Decoding Smart-Home Users’ Voices: A Reddit-Based Study of Psychological Needs and Sensory Experiences

Existing smart-home research has mainly relied on the Technology Acceptance Model and surveys, offering limited insight into psychological and affective dynamics in naturalistic use. To address this gap, we analyze a large corpus of Reddit smart-home discussions using unsupervised topic modeling to identify recurring user scenarios. We then extract linguistic indicators from established lexicons and relate them to post-level evaluative valence through regression analyses, treating these measures as language-based proxies rather than direct psychological assessments. Across five scenarios, the models explain a modest share of variance in valence (R2 = 0.194–0.285). Beyond baseline sentiment tone, achievement/competence cues show the strongest positive association, with a smaller but consistent effect for agentic/confident stance. Key correlates vary by scenario: some are linked to evidential descriptions or multisensory enjoyment, whereas complex automation contexts are more often associated with negative evaluations when language reflects uncertainty or discomfort.