Supporting Self-Awareness of Smartphone Use with Passive Sensing and LLM-Driven Feedback

Smartphones provide considerable benefits in every day life, but can negatively impact users’ mental well-being when used exces- sively. While individuals can proactively manage their smartphone usage, accurately estimating usage time remains challenging. Re- cent advancements in Large Language Models (LLMs) offer signif- icant potential to increase users’ awareness of their smartphone usage habits through more natural conversational interactions. Cur- rent approaches employing LLMs for smartphone usage regulation primarily focus on directly restricting usage or providing conver- sational coaching and persuasive interactions. However, there is limited research on methods designed specifically to enhance users’ awareness of their habits and support sustained self-regulation. This paper presents a prototype system designed to promote self- awareness and support behaviour alignment around smartphone use. The system integrates passive monitoring of smartphone ac- tivity and physical movement with structured daily self-reflection, goal setting, and LLM-generated feedback. Users receive person- alised, reflective summaries informed by both sensor data and self- reported evaluations, with feedback tailored based on the alignment between observed and intended actions. The system was developed as a rapid prototype during a summer school workshop, and its feasibility was demonstrated through a series of simulated usage scenarios.