User Generated vs. Supported Contents: Which One Can Better Predict Basic Human Values?

摘要

Every individual possess a set of Basic Human Values such as self-direction, power, and hedonism. These values drive an individual to commit actions in various situations in her daily lives. Values represent one's attitudes, opinions, thoughts and goals in life, and can regulate a variety of human behaviors and manners that an individual shows in the society. In this paper, we identify five higher-level values from social media interactions by analyzing two types of contents: user generated and user supported. More importantly, we identify which type of content can better predict which human values in different scenarios, which ultimately helps us to predict human values for both silent and active users. We also build a combined value prediction model by integrating different types of interaction features, which can more accurately capture the human values than that of a single feature based model. We also build separate models for silent and active users of SNS to effectively predict values for different types of SNS users. Finally, we compare the strength of different types of models to predict values from social media usage effectively.