A meso-scale cartography of the AI ecosystem

TL;DR

This paper provides a comprehensive large-scale analysis of the AI ecosystem, mapping its internal structure, disciplinary spread, and collaboration patterns across scientific fields.

Contribution

It reconstructs the AI domain's internal structure using co-occurrence networks and analyzes its interdisciplinary diffusion and collaboration dynamics.

Findings

Identified 15 distinct AI specialties with unique temporal patterns.

Mapped the disciplinary landscape of AI applications across science.

Found limited collaboration between AI developers and users, with few bridging researchers.

Abstract

In recent decades the set of knowledge, tools and practices, collectively referred to as "artificial intelligence" (AI), have become a mainstay of scientific research. Artificial intelligence techniques have not only developed enormously within their native areas of development (computer science, mathematics and statistics) but have also spread fast, in terms of application, to multiple areas of science and technology. In this paper we conduct a large scale analysis of artificial intelligence in science. The first question we address is the composition of what is commonly labeled AI, and how the various elements belonging to this domain are linked together. We reconstruct the internal structure of the AI ecosystem through the co-occurrence network of AI terms in publications' abstracts and title, and we propose to distinguish between 15 different specialities of AI, with different temporal patterns. Further, we investigate the spreading of AI outside its native disciplines. We reconstruct the temporal dynamics of the diffusion of AI production in the whole scientific ecosystem and we describe the disciplinary landscape of AI applications. Finally we take a further step analyzing the role of collaborations for the interdisciplinary spreading of AI techniques. While the study of science frequently emphasizes the openness of scientific communities, we show that there are rarely any collaborations between those scholars who primarily develop AI, and those who apply it. Only a small group of researchers is able to gradually establish a bridge between these communities.