Stance Influences Your Thoughts: Psychology-Inspired Social Media Analytics

There are abundant user posts in social media which contain valuable information. Lots of previous studies focus on social media analytics, such as topic detection, sentiment prediction, and event trend analysis. According to psychological theories, namely affective forecasting, endowment effect, and negativity bias, user stance (one’s role in a specific social event, e.g. involvement) results in biased sentiment and attitude in real scenarios. However, user stance has not been taken into consideration in previous work. In most cases, user stance is a visible factor, so we argue that it should not be ignored. In this paper, we introduce user stance into two real scenarios (sentiment analysis and attitude prediction). Firstly, analyses on two real scenarios indicate that user stance does matter and provides more useful information for event analyses. Different user stance groups have significantly distinct sentiments and attitudes on an event (or a topic). By taking the differences into consideration, it is easy to get better mining results. Secondly, experimental results show that taking user stance information into account improves prediction results. Instead of designing a new algorithm, we propose that different algorithms should incorporate user’s stance information in online social event analysis. To the best of our knowledge, this is the first work which integrates psychological theories of user stance bias on understanding social events in the online environment.