Climate change on Twitter: Implications for climate governance research

There is increasing public debate about the governance of climate change and its repercussions for nature and human livelihoods. In today's digitalized communication landscape, both public and private actors involved in climate change governance use social media to provide information and to interact with stakeholders and the broader public. This Focus Article discusses two main aspects of debates about climate change and climate governance on Twitter, which previous theories suggest to shape climate governance across domestic and global levels: non‐state climate action and public opinion formation on the social media. We see significant advancement in the environmental social sciences studying these two areas. Yet, we also see the need for a better understanding of how public and private actors in the climate governance complex interact on Twitter, and how these actors shape, and are shaped by, experiences, values, and positions. This understanding will help to advance climate governance theories. This article proceeds in three steps. We first discuss previous social media research on non‐state climate action and public opinion formation related to climate change and its governance. Then we sketch avenues for future research, elaborating how Twitter data might be used to investigate how non‐state climate action and public opinion formation on social media are linked to and influence climate governance. We conclude by making the case for drawing together Twitter data and climate governance research into a more coherent research agenda.

At both the domestic and international levels, the importance of social media debates about climate change and its governance is growing.Members of the public in different countries have increasing awareness and experience of the influences that climate risks have on human livelihoods (Ara Begum et al., 2022).As a result, these climate risks are being extensively debated and the governance of them, by both domestic authorities and international organizations such as the European Union, the African Union and the United Nations, is being increasingly contested.Social media plays crucial roles in this debate and contestation.Platforms like Twitter can function as a discussion forum (Pearce et al., 2019), a soft power tool (e.g., Mavrodieva et al., 2019), and a vehicle for transnational advocacy and climate action (e.g., Segerberg & Bennett, 2011).In turn, communication about climate change is nowadays part and parcel of policymakers' engagement with climate change.These policymakers often use social media, and Twitter in particular, to frame the debate and to showcase particular achievements (Ecker-Ehrhardt, 2020;Pearce et al., 2019).
As a result, Twitter is an important part of the context in which many actors in the global climate change regime develop political views, contest climate norms, but also spread misinformation and disinformation.This has led to increasing scholarly interest in the ways in which members of the public in different countries use social media to engage with the topic of climate change and with climate policy solutions.In this Focus Article, we discuss two central aspects of these discussions: non-state climate action and public opinion on climate change and its governance.Theories of global governance lead us to expect that non-state climate action that uses strategies such as public opinion mobilization or norm diffusion, be it online or offline, or a combination thereof, influences climate governance (Allan & Hadden, 2017;Keck & Sikkink, 1998).In turn, previous social media research commonly assumes that non-state climate action and public opinion on social media shape climate governance through norm diffusion, opinion leadership, and public opinion formation.However, these mechanisms are usually not systematically studied.
To date, social media and climate governance research have made only sporadic contact, despite a strong potential for highly productive cross-fertilization.Important research questions remain unanswered: How is climate change and its governance debated on social media?Why are some climate change and related policy issues debated on social media while others are not?How do state and non-state actors, as well as individual transnational actors, communicate about climate change and governance on social media, and what effects does this have on governance?How do policymakers understand and are influenced by social media debates and how do they sort out what is valuable when forming their opinions and deciding about climate policy solutions?In the widest sense, how are social media debates and climate governance related?
This Focus Article uniquely links climate governance to a relatively new type of media.Social media is often defined in differentiation from traditional media, where traditional media is single directional one-to-many communication, while social media is multi directional many-to-many communication and thus fundamentally defined by their sociality (Sloan & Quan-Haase, 2022, ch. 1).Examples of social media platforms are Twitter, Facebook, Youtube or Weibo.While previous research has provided invaluable knowledge of non-state climate action and public opinion formation on social media, we need a better understanding of the knock-on question of how public and private actors interact on social media, and how such interactions shape climate governance.In this article, we argue that future research could provide such an understanding by using one of the most frequently used social media platforms for political debates: Twitter.

| THE EFFECTS OF SOCIAL MEDIA ON CLIMATE GOVERNANCE
Non-state climate action and public opinion formation processes on Twitter have emerged as focal points in previous research.We discuss both aspects in turn and pay close attention to how they have been linked to climate governance.Climate governance refers to the structures, processes, and actions through which private, public, and hybrid actors interact to address societal goals related to climate change (IPCC, 2022(IPCC, , p. 2910)).

