Cognitive-Aware User Search Behavior Simulation

Saber Zerhoudi, Michael Granitzer

Abstract: In Information Retrieval (IR), evolving models have begun to address the complexity of user search behavior, as traditional approaches often oversimplify the nuanced needs of web search users. While effective for simpler informational needs, these models struggle to capture the dynamic intricacies involved in complex search tasks. To address these limitations, we introduce the Cognitive-Aware Complex Searcher Model (CACSM), an advanced model that simulates user actions and integrates a deep understanding of users' cognitive states into the search process. CACSM enhances the existing Complex Searcher Model by incorporating cognitive state updates that dynamically reflect users' knowledge and information needs throughout the search session. By utilizing traditional RNN-based (CACSMum) and Large Language Model-based (CACSMllm) user model strategies, CACSM simulates user actions and adapts the system's responses based on the evolving user states. This approach offers a granular understanding of user interactions and provides a comprehensive framework for capturing and responding to the complex cognitive processes involved in digital libraries.

External IDs:doi:10.1145/3677389.3702598