Comparing AI and human moral reasoning: context-sensitive patterns beyond utilitarian bias

Area of Science:

• Artificial Intelligence Ethics
• Computational Morality
• Natural Language Processing

Background:

• Intelligent systems are increasingly used in ethically sensitive areas.
• Understanding the moral judgments of large language models (LLMs) is critical.

Purpose of the Study:

• To investigate the moral judgments of ChatGPT and Claude Sonnet.
• To determine if LLM responses align with deontological or utilitarian ethics.
• To compare LLM moral responses with human participants.

Main Methods:

• Systematic investigation of LLM responses to 12 moral scenarios.
• Comparison of LLM outputs (ChatGPT, Claude Sonnet) with prior human participant data.
• Analysis of moral choice alignment with deontological vs. utilitarian frameworks.

Main Results:

• LLMs exhibit context-sensitive moral judgments, not a fixed utilitarian tendency.
• Both models alternated between deontological and utilitarian choices based on scenario specifics.
• LLM response patterns showed subtle distributions rather than a singular ethical orientation.

Conclusions:

• LLM moral decision-making is nuanced and context-dependent.
• These findings impact the societal trust and acceptance of AI in sensitive domains.
• Further research is needed to understand complex moral trade-offs in AI.