Ideological biases in social sharing of online information about climate change

shared readerships. For the first time, we track network structure over a 7-week period, includ- ing a disruptive media event (the announcement of the withdrawal of the USA from the Paris Agreement on climate change), to show the persistence of the observed polarisation and politi- cisation of media sharing related to climate change over time and with varying background levels of public interest. Social media data has been used to study several aspects of public opinion around climate change, including attitudes towards and engagement with climate change mitigation strategies [15], media framing of the leaked “Climategate” emails on YouTube [5], and the spreading of calls for collective action at the COP15 conference [16]. Networks are an intuitive representa- tion, and come with a host of analytical tools to understand the shape of online discussions around climate change; for example Elgesem et al. [17] explore the network of hyperlinks between blogs, while Williams et al. [18] study the structure of follower, retweet and mention networks on Twitter. In both of these studies, user communities manifest as densely intercon- nected clusters with similar characteristics. These communities are highly polarised, such that each community is well described by a single viewpoint, with few moderate voices. Similar pat- terns have also been observed for online political discourse (e.g. [19–24]). This pattern of opin- ion polarisation and segregation has important implications for opinion change and the

likelihood of global consensus [25]. Polarisation in online social media is most frequently stud- ied in the political sphere, especially for two-party political systems with an ideological split along a left-right axis. However, the phenomenon also extends to the competing opinions around climate change, which are often simplified as a debate between environmentalists (sup- porting the scientific consensus and promoting action) and sceptics (doubting or opposing the consensus and need for action), notwithstanding the diversity of viewpoints and representa- tions of this complex issue [26]. These previous network-based studies generally treat datasets as single snapshots, along with the implicit assumption that the phenomena under study varies slowly. The intervals studied range from months to years, but by choosing such a timescale it is possible to overlook the changes that social networks can experience in shorter periods. The pattern of polarisation in both political and climate change contexts is often associated with the existence of echo chambers in the social media ecosystem, whereby users choose to associate with people and news-media sources which conform to and reinforce their existing beliefs [18]. Echo chambers have been proposed to contribute to the spreading of misinforma- tion [27], political networks of environmental actors [28], exposure to political information on social media [21], and online content around climate change (e.g. [17, 18, 29]). In this work we focus on the structural phenomenon of echo chambers but other scholars have looked at how

they are linked to psychological processes such as confirmation bias (e.g. [30]). Previous studies of information-sharing around climate change have focused on the promi- nence of different sources. Newman [31] analysed the tweets and information sources shared alongside the release of the IPCC AR5 WG1 report, finding a focus on the public engagement with science, and a dominance of mainstream media sources. Segerberg and Bennett [16] examined the breakdown of different link sources used alongside calls to collective action at the COP15 conference. Kirilenko and Stepchenkova [32] studied the URLs shared on Twitter over the course of one year in five different languages, finding that by country, the US domi- nated total tweet counts, and a mix of traditional media, activist and sceptic sites were shared. Polarisation is a common observation in the climate change debate. Notable studies such as Dunlap et al. [3] have shown that the polarisation effect in climate change opinion has grown between 1997 and 2016. An impact of polarisation can be observed in the frames used to dis- cuss climate change, such as by Jang and Hart [33], who analysed the themes present in Twit- ter’s climate change debate across two years. They found differences in the terminology used by opposing groups, with Republican-leaning states in the US using global warming in PLOS ONE Ideological biases in social sharing of online information about climate change PLOS ONE | https://doi.org/10.1371/journal.pone.0250656 April 23, 2021 3 / 25