Scientific research collaboration in Artificial Intelligence: global trends and citations at the institution level

In order to gain a deeper understanding of collaboration and the relationship between collaboration patterns and citations in global Artificial Intelligence (AI) research, this present paper defines institution types and collaboration patterns from a new perspective. According to a variation of H-index, it classifies institutions into two types: Main institutions and Normal institutions. Based on institution types of the first and remaining institutions in a paper, it divides collaboration publications into six parts: M, M&M, M&N, N, N&M and N&N. In this study, all publications were collected from papers listed in Web of science from 1997 to 2017, published in the field of AI. According the number of units in a paper, results show that five or more authors have a great chance to be the primary pattern in Al field in the future; single-institution papers are the primary pattern but decreasing sharply during a long time; single-country papers keep playing a dominant role in past almost 20 years. According to different collaboration types, results show that five or more author publications are the primary form in M&M, M&N and N&M types, while three-author papers in N & N; Domestic two-institution papers in M&N and N&N are obviously more than that in M&M and N&M types; Single-country papers account for a large share in M&N, N&N and N&M, while two-country papers are more than single-country papers and become the most important part since 2010 in M&M. According to the relationship between collaboration types and citations, results show that the number of Main institutions has a positive relationship with the citation values, while the number of Normal institutions has a little negative influence on N&N type.