A new research paper published on arXiv explores the moral reasoning of large language models (LLMs) in healthcare decision-making, specifically concerning the allocation of scarce resources. The study found a significant "judgment-consequence gap," where LLMs, unlike humans, largely fail to let a patient's responsibility for their own health-harming behaviors influence decisions about care denial or resource allocation. Instead, LLMs tend to default to random allocation, even when acknowledging patient culpability, and place a greater emphasis on information access when determining responsibility. AI
IMPACT Reveals a potential disconnect in LLM ethical frameworks compared to human reasoning, highlighting risks in high-stakes applications like healthcare.
RANK_REASON Research paper published on arXiv detailing LLM behavior in a specific domain.
Read on Hugging Face Daily Papers →
- health care
- Homo sapiens
- patient
- arXiv
- LLM
- The Judgment-Consequence Gap: LLM Moral Reasoning in Healthcare Decisions
AI-generated summary · Google Gemini · from 2 sources. How we write summaries →
