Smart med reminders: full automation hurts trust, study finds limits

A new study from researchers Liqian You and Jianlong Zhou examines how much automation is appropriate for smart medication support systems—tools that recognize pills, send reminders, and log doses. The team built a prototype and tested it with 53 participants across three conditions: confirmation required, automatic logging with undo, and fully automatic support. They also conducted in-depth interviews with 11 older adults, uncovering nuanced preferences tied to privacy concerns, digital confidence, and caregiver involvement.

Contrary to the assumption that more automation is better, the fully automatic condition was rated significantly lower in autonomy, trust, transparency, dignity, and satisfaction. Users valued systems that reduced routine effort but still allowed them to override or correct actions. The research provides empirical evidence for designing 'automation boundaries'—sweet spots where the system handles low-risk tasks while leaving high-stakes decisions to the user. These findings have direct implications for the growing market of smart health devices, especially for aging populations who may have varying comfort with AI.