The Broad View of Task Type Using Path Analysis

Abstract

Many past research efforts have examined the relationships between observable search behaviors, the task that drives searching, and user characteristics such as search expertise and topic familiarity. These studies often look at pairwise relationships between characteristics to show that task characteristics or user characteristics can be distinguished by differences in browsing behavior. Recent work has moved toward detecting and predicting task from browsing behavior but has not additionally considered user or session characteristics that clearly affect browsing. To what extent should user characteristics and unlogged traits be considered when inferring task type from browsing behavior? This paper examines this complex relationship through the lens of path analysis, showing that such a holistic view should be considered in future task prediction research. Future task prediction work should consider direct links from task to behavior, indirect links through other factors, and other characteristics that affect browsing.