Dynamic creation of nationwide virtual panels for collective behavior prediction

Conceptual UI-design of profile-based behavior predictor

The ability to predict, explain, and change the behavior of other humans has long been of interest to both the scientific community and the business world. Decades of research has made great advances in improving our understanding of human behavior online and offline, especially with the advent of social media. Predicting how individuals will react to future social, political, and economic events can be challenging and complex, however. More importantly, providing an explanation as to why such behavior occurs and determining if it can be changed is even more challenging as this would require extensive and detailed profiling of individuals over time.

The overall objective of this research is to develop a predictive system that enables stakeholders (e.g. policymakers, politicians, celebrities, and business organizations) to know how people will behave in the future and that provides a solid understanding of these behaviors to help stakeholders’ strategic planning. In particular, we will exploit past behaviors to predict future behaviors as the former are considered to be vital for personalization and recommendation systems but are yet to be explored in public opinion and human behavior studies.

The specific aim of this research is to develop a system that predicts behaviors (i.e. sentiments and stances expressed by individuals on social media) regarding a future news event based on the individuals’ background factors and past behaviors and provides an explanation of the predicted behaviors with detailed user profiles.

This research makes empirical contributions to both the social science and computer science fields. Specifically, it advances the state of knowledge regarding how individuals’ intrinsic and extrinsic characteristics and past behavior determine their future behaviors, how human behaviors are connected to each other, how to model news events and individual behaviors (sentiments and stance regarding a news event) on social media, and the extent to which individuals’ intrinsic and extrinsic characteristics and past behaviors can determine their future behaviors. The resulting system will enable us to have a precise understanding of what makes the public happy, concerned, angry, and prone to act so that better solutions can be provided for the stakeholders and the risk of a bad reputation can be reduced.

Jisun An
Jisun An
Assistant Professor

My research interests include computational social science, social media analysis, natural language processing and artificial intelligence.