What makes papers visible on social media? An analysis of various document characteristics

In this study we have investigated the relationship between different document characteristics and the number of Mendeley readership counts, tweets, Facebook posts, mentions in blogs and mainstream media for 1.3 million papers published in journals covered by the Web of Science (WoS). It aims to demonstrate that how factors affecting various social media-based indicators differ from those influencing citations and which document types are more popular across different platforms. Our results highlight the heterogeneous nature of altmetrics, which encompasses different types of uses and user groups engaging with research on social media. INTRODUCTION Five years after the introduction of the term (Priem, et al. 2010), altmetric indicators can be found on most of major publishers platforms, and are increasingly used in research evaluation (Wilsdon et al., 2015). Although some factors such as document age (Thelwall, et al, 2013), discipline (Haustein, Costas, & Larivière, 2015), topic (Costas, Zahedi, & Wouters, 2015) as well as countries (Alperin, 2015), have been shown to affect the various indicators, the processes which make a scientific paper visible on social and mainstream media are still not yet fully understood. Haustein et al. (2015) showed that factors which typically influence citations counts had a smaller or opposite effect on social and mainstream media mentions and that the usage pattern differed in particularly regarding document types. This study builds upon this work, taking into account a longer citation and social media window and expanding it by Mendeley readership counts. It addresses the following research questions: What is the effect of document characteristics on the number of Twitter, Facebook, blogs and mainstream media mentions as well as on Mendeley readership counts? Particularly, 1. How do these effects compare with that observed for citations? 2. How do these effects differ across document types? 1 This work was supported by the Alfred P. Sloan Foundation Grant #2014-3-25, Leiden University Fund (LUF) Grant # 4509/22-1-14\T, Vw and by funding from the DST-NRF Centre of Excellence in Scientometrics and STI Policy (South Africa). This work is licensed under a Creative Commons License: Attribution-NonCommercial-NoDerivatives 4.0 International.