Discovering AI adoption patterns from big academic graph data

Although AI has been widely adopted by researchers in non-AI disciplines, the path to fully realizing its benefits through adoption remains unclear. A comprehensive understanding of AI adoption patterns can reveal who is able to leverage this emerging technology and in what ways, providing insights into the future direction of AI applications and research collaboration. This study leverages the Microsoft Academic Graph, a massive bibliographic dataset with detailed subfield information, to investigate AI adoption patterns among researchers in various disciplines (18 non-AI disciplines ranging from the humanities and social sciences to STEM), career stages (early, mid, and senior), and the interactions between these two aspects from 2006 onwards. Our findings indicate that researchers in economics and business can play an important bridging role in AI-related collaborations between STEM and social science researchers, who currently exhibit substantial disparities in AI adoption patterns. Late early-career to early mid-career researchers tend to adopt AI more actively than others, although this pattern varies across disciplines. In some fields, such as materials science, chemistry, and physics, early-career and senior researchers share a considerable level of common understanding and interest in AI, implying the potential for fruitful cross-seniority collaboration.