Can 280 Characters Speak for Researchers? Leveraging Twitter Data for Unobtrusive Measurement of Academics’ Well-Being

Current approaches for researcher identification on Twitter prove to be effective, but rely on external data sources. This dependency can be a challenge to their sustainability. Here, we report a chain-referral sampling algorithm that uses solely data from the Twitter API. Researchers are identified by crawling the mentions network of a seed sample of verified researchers. We address the two research questions of validity (RQ1) and representativity (RQ2) of the Twitter accounts identified by the algorithm. To answer the first research question, a precision-recall analysis was performed, while to answer the second research question, the distribution of gender, location, and subdiscipline criteria on Twitter was compared to that of publishing authors using the Chi-square test and Fisher's exact test. The results suggest our approach as a solid alternative for the case of missing external data sources. Moreover, our study provides further evidence that Twitter-active researchers should not be regarded as representative of the whole research community.