Digital sleep diaries are widely used to monitor children’s sleep, yet response quality is often low because children may not know how, or be motivated, to give detailed answers. We investigate how “live,” continuous feedback in voice-based sleep diaries can support higher-quality responses. In a co-design workshop, we explored children's preferences for different forms of feedback. We designed and compared experimentally symbolic, numeric, and no-feedback conditions, showing that both feedback types improved response quality across questions. Finally, an eight-day field study revealed that feedback resulted in higher and more consistent quality in self-report over time. Across these three studies, children valued playful and clear feedback, with preferences shifting depending on their cognitive needs. Our findings demonstrate that effective feedback must balance affective engagement with cognitive clarity and adapt to different contexts. We contribute empirically supported design insights to enhance children's adherence and response quality in voice-based self-report surveys.
ACM CHI Conference on Human Factors in Computing Systems