Simulating realistic patient interactions with GenAI

How many of our students graduate with a firm grasp of theory, but feel unprepared to apply what they've learned in a professional setting? And can educators do more to bridge that gap?

This question was central to the development of the ATLAS platform. While educators are committed to providing immersive, practice-based learning experiences and traditional simulation models, these are often limited to small cohorts and are resource-intensive and difficult to scale.

ATLAS leverages artificial intelligence and large multimodal models to deliver high-fidelity, personalised simulations that replicate professional interactions at scale. The platform enables students to engage in dynamic clinical scenarios, receive targeted feedback, and iteratively refine their performance—without compromising educational rigour or authenticity.

With funding from the MEA (Monash Education Academy), we developed a couple of different models in the ATLAS platform, the most frequently used model is one where the student interacts with a simulated persona.

The student can input text which goes through our algorithm and provides structured parameters for response, which is then fed through a large language model API ChatGPT in a very constrained way to give a response that's appropriate to that professional situation.

For example, a pharmacy student may take on the role of a professional pharmacist interacting with a simulated patient.

	The student pharmacist conducts a medication history interview, asking relevant questions, documenting current medications, and identifying adherence concerns.
	The patient background including medical history and contextual details is stored in a secure database. When the student interacts with the simulated patient, their voice is recorded and transmitted as text to the large language model. The persona then responds in the voice of the persona transmitted as audio, read back using text-to-speech with a very lifelike voice.
	During the interaction, the student spoken questions are transcribed into text and processed by a large language model. The simulated patient responds in character, with audio generated via lifelike text to speech technology.
	The patient visual representation dynamically reflects emotional states such as frustration or appreciation depending on the pharmacist actions and communication style. Meanwhile, the student body language is captured via webcam allowing their facial expressions and nonverbal cues to be recorded and reviewed.
	At the conclusion of the interaction, the student receives detailed educator led feedback tailored to specific evaluation criteria. This includes reflections on their body language, the appropriateness of their verbal communication, and their professionalism within the simulated context.

Watch the video below to hear from an educator about their experience using the ATLAS platform

Watch the video below for a Pharmacy student's experience using the ATLAS platform

If you want to explore case based teaching or simulations that you think might be beneficial for students, reach out to ATLAS.AI@monash.edu to work with the ATLAS team to create a specialised ATLAS application for your educational use case.

Think about the professional future of your students and what kinds of communication skills would be empowering for students to practice in their learning. As in the case of medicine and pharmacy, this is a student taking the role of a professional pharmacist interacting with a patient or persona. For example in the Education faculty, students could undertake a mock job interview.