Large-scale moral machine experiment on large language models

2024-11-11

arXiv.org

AI Summary

This study systematically evaluates the alignment of 52 large language models (LLMs)—including GPT, Claude, Gemini, Llama, and Gemma variants, spanning multiple generations and both closed- and open-source paradigms—with human moral judgments in autonomous driving ethical dilemmas. Leveraging the Moral Machine framework and conjoint analysis, it conducts the first large-scale, cross-model quantitative assessment of ethical consistency. Key findings reveal: (1) open-source models exceeding 10B parameters exhibit significant *negative* correlation with human preferences—challenging the implicit assumption that model scaling inherently improves ethical alignment; (2) closed-source models and select open-source models >10B parameters achieve higher alignment; and (3) widespread overreliance on utilitarian principles is observed across models. The work uncovers a non-monotonic relationship among model scale, training paradigm, and ethical alignment, establishing an empirical benchmark and methodological foundation for trustworthy AI governance.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Philosophy and Ethics of AI: Artificial General Intelligence
• Humans and AI: Learning Human Values and Preferences

Application Category

• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems

Abstract

The rapid advancement of Large Language Models (LLMs) and their potential integration into autonomous driving systems necessitates understanding their moral decision-making capabilities. While our previous study examined four prominent LLMs using the Moral Machine experimental framework, the dynamic landscape of LLM development demands a more comprehensive analysis. Here, we evaluate moral judgments across 52 different LLMs, including multiple versions of proprietary models (GPT, Claude, Gemini) and open-source alternatives (Llama, Gemma), to assess their alignment with human moral preferences in autonomous driving scenarios. Using a conjoint analysis framework, we evaluated how closely LLM responses aligned with human preferences in ethical dilemmas and examined the effects of model size, updates, and architecture. Results showed that proprietary models and open-source models exceeding 10 billion parameters demonstrated relatively close alignment with human judgments, with a significant negative correlation between model size and distance from human judgments in open-source models. However, model updates did not consistently improve alignment with human preferences, and many LLMs showed excessive emphasis on specific ethical principles. These findings suggest that while increasing model size may naturally lead to more human-like moral judgments, practical implementation in autonomous driving systems requires careful consideration of the trade-off between judgment quality and computational efficiency. Our comprehensive analysis provides crucial insights for the ethical design of autonomous systems and highlights the importance of considering cultural contexts in AI moral decision-making.

Problem

• Large-scale Language Models
• Moral Decision-making
• Autonomous Vehicles

Innovation

• Ethical Machine Experiment
• Large-scale Language Models
• Moral Decision-making in AI