Divination by Prompt: LLM-Mediated Xuanxue on Chinese Social Media

2026-06-12 · Source: cs.AI updates on arXiv.org · Field: Science & Research — Social Sciences & Behavioral Studies, Research Methodology & Innovation, Human-AI Interaction Studies · Depth: Expert, quick

Summary

A systematic study by Chuang Li, Lixuan Wang, Yuqi Chen, and Ze Hong investigates LLM-mediated divination, termed Xuanxue, on Chinese social media. Analyzing over 23,000 posts and comments from Xiaohongshu and conducting 32 interviews, the research reveals users primarily consult large language models for pragmatic concerns like relationships, careers, and exams. This practice is driven by viral trends and anxiety, with users actively refining prompts to become "prompt engineers." While perceived efficacy is often positive, attributed to biographical fit and confirmation bias, users also employ verification methods such as repeated trials and cross-model comparisons. Professional diviners, however, dismiss LLMs as lacking genuine "spiritual power," highlighting a conflict between traditional beliefs and economic interests. The study concludes that LLM divination maintains core traditional functions while introducing scalability and prompt-driven co-production, altering how divinatory authority is established.

Key takeaway

For AI scientists and researchers studying human-AI interaction, this research highlights how LLMs are integrated into cultural practices like divination. You should consider the role of user-driven prompt engineering and cognitive biases in shaping perceived AI efficacy. This understanding is crucial for developing LLMs that are culturally sensitive and for anticipating their societal impacts beyond conventional applications.

Key insights

LLMs facilitate a new form of divination, Xuanxue, on Chinese social media, driven by user curiosity and anxiety.

Principles

• LLM divination efficacy is often justified by biographical fit and confirmation bias.
• Professional diviners assert LLMs lack "spiritual power" to protect their economic domain.
• LLM divination introduces scalability and prompt-driven co-production to traditional practices.

Method

Users engage in collaborative prompt refinement and verify LLM readings via repeated trials and cross-model comparisons.

In practice

• LLMs are used for pragmatic concerns like romantic relationships and career advice.
• Users can act as "prompt engineers" to guide AI-generated divinatory readings.

Topics

• Large Language Models
• Divination
• Xuanxue
• Chinese Social Media
• Prompt Engineering
• Human-AI Interaction

Best for: AI Scientist, Research Scientist, Domain Expert

Related on AIssential

• 7 Specific Unconventional Things to Do with Language Models — LLMs offer diverse unconventional applications beyond chat, enabled by intentional, role-based prompting.
• When Do LLMs Generate Realistic Social Networks? A Multi-Dimensional Study of Culture, Language, Scale, and Method — LLM-generated social networks are significantly influenced by prompt design, culture, language, and model scale.
• Mapping the Design Space of User Experience for Computer Use Agents — Understanding user experience is crucial for designing effective LLM-based computer use agents.
• Are we getting better at explaining things because of our interaction with LLMs? — Crafting clear LLM prompts may enhance human explanatory and communication abilities.
• Occupational Prompting Reveals Cultural Bias in Large Language Models — Occupational prompts elicit structured cultural value patterns in LLMs, even within a Western-leaning default.
• Leveraging LLMs in your Obsidian Notes — LLMs can transform note-taking by enabling conceptual search and conversational interaction with personal knowledge bases.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by cs.AI updates on arXiv.org.