AI language model rivals expert ethicist in perceived moral expertise.

Area of Science:

• Artificial Intelligence Ethics
• Human-Computer Interaction
• Moral Psychology

Background:

• Public perception of AI expertise is widespread, yet the quality of AI's moral judgment is uncertain.
• Large language models (LLMs) show promise in moral alignment tasks, but their reasoning capabilities are under scrutiny.
• Increasing LLM integration in decision-making necessitates evaluating their moral reasoning alongside aligned judgments.

Purpose of the Study:

• To assess public perception of AI-generated moral expertise compared to human ethical advice.
• To evaluate GPT-4o's ethical advice against a renowned human ethicist and a general American sample.
• To explore the potential of LLMs as sources of moral guidance and decision-making support.

Main Methods:

• The study advanced the Moral Turing Test framework.
• Participants rated ethical advice from GPT-4o and 'The Ethicist' column.
• Comparisons were made between GPT models, a representative American sample, and a renowned ethicist.

Main Results:

• Americans rated GPT-4o's ethical advice as more moral, trustworthy, thoughtful, and correct than 'The Ethicist'.
• Participants perceived GPT models as superior to both the general American sample and the renowned ethicist in moral justification.
• Findings suggest a growing tendency to view LLM outputs as credible sources of moral expertise.

Conclusions:

• LLMs like GPT-4o may be perceived as viable sources of moral expertise, complementing human guidance.
• The study highlights the potential for LLMs to influence user moral reasoning.
• Careful programming of ethical guidelines within LLMs is crucial due to their potential societal impact.