The rapid di!usion of short-form video platforms has intensified highly immersive interaction patterns, often leading to excessive and unintentional consumption behaviors commonly referred to as doomscrolling. Despite growing concern about its psychological and behavioral impact, limited research has focused on its real-time detection through objective behavioral data. This thesis presents the design and experimental validation of a mobile applica- tion for the detection of doomscrolling behavior on the most widely used short-video platform, namely TikTok. The system uses Android Accessibility Services to collect processed interaction data, including scroll velocity, content skipping rate, interaction frequency, watch time, and session duration. A set of behavioral metrics was defined to model user activity and compute personalized detection thresholds.