Co-Speculation: Public Perceptions and Ethical Boundaries of AI-Enabled Household Products Based on LDA Topic Modeling and Speculative Design

Abstract

This study examines public perceptions and ethical boundaries of AI-enabled household products in the Chinese consumer market. A multi-source corpus was organized by analytical function rather than platform type. During the retrieval window from February 7 to 12, 2026, 3,431 raw public records were collected, and 3,186 valid records were retained after initial cleaning. A final LDA-cleaned modeling corpus of 567 texts was then used for both LDA topic modeling and SnowNLP sentiment analysis, yielding 452 neutral, 107 negative, and 8 positive texts. The results indicate that public perception has shifted from isolated smart devices toward domestic intelligent agents and whole-home systems, while concerns over privacy, cameras, connectivity, and loss of control form a distinct critical dimension. Based on these findings, a speculative design workshop translated empirical concerns into three scenarios with low, medium, and high levels of intelligence. The study shows that convenience and trust remain structurally unresolved in AI-enabled household products and proposes design implications for preserving user agency, privacy transparency, and controllable human-AI interaction.