Disappointed with Siri: Expectation-experience gaps in human-AI interaction

초록

Voice assistants such as Siri increasingly mediate everyday tasks, yet negative experiences with these systems remain understudied. Guided by social representation theory, we use a sequential mixed-methods design-topic modeling of user narratives followed by in-depth interviews-to characterize dissatisfaction with a voice-based AI and to explain how its meanings differ by users' gender. Topic modeling surfaces a broad "inconvenience/ disruption" cluster alongside frequent references to speech-recognition errors. Interviews then reveal the interpretive logics beneath these signals: men tend to read failures as breaches of technical performance and task logic, whereas women more often construe the same events as violations of social expectations. These gendered interpretations show that dissatisfaction is not merely an individual usability outcome but a socially anchored perception organized by shared representational frames. The study contributes (1) a theoretically grounded account of how gender structures sense-making around AI malfunctions, (2) a methodological synthesis that links computational signals to qualitative representation mapping, and (3) design implications that anticipate divergent expectations without reinforcing stereotypes. By moving beyond frequency counts to interpretive coherence, the work advances understanding of why the same Siri behavior can produce different forms of dissatisfaction across users.

키워드