The article discusses the rapidly expanding adoption of generative AI technologies, particularly in services providing increasingly precise responses. The authors evaluated how current AI services meet user needs and examined society’s depiction by Large Language Models (LLMs). A query experiment was conducted about social conventions in various countries. The researchers compared the LLMs’ value judgments with public data and suggested a decision-making model for value-conflicting scenarios. The study argues against AI making value judgments and advocates for a more critical perspective that defers judgmental capabilities to individuals. The research aims to empower anyone to receive safe and accurate value judgment-based outputs effectively.
Publication date: 17 Nov 2023
Project Page: Not Provided
Paper: https://arxiv.org/pdf/2311.09230
