How does collaboration affect researchers’ positions in co-authorship networks?

Highlights

• We want to validate the generalized friendship paradox from researcher's academic level.

• From the angle of sociability of researchers, we study the benefits of collaboration on enhancing researchers’ social circle.

• We evaluate the scope and dynamic changing of researcher's ego-network with some network centrality metrics.

• According to all these analysis, we explore the relation between collaboration and researchers’ influence in computer science. Experimental results show that collaboration can help researchers improve their influence to some extent.

Abstract

Collaboration usually has a positive effect on researchers’ productivity: researchers have become increasingly collaborative, according to recent studies. Numerous studies have focused on enhancing research collaboration by recommendation technology and measuring the influence of researchers. However, few studies have investigated the effect of collaboration on the position of a researcher in the research social network. In this paper, we explore the relationships between collaboration and influence by social analytical methods, which are pertinent to analyzing the network structure and individual traits. We evaluate three aspects of the researchers’ influence: friendship paradox validation, social circle, and structure of a researcher's ego network. Furthermore, the ”six degrees of Bacon number” theory, generalized friendship paradox, and triadic closure theory are introduced to support our analysis. Experimental results show that collaboration can help researchers increase their influence to some extent.

Introduction

We live in a world of networks with numerous nodes and connections. Social network analysis can help us understand some social phenomena and behaviors (Schoenebeck, 2013; Zhong, Fan, Zhu, & Yang, 2013), and helps to reveal and understand the network structure and features of individual nodes (Staiano et al., 2012). For example, analysis of online social networks such as Facebook and Twitter shows users’ personality and a tendency to associate with similar users (Liu, Venkatanathan, Goncalves, Karapanos, & Kostakos, 2014; Weitzel, Quaresma, & de Oliveira, 2012). Nowadays, the grand challenges that humans confront include rapidly changing human society and population structure, global crises, and increasing rates of crime. To solve these problems, it is necessary and meaningful for social scientists to analyze the structure of society and patterns based on massive amounts of data (Conte et al., 2012). For example, Liu et al. (Liu et al. 2018) studied the detailed structure of citation networks, in terms of the evolution of topics in artificial intelligence (AI), to improve the comprehension of the development of human society. Computational social science aims to model social problems quantitatively, understand social systems (Conte et al., 2012), and explore some mechanisms behind these phenomena. As a result, it promotes more studies of application services, such as recommender systems (Hsiao, Kulesza, & Hero, 2014; Kong, Mao, Wang, Liu, & Xu, 2018; Rafailidis & Crestani, 2016).

Collaboration has been a vitally important behavioral phenomenon for social scientists to analyze social patterns, whether in education (Barr & Gunawardena, 2012), development teams (Ghobadi, 2015), or research (Coccia & Wang, 2016). Social analytical studies of co-authorship networks give a quantitative description of collaboration and may be pertinent for mining the essence of co-authorship networks ([Kong et al., 2016], [Kong et al., 2017], [Wang et al., 2017]; Xia, Wang, Bekele, & Liu, 2017). Collaboration is a very important aspect of research activity. In particular, persistent academic collaborations promote the progress and development of the whole academia. Studies of research collaboration reveal the facilitating relationships between research achievements and researchers’ cooperation (Amjad et al., 2017; Mavin, 2015; Zhang, Bu, & Ding, 2016). A two-stage least squares analysis proposed by Lee et al. (Lee & Bozeman, 2005) confirmed that collaboration has a positive effect on researchers’ productivity. Triadic closure can be applied to find more collaborations and understand transitivity in a co-authorship network. There is another social analytical theory, the friendship paradox, which states that one's friends have more friends than oneself on average. The friendship paradox has been confirmed to hold for more than 98% of Twitter users (Hodas, Kooti, & Lerman, 2013). The friendship paradox can be extended to some research characteristics, in which case it is called the generalized friendship paradox. One's coauthors are likely to be more prominent than oneself, on average, with more publications, more citations, and more collaborators ([Eom and Jo, 2014], [Grund, 2014]). We have noticed that a small amount of elite researchers, who become influential in their social circles, can break the law of the friendship paradox. This raises the question, how does collaboration benefit researchers in co-authorship networks? In this work, we will analyze co-authorship networks by social analytical methods to investigate this phenomenon.

The goal of this paper is to investigate what an influential researcher looks like in a co-authorship network. An influential researcher usually performs prolifically and has a higher rank in his/her social circle, attracting more people to collaborate (Zhang, Bu, Ding, & Xu, 2017; Zhang et al., 2016). His/her social distance from famous researchers is shorter than that of others. In this work, we mainly analyze the influence that collaboration can have on research, based on the DBLP dataset. We focus on three aspects. First, we explore the friendship paradox based on the researchers’ rank. Second, from the viewpoint of the sociability of researchers, we study the benefits of collaboration on enhancing researchers’ social circles. Finally, we introduce the Bacon number (Backstrom, Boldi, Rosa, Ugander, & Vigna, 2012), which measures the distance from one actor to Kevin Bacon. We can measure the social distance between an arbitrary researcher and a famous researcher, and explore the importance of the social circle of a given scholar. We evaluate the scope and dynamic change of the Bacon Number in a co-authorship network. We also consider two network centrality metrics, i.e. clustering coefficient and network density to measure the importance of scholars in the network. According to this analysis, we explore the effect of collaboration on the position of a researcher in a co-authorship network.