“YOU CAN'T SIT WITH US”: WHAT TOPIC MODELING CAN REVEAL ABOUT PALEONTOLOGY ON TWITTER

The scientific community often relies on social media to connect and converse about topics of interest. To date, scarce research examines how members of various scientific communities use social media and the topics which generate conversations within different social media niches. On Twitter, the paleontological community is made up of a diverse group of participants including scientists from various fields, museums and their representatives, amateur fossil collectors, paleontological artists, and commercial retailers. The participants in this community have varying interests and often use different types of messaging strategies to connect with others.

This longitudinal study used social network analysis and text mining to better understand how people participate in paleontology on Twitter. We used Netlytic, a network extraction software application, to pull tweets from the public Twitter search API every 15 minutes over a one-year period (July 2017 – August 2018) for the social network of an NSF-funded initiative, the FOSSIL Project (NSF-DRL 1322725), which focuses on uniting, developing, and growing the paleontology community. Sampling resulted in a corpus of 7,753 connections which were subjected to social network analysis and text mining using Latent Dirichlet Allocation (LDA). Participants (n = 1,369) were classified based upon a taxonomy for how they self-identify with paleontology (Lundgren, Crippen, Bex, 2018). Topics that reached diverse members and spanned the entire network included: opportunities to contribute to paleontology regardless of status and contributions by amateurs. One example of a topic in which we saw insular conversations and minimal bridging between diverse members occurred when paleontologists communicated about women in paleontology. Our results illustrate the diversity of the paleontological community, including how certain topics and messaging strategies reach different audiences within a single niche. Topic modeling provides a strategy for describing a network’s content, which can lead to the development of approaches to increase discourse related to science.