A Conversational Ethics Bot (CALEB) Versus GPT-4o for Medical Ethics Education: Transcript Analysis and Extended...

Area of Science:

• Medical Education
• Artificial Intelligence
• Ethics

Background:

• Medical ethics education struggles with standardization and often relies on didactic methods.
• Current large language models (LLMs) lack embodied, conversational engagement for realistic ethics discussions.

Purpose of the Study:

• To develop and evaluate CALEB (Conversational Agent Learning Ethics Bot), an XR-enabled agent for pragmatic, case-based medical ethics deliberation.
• To compare CALEB's pedagogical distinctiveness against general-purpose LLMs like GPT-4o.

Main Methods:

• CALEB integrates a medical ethics knowledge base, persona design, and bounded generative architecture.
• A two-phase feasibility study compared CALEB and GPT-4o via transcripts and live XR interaction with medical ethicists.

Main Results:

• GPT-4o produced longer, comprehensive responses; CALEB generated shorter, principle-dense responses with first-person and emotional language.
• CALEB demonstrated significantly higher empathy scores (p < 0.001) in interpretive dimensions.
• Live XR interaction favored CALEB's perceived effectiveness over transcript-only review.

Conclusions:

• Domain-specific agents like CALEB are feasible and offer a distinct pedagogical approach compared to general LLMs.
• Specialized and foundational AI systems may complement each other in scalable medical ethics training.