Chen et al. (2009) did a research on evaluating the effectiveness of people-recommending algorithms on social networking sites. Based on a very solid literature survey, two types of recommender algorithms are elicited. One is based on the social network established by the website (Friend-of Friend, Sonar), while the other is on the similarity of the content of the user profiles (Content Matching, Content-plus-Link). Term-frequency and inverse-document-frequency (TF-IDF), which is a very important algorithm for information retrieval, is used to calculate the content similarity.
The research design of the study is very systematic and rigid in terms of recruiting users, evaluation instruments used, and data analysis methods. On the one hand, a survey was carried out to capture users’ attitude toward the recommendations given by the website with different algorithms. On the other hand, with higher ecological validity, they did a field study on whether user would make a connection or not, when a set of connections are recommended to him/her. More interestingly, they also recorded many qualitative data to support the analysis of their quantitative data.
The results of the study show that, generally, relationship based algorithms are more effective than content similarity algorithms. However, the latter is better at discovering potential friends.
Reference:
Chen, J., Geyer, W., Dugan, C., Muller, M. & Guy, I. (2009). Make new friends, but keep the old recommending people on social networking sites. In Proceedings of CHI 2009. 201-210.
