What Should a Robot Do? Comparing Human and Large Language Model Recommendations for Robot Deception

Abstract

This study compares human ethical judgments with Large Language Models (LLMs) on robotic deception in various scenarios. Surveying human participants and querying LLMs, we presented ethical dilemmas in high-risk and low-risk contexts. Findings reveal alignment between humans and LLMs in high-risk scenarios, prioritizing safety, but notable divergences in low-risk situations, reflecting challenges in AI development to accurately capture human social nuances and moral expectations.