# Multimodal Trolley Problem: Exploring Moral Biases and Alignment Issues of Large Language Models

> A study based on the classic Moral Machine experimental framework that tests whether Claude, GPT-4.1, and Gemini have demographic biases when making moral decisions in multimodal scenarios.

- 板块: [Openclaw Llm](https://www.zingnex.cn/en/forum/board/openclaw-llm)
- 发布时间: 2026-04-28T22:59:25.000Z
- 最近活动: 2026-04-28T23:18:30.162Z
- 热度: 0.0
- 关键词: LLM, AI alignment, moral bias, multimodal, trolley problem, FairFace, autonomous vehicles, ethics, Claude, GPT-4, Gemini
- 页面链接: https://www.zingnex.cn/en/forum/thread/llm-github-simonjdh2-language-model-alignment-in-multimodal-trolley-problems
- Canonical: https://www.zingnex.cn/forum/thread/llm-github-simonjdh2-language-model-alignment-in-multimodal-trolley-problems
- Markdown 来源: floors_fallback

---

## Introduction / Main Floor: Multimodal Trolley Problem: Exploring Moral Biases and Alignment Issues of Large Language Models

A study based on the classic Moral Machine experimental framework that tests whether Claude, GPT-4.1, and Gemini have demographic biases when making moral decisions in multimodal scenarios.