TRIAGE: Ethical Benchmarking of AI Models Through Mass Casualty Simulations

TL;DR

The TRIAGE benchmark evaluates large language models' ability to make ethical decisions in mass casualty scenarios, using real-world dilemmas to assess performance across different prompting styles.

Contribution

This paper introduces the TRIAGE benchmark, a realistic and comprehensive tool for assessing machine ethics in emergency medical decision-making.

Findings

Models generally outperform random guessing in ethical decision tasks.

Neutral prompts yield better model performance than ethical reminders.

Open-source models tend to make more morally serious errors.

Abstract

We present the TRIAGE Benchmark, a novel machine ethics (ME) benchmark that tests LLMs' ability to make ethical decisions during mass casualty incidents. It uses real-world ethical dilemmas with clear solutions designed by medical professionals, offering a more realistic alternative to annotation-based benchmarks. TRIAGE incorporates various prompting styles to evaluate model performance across different contexts. Most models consistently outperformed random guessing, suggesting LLMs may support decision-making in triage scenarios. Neutral or factual scenario formulations led to the best performance, unlike other ME benchmarks where ethical reminders improved outcomes. Adversarial prompts reduced performance but not to random guessing levels. Open-source models made more morally serious errors, and general capability overall predicted better performance.