This paper proposes an innovative relevance feedback algorithm for wall-content selection in social media. The procedure results in an iterative loop, which recursively updates a weighted distance. The distance is then used for finding multimedia items that are relevant to a user's preferences. To do so, the activity log of the user under investigation is considered and his/her attention at previous intervals is analyzed. Another novel point of the proposed approach is the incorporation of friends' attention into the relevance feedback scheme. In particular, interactions among users and posted multimedia items are considered as an explicit crowdsourcing activity. By this way some multimedia items receive more attention, while some others receive less or no attention. By analyzing these social interactions, a social computing framework is formed, which affects the evolution of the content selection process. Overall, the iterative relevance feedback algorithm takes into consideration visual features, activity log and social attention, in order to select the wall information of each social media user. Experimental results and comparisons on real data, exhibit the advantages of the proposed scheme and future directions are also discussed.