Recent advances in wearable sensing and digital phenotyping have enabled mental health tracking by combining self-reports with physiological and behavioral indicators. However, conventional interfaces record only final answers and overlook how users hesitate and introspect. This gap contributes to discrepancies between self-reports and behavioral signals. In this study, we propose an introspective sensing framework that treats the response process itself as a sensing target. We combine an explicit uncertain option ‘?’ with decision latency measured by a smartwatch accelerometer. Forty-one participants answered ten introspection-inducing questions about personality-related tendencies on a smartwatch interface offering three symbolic response options: ‘◯’ (yes), ‘?’, and ‘×’ (no). Experimental results showed that 61% of participants used ‘?’ at least once. Decision latency was longest for ‘?’, followed by ‘×’ and ‘◯’. When responses were grouped into three types (immediate, internalized, and externalized) based on latency and ‘?’ usage, each type exhibited distinct decision patterns. These patterns suggest different balances of intuitive versus deliberative processing and different ways of expressing uncertainty. Our findings indicate that ambiguous responses and decision latency provide behavioral features that complement physiological indicators and support low-burden, stigma-sensitive mental health tracking.