach to developing diagnostic and therapeutic skills. Conversational virtual patients, driven by generative AI, enhance this model by enabling dynamic, realistic interactions between learners and simulated patients, closely resembling real-life clinical encounters. Our project introduces a set of gender-informed virtual patients tailored to medical, nursing, and psychology education. These simulations focus on diseases that present distinct symptoms depending on the patient’s gender, aiming to improve early recognition and decision-making accuracy.

The project is structured around three core phases. First, a flexible platform was developed to host AI-based virtual patients, enabling easy creation of new cases and automated performance assessment of learners. Second, this platform is being used to design and implement a library of gender-differentiated virtual patient scenarios—such as myocardial infarction and depression—ensuring diversity and clinical relevance. Third, the educational impact is being evaluated through pre- and post-intervention studies involving students and professionals.

While still in progress, the project has already demonstrated promising results. A prototype version of the platform has been developed and is currently being used in medical, nursing, and psychology courses. Upcoming phases will focus on the development of more specialized patient sets and the collection of quantitative data on learning gains. These findings will help assess the potential of gender-specific virtual patients to mitigate diagnostic disparities and strengthen clinical training programs.
By integrating AI with evidence-based pedagogical design, this initiative advances the use of virtual patients as a powerful tool for health professions education. It supports the development of future clinicians who are better equipped to recognize and address gender-based variations in disease presentation, ultimately contributing to safer, more equitable patient care.

Keywords: Virtual patients, gender bias, healthcare education, generative AI, clinical training tools.