NASA’s new dark energy space telescope can also detect killer asteroids
technologyreview.com·2h ago
AI Research
arXiv CS.AI·Jun 12·5 min read
Divination by Prompt: LLM-Mediated Xuanxue on Chinese Social Media
TL;DR
A gap existed in understanding how large language models (LLMs) are used for divination on Chinese social media. Researchers conducted a mixed-methods study analyzing over 23,000 posts and 32 interviews to explore user interactions with LLMs in the context of Xuanxue, a term for mystical practices.
✦ Why It Matters
Engineers and researchers can explore user interaction patterns with AI to enhance AI-driven applications in cultural contexts.
Key Takeaways
- Investigate user engagement with AI in cultural practices.
- Consider the impact of confirmation bias in AI outputs.
- Explore collaborative features in AI tools for user empowerment.
- Assess the balance between technology and traditional practices.
Full Summary
The rise of large language models (LLMs) has led to their novel use in divination practices on Chinese social media, particularly within Xuanxue, which encompasses various mystical and spiritual traditions. Researchers employed a mixed-methods approach, analyzing over 23,000 posts from Xiaohongshu and conducting 32 semi-structured interviews with users and professional diviners.
Users typically consult LLMs for practical concerns like relationships and career advice, driven by curiosity and anxiety. A key finding is the collaborative refinement of prompts, where users actively shape their queries, leading to perceived accuracy through personal relevance and confirmation bias.
In contrast, professional diviners express skepticism about LLMs' spiritual capabilities, highlighting a divide in understanding between technological and traditional practices. This study situates LLM-mediated divination within broader anthropological and cognitive frameworks, suggesting that while it retains core functions of traditional divination, it also introduces new dynamics in authority and evaluation.