Session-based Search Behavior in Naturalistic Settings for Learning-related Tasks

Abstract

In this research, we investigate the behavioral patterns exhibited in different search sessions as users attempt to complete search tasks of increasing cognitive complexity. The search tasks, which are exploratory in nature, have been designed using the Taxonomy of Educational Objectives, and are presented to the users hierarchically. We capture naturalistic search behavior of the users in real world (non-lab) setting using a Chrome browser plugin. The research analyzes the web log data of the users to assess if and how the web search behavior of the users changes over different search sessions. We also look at the different demographic factors like age and gender, educational factors like the academic background, read and write proficiency in English, and search skills to determine if these factors influence the web search behavior of the users. Our results indicate that search sessions have significant effects on the web search behavior of the users. Most of the web search behaviors differed significantly across search sessions. Of the secondary factors, gender showed significant effect on the query reformulations (measured using average edit distance between queries) and query length (measured using number of words per query) while year of study affected only the average query length. Search experience had significant effect on all the web search behaviors.