How ethical rules shape large language models in medical dilemmas
MedConstBench: Benchmarking How Constitutions Steer LLMs in Medical Ethics
Artificial IntelligenceComputation and Language
Summary
Large language models are increasingly used to answer medical questions, but their ethical behavior depends on the 'constitution' rules set when they are used. The authors created MedConstBench, a test that measures how different ethical guidelines affect model decisions in medical dilemmas. They found that the impact of these guidelines varies by model and ethical principle, so no single set of rules works best for all. To improve this, they designed a method that selects the best ethical rules for each model to reduce harm and adapt to new models effectively.
What this means in practice
- •For medical ai developers: Select ethical guidelines tailored to each AI model when deploying medical question answering systems to reduce harm in clinical decisions.
- •For ai ethics teams: Use systematic benchmarking to evaluate and compare how various ethical rule sets affect AI behavior in medical scenario responses.
Authors
Yibo Yang, Zerui Huang, Haomiao Yang, Zixia Wang
Abstract
As large language models advance rapidly, they are increasingly used to answer medical questions, and many such models are now deployed locally. Once a model is deployed, its behavior is governed mainly by the ethical policy written into its constitution. However, most existing work focuses on factual correctness rather than the trade-off between competing ethical values. Even benchmarks that do evaluate ethics treat it as a fixed property of the model rather than as a variable, so how a constitution reshapes medical-ethics behavior remains largely unexplored. We introduce MedConstBench, a medical-ethics benchmark that treats the deployment constitution as a controlled variable, is organized around Beauchamp's four principles, and is scored by human-calibrated multi-model peer review. It includes MedDilemma, our own dataset of clinical dilemmas, and supports a systematic analysis of how constitutions steer LLMs. Evaluating ten open-source LLMs under four constitutions, we find that the effect of a constitution is dimension-selective and model-dependent, and that no constitution is universally optimal. Motivated by these findings, we propose a do-no-harm adaptive selector that chooses a constitution per model, matching the per-model oracle while guaranteeing a worst-case floor and generalizing to unseen models at low cost.