Abstract: [Purpose/Significance] This study explores the formation mechanism of continued usage intention for academic retrieval agents, providing theoretical frameworks and practical guidance for AI4S applications in academic retrieval. [Design/Methodology] Grounded in subject-based information retrieval needs, a three-dimensional construction framework for academic retrieval agents was proposed, encompassing technical architecture design, interaction experience optimization, and evaluation feedback mechanisms. The "CNKI Retrieval Master" agent was developed accordingly. Seven conditional variables were selected from three dimensions: system performance, user characteristics, and interaction utility. Through questionnaire surveys and in-depth interviews, the Fuzzy-Set Qualitative Comparative Analysis（fsQCA) was employed to uncover configurational pathways of the intention to continue using. [Findings/Conclusion] Four distinct configurations driving high intention to continue using are identified: system advantage-driven, learning empowerment-driven, key factor-driven, and goal achievement-driven. High-performance platform access emerges as a necessary condition for forming high intention to continue using, while "understanding of the trade-off of recall and precision" proves pivotal across nearly all pathways. The academic retrieval agent demonstrates a significant enabling effect on novices, embodying an important shift from purely "tool-oriented" to the integration of "tool + education" value. [Originality/Value] This study proposes a three-dimensional framework for academic retrieval agents and employs fsQCA to reveal diverse configurational pathways of continued usage intention formation, providing empirical evidence for differentiated agent design strategies.

Keywords: Agents, Academic retrieval agent, Continued usage intention, Information literacy education, Configurational pathway