Short-form video platforms such as Instagram Reels have become highly popular among students, raising concerns about their impact on sustained attention and academic performance. This study examines the relationship between Instagram usage patterns, attention span, and academic outcomes using survey-based data and machine learning techniques. Data were collected from undergraduate students through structured questionnaires covering screen time, study duration, sleep habits, and perceived distraction. A composite cognitive metric termed Brain Score was derived to categorize attention levels into low, medium, and high classes. Several supervised learning models were applied, and the Random Forest classifier achieved the best predictive performance. The findings indicate that excessive short-form video consumption is strongly associated with reduced attention span and lower academic performance, highlighting the need for balanced digital engagement.