Ideological biases in social sharing of online information about climate change

preference to climate change, and often using hoax frames to cast doubt on the scientific con- sensus. O’Neill et al. [34] found that this trend extended to the media coverage of the publica- tion of the IPCC AR5 working group reports. By studying the frames used in newspaper and television broadcasts, they found clear preferences for different frames amongst the various news organisations. Some work has looked at countering the growing levels of polarisation, including Zhang et al. [7] who study the effect of clarifying messages on accuracy over the per- ceived levels of consensus among climate scientists. Among their experimental group, expo- sure to the clarifying message lead to more uniform accuracy around the scientific consensus through greater impact in the areas of lower baseline belief. The effects of polarisation do not always manifest equally on each end of the opinion spec- trum. Schuldt et al. [35] compared the usage of the terms climate change and global warming across the websites of a series of think tanks. Right-wing think tanks were more likely to use global warming, with the opposite trend observed in left-wing think tanks. These findings of content differences from polarised sources extends beyond the climate change debate. Further analysis of climate sceptic organisations and their funding sources by Farrell [36] found shared sources of funding across many of them. Beyond the topic of climate change, Freelon et al. [37] present an overview of how the different online activism strategies of left- and right-wing

groups manifest different types of content and audiences. They highlight a key difference in the perceived strategies of the different groups. Left-wing groups target “hashtag activism” leading to social promotion of movements whereas right-wing groups engage much more readily with sympathetic media organisations to promote their messages and goals. Freelon et al. also recognise a similar trend in group coherence where right-wing groups are tighter when compared to left-wing groups formed from a loose coalition of multiple issue-led groups. Considering this coordinated funding of multiple groups and engagement with media organi- sations, it is to be expected that climate sceptic messaging is likely to be more consistent than competing environmentalist information. Along with climate change, studies of information-sharing on social media have embraced a diversity of different topics; Starbird [38] studied the spread of misinformation around mass shooting events in the United States, finding a cluster of alternative news sites separate from mainstream media sources. Arif et al. [39] and Del Vicario et al. [27] study the spreading behaviour of rumours and misinformation on Twitter and Facebook respectively and classify observed trends. Schmidt et al. [40] analyse a network of news-related pages on Facebook, where two pages are connected if they have posts that are liked or commented on by the same user. Their cluster analysis reveals a highly polarised structure, as seen in a number of other

contexts (e.g. [19]). Williams et al. [41] also study the community structure of political news- sharing via Twitter, finding communities characterised by both geographical and political fac- tors. Weaver et al. [13] examined information-sharing on Twitter during the UK General Elec- tion in 2015, showing strong community structure explained by ideological, geographical and topical preferences. Each of these studies should be considered in the context of the typical sharer and the information they are exposed to. Not all users on social media are exposed to the same information, and typically they are exposed to information in which they have already shown an interest through their own decisions of which users to follow, along with algorithmic filtering effects. Sharing information requires action on the user’s part and as such those who choose to share are likely part of a more highly invested subset of the users exposed to a certain piece of information. The rest of the paper proceeds as follows. Section Methods details the method of data collec- tion and preparation, along with the techniques used to construct and analyse the informa- tion-sharing networks. Our main results follow in Section Results, including network analysis and characterisation of community structures, and finally Section Discussion provides a 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 4 / 25