Practitioner-Centered Design: A Case Study of a Human-Computer Interaction-Clinician Collaboration on the Development of Prototype Virtual Reality Environments for the Treatment of Posttraumatic Stres

Ethics is a core element of medical practice, shaping clinicians’ responsibilities, guiding patient care, and shaping the education of future physicians. Ethics education requires physicians to not only learn ethical principles, but also to engage in applying them in practice. Thus, ethics is taught through multiple paradigms. One established model is the epistemological approach, where learners acquire knowledge of ethical theories and principles through didactic instruction. Another is the pragmatic-hermeneutical, dialogical approach, which emphasizes moral case deliberation with competing perspectives and values and fosters reflective judgment. Despite its importance, ethics training in medical education remains inconsistent, with a scarcity of dedicated resources, and significant variation in curricular emphasis, standardization, and reinforcement. Advancements in generative artificial intelligence (AI), particularly through natural language processing (NLP) and large language models (LLM)s, raise questions on whether AI can close the gaps in medical ethics training. LLMs are powerful tools for interpreting semantics, engaging complex prompts, and generating naturalistic dialogue, all of which may allow for practice in ethical reasoning. While early trials suggest LLMs may improve performance in clinical assessments such as Objective Structured Clinical Examinations (OSCEs), evidence for their use in ethics education remains limited. Most commercially available LLMs, such as GPT-4o, are not designed with pedagogical needs in mind. They are trained on vast general-domain corpora (e.g., “Common Crawl,” Wikipedia, professional exams) but lack domain-specific fine-tuning for health care ethics, mechanisms for weighting clinically authoritative sources, or the capacity to manifest native empathy. In response, we developed CALEB (Conversational Agent Learning Ethics Bot), an artificially intelligent conversational agent (AICA) designed to support the pragmatic-hermeneutical approach to ethics instruction. Leveraging prior work in extended reality (XR) AICAs and guidelines for the safe and ethical creation of AICAs, we designed CALEB as a holographic avatar to enhance presence, interactivity, and engagement in moral case deliberation.