Research: Do LLMs Have Values?

The article investigates the values of popular generative Large Language Models (LLMs) such as ChatGPT, DeepSeek, and Gemini, using the Portrait Values Questionnaire-Revised (PVQ-RR) to assess their alignment with human values. The research highlights that while LLMs generally emphasize universalistic and pro-social values, there are notable differences among them, particularly in areas like benevolence, self-direction, and adherence to rules. The findings suggest that understanding these value profiles can help organizations strategically select LLMs that align with their specific needs and ethical considerations. The authors advocate for greater transparency and accountability in AI development to mitigate biases and enhance the alignment of AI outputs with human values. [Extracted from the article]