Mapping the Evolution of Human-in-the-Loop Artificial Intelligence: A Bibliometric Analysis

Context: Human-in-the-loop (HITL) artificial intelligence has evolved from human-assisted model training toward broader human–AI collaboration, oversight, and decision-making. Its rapidly expanding and multidisciplinary literature, however, remains fragmented across technological and application domains, creating a need for systematic mapping of the field.

Aim: This study maps the evolution, scholarly impact, geographical and institutional structure, collaboration patterns, and conceptual development of HITL AI research from 2016 to 2026.

Approach: A bibliometric analysis was conducted on 2,071 Scopus-indexed publications. Bibliometrix and its Biblioshiny interface were used for bibliometric performance analysis, and VOSviewer was used for keyword co-occurrence mapping and visualization.

Key Findings: HITL AI research recorded a 55.9% annual growth rate and involved 8,319 authors publishing across 1,029 sources. The United States and China led publication output. Keyword analysis identified 462 keywords organized into five clusters, with prominent themes including deep learning, reinforcement learning, generative AI, decision support, responsible AI, ethics, and human oversight.

Interpretation: The findings indicate that HITL AI is developing into an increasingly multidisciplinary field in which human participation extends beyond model training toward collaboration, oversight, and accountable AI deployment.

Contribution: This study provides an integrated bibliometric mapping of HITL AI that consolidates its growth, scholarly impact, leading contributors, collaboration patterns, and conceptual structure while identifying emerging directions for future research.