일상생활 속의 AI

Convenience does not remove the need to understand what data is used, what the system can change, and how to correct a mistake.

주요 시사점

• Identify the system’s action level.
• Check data and permission boundaries.
• Preserve manual control and correction.

Identify whether a feature predicts, recommends, generates, or acts. A suggested route and an automated purchase have different consequences. Review the permissions and data involved, and keep confirmation for actions that affect money, accounts, safety, or other people. Treat personalized outputs as conditional suggestions. Recommendations can reflect incomplete data, feedback loops, or business objectives. Check important information against a trusted source and retain a manual alternative when the service is unavailable. Consider accessibility and household context. Voice controls, captions, translation, and reminders can increase independence, while poor recognition or an inaccessible fallback can create new barriers. Test with the people who will actually use the feature. Review privacy and retention settings periodically. A device, app, and connected cloud service may each store different information. Make it easy to see, edit, delete, or stop a feature without losing unrelated functionality.

1. Imagine an assistant recommending a cheaper flight and offering to buy it.
2. Show the route, date, price, baggage terms, and destination before any purchase step.
3. Require the appropriate confirmation and verify the booking record afterward.

This hypothetical workflow separates helpful suggestion from consequential action.