This paper argues that statistical inference of social data (e.g., stock market, human/organizational behavior, etc.) cannot be analyzed with the same statistical tools used in natural sciences. In traditional statistical inference, when one examines the probability of events, the uncertainty lies in unknown outcomes. In social-science settings, however, in addition to outcome uncertainty, one often does not know the underlying distribution of events. So, on top of outcome uncertainty, we have to deal with distribution uncertainty (also called ambiguity in the economics literature). We discuss how inference should be done when there is ambiguity in social data.
Professor Michael Zhang is the Associate Dean of Innovation and Impact and a Professor of Decision Sciences and Managerial Economics at the CUHK Business School, Chinese University of Hong Kong. He holds a PhD in Management from MIT Sloan School of Management, an MSc in Management, a BE in Computer Science and a BA in English from Tsinghua University. His works study pricing of information goods, online advertising, innovation and incentives, and use of machine learning in financial markets. His research has appeared in premier academic journals such as American Economic Review, Management Science, Marketing Science, Journal of Marketing, MIS Quarterly, Information Systems Research, Journal of MIS, Decision Support Systems, and Journal of Interactive Marketing. He serves as Senior Editor for Information Systems Research, Associate Editor for Management Science, Guest Associate Editor for MIS Quarterly.