A Journey into Using Large Scale Datasets to Address Behavioral Research Questions

ABSTRACT With the growing availability of large datasets reflecting real human behavior, multimethod research in behavioral science has expanded significantly. In this talk, I will share my personal journey of acquiring and analyzing large datasets of real human decisions to address questions in consumer behavior. This includes techniques such as web scraping, application programming interfaces (APIs), and collaborations with online platforms to gather data, as well as transforming unstructured information into structured datasets suitable for statistical analysis. I will illustrate these methods through an exemplary project investigating the main determinants of consumers’ decisions about whom to follow on various online platforms. Specifically, we ask: (1) What determines the value consumers place on communicator credibility in following decisions? and (2) Which communicator cues influence these decisions under conditions of high versus low credibility premium? Analysis of four large datasets from following-enabled platforms reveals that consumers’ content consumption orientation—goal-directed (“search”) versus experiential (“scroll”)—plays a pivotal role. On search-oriented platforms (Yelp, Goodreads), perceived communicator credibility drives following decisions, while on scroll-oriented platforms (Twitter, Instagram), communicator likability takes precedence. Consequently, on scroll-driven platforms, communicators with positive sentiment attract the most followers, whereas on search-driven platforms, communicators with mixed sentiment gain the most followers. I will discuss both the challenges and advantages of incorporating large datasets into behavioral research, highlighting their potential to enhance understanding through multimethod research methodologies.