ScholarWorks@성균관대학교

초록

Short-form videos offer advantages of rapid information acquisition and intuitive content consumption. However, the combination of brevity and infinite scroll structures exacerbates habitual overuse problems. Existing intervention studies have failed to achieve fundamental behavioral change by relying on simple usage limits and one-time notifications, taking a uniform approach that does not consider individual viewing contexts. To overcome these limitations, this study proposes SHIFT, which applies If-Then theory. SHIFT guides users to establish specific plans of “if situation X occurs, then I will do action Y” and tracks actual execution to induce automatic behavioral change. The system collects real-time scroll patterns, automatically classifies viewing content using Vision-Language Models, and generates personalized intervention messages through LLM-based four distinct prompting strategies. A 4-week user study confirmed a 50% reduction in average daily usage time (p < 0.001) and achieved an 83% (p < 0.001) intervention success rate. © 2025 Elsevier B.V., All rights reserved.

키워드