Selective Consolidation in AI Agent Research: A Co-Citation Analysis of Its Intellectual Structure and Relevance for Human–Computer Interaction

Research on Artificial Intelligence (AI) agents has expanded rapidly across dis-ciplines, generating a diverse but fragmented body of knowledge. Despite the proliferation of domain-specific reviews, a field-level understanding of how AI agent research is intellectually structured and how it evolves over time remains limited, particularly from the perspective of human–computer interaction (HCI). This study aims to map the intellectual structure and temporal evolution of AI agent research and to identify the major conceptual streams shaping the devel-opment of the field. To achieve this, we conduct a co-citation analysis of AI agent publications indexed in the Web of Science. Using CiteSpace, the analysis identifies ten co-citation clusters that capture the main conceptual streams shaping AI agent research. The results show that the field is organized around multiple, partially connected research logics, includ-ing human–agent interaction, application-driven studies, organizational contexts, and governance-related concerns. An examination of citation bursts and intellec-tual turning points reveals that the field evolves through a process of selective consolidation, whereby a limited number of influential works periodically reshape research trajectories and establish dominant reference points. By uncovering the intellectual structure and dynamics of AI agent research, this study provides a field-level perspective that can support more cumulative and integrative scholarship. For the HCI community, the findings offer a struc-tured overview of the conceptual foundations informing the design, study, and evaluation of AI-driven interactive systems.