Nowadays, some applications are having difficulty in retaining users after pulling them in. To enhance user stickiness, algorithms are widely used by major applications due to their advantage of accurate push. Recommendation algorithms have changed the Internets business logic and information dissemination modes from pursuing "information bombing" to "precision strike". Based on the S-O-R model, this paper investigates the influence of recommendation algorithm characteristics (recommendation quality, personalization, usability, and user trust) on college students' satisfaction and continuance use intention of Douyin. Structural equation modelling analysis was conducted in this paper with the help of a questionnaire survey and partial least squares (PLS). It was found that personalization positively affects user attitude and user cognition; usability positively affects user cognition and emotional response; and user trust positively affects user attitude, user cognition and emotional response. While none of the stimuli, recommendation quality, has a significant effect on users' psychological emotions. User attitude, cognition, and emotional response all significantly affect user satisfaction, which in turn significantly affects user continuance use intention. This study provides a quantitative study for understanding user behaviors on social media platforms, and offers suggestions for the optimization of recommendation systems.