A Theory-Driven Design Framework for Social Recommender Systems

Social recommender systems utilize data regarding users’ social relationships in filtering relevant information to users. To date, results show that incorporating social relationship data – beyond consumption profile similarity – is beneficial only in a very limited set of cases. The main conjecture of this study is that the inconclusive results are, at least to some extent, due to an under-specification of the nature of the social relations. To date, there exist no clear guidelines for using behavioral theory to guide systems design. Our primary objective is to propose a methodology for theory-driven design. We enhance Walls et al.’s (1992) IS Design Theory by introducing the notion of “applied behavioral theory,” as a means of better linking theory and system design. Our second objective is to apply our theory-driven design methodology to social recommender systems, with the aim of improving prediction accuracy. A behavioral study found that some social relationships (e.g., competence, benevolence) are most likely to affect a recipient’s advice-taking decision. We designed, developed, and tested a recommender system based on these principles, and found that the same types of relationships yield the best recommendation accuracy. This striking correspondence highlights the importance of behavioral theory in guiding system design. We discuss implications for design science and for research on recommender systems.