By: Anne Sofie Talleruphuus, M.Sc. in computer science, and Majken B. E. Christensen, cand.scient. in physics
In September 2019 Anne Sofie Talleruphuus was an intern at Astronomicca under supervision of Majken B. E. Christensen. During these four weeks she worked with statistics and network analysis to investigate how the number of research publications in science communication has developed over time. This article sums up the work.
Summary
Some foundations and universities have increased their requirements for scientists’ outreach efforts and engagements with the public over the past years.
We find it relevant to investigate if research in science communication has also increased over time, as we would expect researchers to need more scientific knowledge on which outreach initiatives work to ensure an efficient use of their time and efforts.
In this study we examine a sample of research articles within the field of science communication. We investigate how the number of research articles has developed over time to evaluate if the amount of research in science outreach has increased as we see it in fields of natural science in general.
Our study is based on science communication articles published by SAGE journals and Journal of Science Communication (JCOM). The two journals are chosen, because they are the most frequent when we search Google for scientific words and phrases related to science communication.
In part 1 we investigate the development of research articles over time. Here we find that science communication publications in SAGE journals do not follow an increasing trend over time, but instead seem to be fluctuating from year to year. In JCOM we find a slightly increasing trend over the years.
In part 2 we investigate the connection between the publishing authors. Our motivation is to understand if science communication research is generally published by single authors or in collaborations. Here we find that authors in general do not publish together, but instead they tend to publish alone. A small group of authors published with 2-4 fellow authors.
Single-author publications could suggest that (some of) their findings are anecdotal and do therefore not serve as proper outreach material for researchers. However, this is purely speculative.
Part 1: Number of research articles over time
We want to know if the amount of research articles in science communication increases over time. With the increased requirements to the outreach efforts of scientists we claim that this calls for an increase in the amount of research in science communication. If scientists do not know which methods work, especially in today’s digital universe of social media, then how are they expected to perform well in their outreach?
We construct histograms of the annual number of research articles published for each of the two journals. We use webscraper.io to scrape a total of 1319 articles from SAGE and 571 from JCOM. The SAGE articles are filtered, and we include 891 research articles from 1979-2018 and exclude editorials, book reviews, obituaries, article commentaries, review articles, introductions, and the categories other and null. The JCOM articles are also filtered and we include 246 research articles from 2002-2018, and exclude editorials, book reviews, conference reviews, letters, comments, practice insights, essays, and focus articles.
We find variation in the annual number of published articles as seen in figure 1 below, and our hypothesis of an increasing trend is not supported by this data from SAGE journals. We wondered if this was because a majority of science communication articles were collected in the special editions, but excluding special editions (which all occur after 2010) from the analysis does not change the overall picture. SAGE did not publish science communication articles within the category “research articles” in 2003, which explains the gap.
JCOM, however, shows an increasing trend in published research articles after its start in 2002, as seen in figure 2 below. The big jump in 2016-2017 is caused by JCOM’s publication of special editions.
Part 2: Single or multiple authors
We use network analysis to investigate the relations between authors. Network analysis is a powerful tool that also work well for illustrative purposes.
Our network consists of nodes and edges. In our case a node represents an author and an edge represents a collaboration of two authors.
Figure 3: A simple network, with nodes (blue dots) representing authors and edges (black lines) collaborations.
A network of authors in all science communication research articles in SAGE journals provides information about how the community in this field works together. If a node does not have any edges, it means that the author has no collaborations (at all) in SAGE journals. If a node has one edge, the author has collaborated with one other author (in total).
The network of SAGE journals’ authors is shown in figure 4. The network is very disconnected and co-authoring seems to only happen among a small group of people.
We investigate this further by computing the degree distribution of the network. A degree is defined as the number of edges (collaborations) connected to a node (author). The degree distribution is shown in figure 5.
The diagram shows that approximately 15% of the 332 authors do not publish with other authors. More than 40% of the authors have never published with more than one author in SAGE journals, and the majority has published with less than 5 other authors (in total) in SAGE journals.
These findings could raise the concern that (some of) the publications can be an expression of a certain author’s opinions or personal experiences, i.e anecdotal. This is further supported by the fact that more than 50% of the authors only have contributed to 1 article in SAGE journals. These are our own speculations.
About the study
All information used in this study has been scraped from the two websites using the Google Chrome extension webscraper.io. SAGE journals was scraped on September 16, 2019 and JCOM on September 20, 2019. The two sitemaps used for scraping are available at this GitHub Gist. Following attributes were scraped for each article in the archives: {title, date, authors, category, university} (note: university was only available on SAGE journals).
Python 3.7 and the libraries numpy and networkx are used for the data analysis.
Code is available on GitHub. All coding was done by Anne Sofie Talleruphuus.
All graphs and illustrations are our own. The cover photo is a word cloud based on the article information used in the study.
This study is not published anywhere else.
