Empirical Analysis of General Bayesian Network-Driven Online Recommendation Mechanism: Emphasis on User Satisfaction and Psychological Reactance

As Internet technology permeates into our daily lives, the importance of online recommendation grows quite rapidly. Needless to say, the success of the online recommendation mechanism depends heavily on how much users are satisfied by the recommendation results. Moreover, it is essential that users feel less psychological reactance about the recommendation results when it comes to the effectiveness of the online recommendation mechanism. To accomplish those two goals, we propose a new online recommendation mechanism on the basis of the General Bayesian Network (GBN). It is common that users have a large number of personal wants and needs in their minds when using the recommendation mechanism. In this case, GBN is very effective in selecting a small number of relevant factors and organizing a set of causal relationships among them. An advantage of GBN like this allows users to have much satisfaction about the recommendation results, and enjoy rich implications from them. We implemented the proposed GBN-driven online recommendation mechanism (GBNOREM), and tested with users who are suffering from minor health problems and therefore in need of proper menus at the right restaurants. Empirical results clearly revealed that the proposed GBNOREM can enhance users’ perceived satisfaction about the recommendation results, while reducing the level of users’ perceived psychological reactance.