Using Trust in Recommender Systems: An Experimental Analysis