Exploratory search represents a critical pathway for users to access high-value and serendipitous information. Despite the recognized diversity of exploratory search tasks, systematic understanding of their behavioral distinctions remains limited, hindering the development of tailored search support mechanisms. Acknowledging queries as the primary interface between dynamic information needs and system capabilities, this study examines differences in query behavior across learning and investigation tasks under varying levels of cognitive complexity.
A controlled laboratory experiment was conducted to analyze 13 query behavior indicators across four task types: simple learning, complex learning, simple investigation and complex investigation. The study involved 37 participants, with 34 providing 136 valid task observations for analysis.
Analysis of Variance (ANOVA) results revealed that learning tasks involved shorter queries and more frequent use of generalization and specification strategies than investigation tasks. Simpler tasks contained more queries and newly reformulated queries compared to complex tasks. Complex investigation tasks required more time for initial query formulation and subsequent reformulations. A gradient boosting tree classifier achieved 84.65% accuracy in task differentiation, with UniQueryNum, SpecificationNum and AvgQueryTerm being the most discriminative indicators.
This study systematically compared learning and investigation tasks through multiple query dimensions-quantity, length, temporal characteristic and reformulation strategies-providing granular insights into exploratory search behavior. The findings offer practical implications for enhancing search system adaptability, refining implicit feedback mechanisms and developing targeted user training protocols.
