Authors: Nicky Chong-White¹, Jhanvi Ramani¹, Catherine Kwok¹, Marisa Poulos¹
¹National Acoustic Laboratories
Good communication skills are essential in hearing care, but they are hard to teach consistently at scale, and a gap in available training tools. Students and even experienced clinicians might find themselves unprepared for complex, emotional, or resistant clients that they face in clinic.
To help address this gap, we created NAL Virtual Personas (NAL-VP) — an AI-powered platform that uses generative AI to simulate realistic, interactive conversations with virtual clients. The tool is designed to help clinicians practise the kinds of conversations they’ll face in clinical scenarios, including taking case histories, explaining test results, and guiding clients through decisions.
The system uses OpenAI GPT-4o combined with structured logic, dynamic dialogue summarisation, and prompt engineering to support realistic, context-aware responses to the users’ questions and guidance. Each persona is crafted based on background research, real patient interviews and clinical case examples, and can respond in different ways depending on how the user communicates. The platform supports both self-guided and educator-led use, with voice input, transcript saving, and multilingual capabilities.
I’ll share how and why NAL-VP was developed, show a short demo, and reflect on early feedback from students, educators, and global hearing care providers. Already in use in a university audiology program, it has potential for training and onboarding across the hearing industry. This tool shows how large language models can be applied in a focused, human-centred way to improve soft skills in audiology.
