Table 5.
Topics in the expert interviews—Connotations, typical quotations, and model mapping.
| Topic | Connotation (key point) | Typical quotation (quoted from what the expert said) | Corresponding relation with the research model |
|---|---|---|---|
| Visual presentation | - Initial attraction and uninstallation threshold - Avoidance of the “uncanny valley” effect and prevention of homogenization - Needed customizable options such as skins/themes |
“Young users will not give you a chance to ‘warm up slowly’; if it does not appeal at first glance, they’ll uninstall it.” said E2. | Corresponding to visual appeal (visceral level); directly influencing user experience. |
| Interactive intelligence | - Natural language understanding and conversation coherence - Emotional recognition and empathetic response - Cross-session learning and memory |
“Users will soon realize it’s just an advanced voice assistant if an AI can only respond mechanically.” said E1. | Corresponding to interaction quality; intersecting with perceived personalization to jointly influence user experience and acceptance intention |
| Personalized perception | - Memorizing personal details and recalling them at the right time - Dynamically adjusting tone/rhythm - Ensuring transparent and controllable data collection |
“I’d be instantly touched if an AI can remember my complaints yesterday and proactively ask me about that today.” said E3. | Corresponding to perceived personalization/intelligent adaptability; positively influencing user experience and acceptance intention |
| Ethical acceptability | - Encryption of chat data and storage location - Psychological risks of excessive dependence - Social stigma and positioning strategies |
“What users are worried about is not the AI itself, but rather who might witness their vulnerabilities.” said E2. | Serving as a moderating variable (trust) influencing the strength of the path from emotional design to behavioral intention |
| Usage scenario | - High-frequency periods: late at night, during commuting, and in fragmented rest intervals - Typical emotions: loneliness, stress, boredom - Cultural differences: Japanese and Korean > European and American > Chinese |
“Many people do not lack friends; rather, they cannot find someone to chat with at 2 a.m.” said E1. | Serving as a control variable (contextual fit) for adjusting the overall model error |