This document presents a research paper that proposes a hierarchical attention model called HASC for social contextual image recommendation. It identifies three key aspects (upload history, social influence, owner admiration) that affect users' preferences for images. A hierarchical attention network is designed to model these aspects and their relationships. Experimental results on real-world datasets show the model outperforms baselines in image recommendation. The paper aims to better capture complex user preferences by fusing social network information with image and user data.