| Non-state climate action
Many studies of social media and climate change focus on climate action in terms of non-state actor advocacy, strategies, and impact on climate governance on Twitter (see Segerberg, 2017, for an overview).Non-state actors are central in climate politics and governance, where they work to mobilize for and influence change in climate governance.This section foregrounds the examples of transnational activists and scientists, as they are central voices in climate change debates on Twitter.Such actors' social media strategies, often in combination with protest activities and participation at climate governance events, have been shown to be influential at both the domestic level and the global level.However, there is much debate about whether such strategies have a positive or negative influence on climate governance (Bakaki, 2022).
For climate governance at large, non-state actors' problem or issue framing has been shown to matter.How problem framing works in debates among experts, non-governmental organizations (NGOs), and transnational advocacy networks, among others, can impact climate policy (e.g., Allan & Hadden, 2017;Keck & Sikkink, 1998).Moreover, corporate advocacy activities can skew framings in climate governance to reflect special interests rather than the public good (e.g., Hopke, 2021;Nasiritousi, 2017).
The use of Twitter allows non-state actors to mobilize support in a cost-effective way.This has for example been shown to be the case during events such as the United Nations' climate change conferences (COPs) (Segerberg & Bennett, 2011).Studies of the Fridays for Future movement have shown that Twitter has enabled a forceful mobilization (Boulianne et al., 2018), even by marginalized populations such as Indigenous Peoples (Hopke & Paris, 2022).The COVID-19 pandemic does not appear to have affected the extent of mobilization on Twitter, rather, the pandemic has led activists to shift their communication from mobilization to thematic norm diffusion, where norms are crucial guiding principles for policymakers (Haßler et al., 2021).
There is also an increasing interaction between state and non-state actors about climate change on social media.For example, the content and tone of communication by scientists in the United States change when they interact with politicians on Twitter and become more strategic (Hodges & Stocking, 2016;Walter et al., 2019).Isolated studies from intergovernmental organizations suggest that such organizations' social media engagement follows a similar pattern (Ecker-Ehrhardt, 2020).
Recent years have seen increasing scholarly attention to the issue of misinformation and disinformation on Twitter.One source of misinformation is fossil fuel industry trade groups that promote their own framings of climate change (Hopke & Hestres, 2018).Those framings can misinform social media users when they frame information in a way that omits information that is essential to understand decarbonization (Hopke, 2021).Related, the discourses and techniques used to promote denialist arguments have been studied (Jacques & Connolly Knox, 2016).
Another question concerns who leads discussions on Twitter and is most likely to influence the opinions of policymakers.We currently know little about the patterns and sources of opinion leadership.Research has found that it can be challenging for non-state actors to use online platforms to influence opinions because they often lack the resources to form a cohesive communication strategy (Takahashi et al., 2015).Those most successful in their opinion leadership are non-state actors from the Global North and Oceania (Vu et al., 2020).However, social movement representatives' Twitter engagement is consequential for offline protest activities despite their social media communication being highly diverse (Boulianne et al., 2018).
In sum, past research into non-state actor communication about climate change has focused on how such actors use Twitter platforms to pursue advocacy and to strategically diffuse norms, as well as studying when and why they become opinion leaders.However, most of this research is not framed in terms of a contribution to climate governance.

| Public opinion
Public opinion is central for climate governance because office-seeking policymakers have incentives to introduce policies that seem favorable to domestic audiences, thereby shaping the political agenda, decision-making, and outcomes of governance.This makes it important to understand the patterns and sources of attitudes toward climate change and climate policy.Moreover, previous research has focused on public support for climate policy and international climate action (see Fairbrother, 2022, for an overview).There are ever-expanding fields of research studying public opinion on Twitter (e.g., Cody et al., 2015;Jang & Hart, 2015;Kirilenko & Stepchenkova, 2014).
The emotionality, ideological biases, and polarization of public opinion on Twitter have attracted scholarly interest (e.g., Cann et al., 2021;Jost et al., 2019).The claim that debates focused on climate change and climate policy tend to become "echo-chambers" (Sunstein, 2001), whereby people communicate mostly with like-minded peers (Pearce et al., 2014), has seen increased attention in recent years.Yet, there is also research problematizing this claim, as the evidence is inconclusive and the framing might distract from the polarization of contemporary societies at large (Bruns, 2021).On Twitter, between 2014 and 2021, the polarization around climate change has increased (Falkenberg et al., 2022).This tendency may be reinforced when people perceive a specific opinion to be in line with a social consensus on climate change.Indeed, opinion leaders such as scientists can impact people's attitudes toward climate change and climate policy solutions by communicating a supposed scientific consensus (Jang & Hart, 2015;Lewandowsky et al., 2019;Williams et al., 2015).
Another prominent research problem studied in the literature concerns who participate in social media debates about climate change.For instance, Kirilenko and Stepchenkova (2014) used a global Twitter sample with five languages to identify the countries in which climate change debates are most prominent.They found debates are most prominent in the United States and the United Kingdom, followed by Canada and Australia.It is also known that there is self-censoring on social media as a consequence of political repression in authoritarian countries.This, in turn, may affect whose opinions are expressed on social media (e.g., Liu & Zhao, 2017).
Similarly, there is inequality in social media participation within countries.It is well known that Twitter data are not representative of broader public debate or public opinion (Mellon & Prosser, 2017).This is because those registered on Twitter tend to be interested in politics and to be part of the working-age population (McCormick et al., 2017).Under-represented groups on Twitter might instead use other social media, such as Facebook, TikTok, or Instagram.Limitations regarding the non-representativeness of social media data and the potential for (self-) censorship require careful reflection when discussing the generalizability of research using social media data.It is an open question whether and to what extent data from social media platforms other than Twitter have similar limitations.
In sum, research into public opinion on climate change and climate policy typically assumes that it is relevant for policymakers mostly in the United States and Europe to know about public opinion expressed on Twitter.Yet, we have not yet seen studies that systematically connect public opinion on Twitter to climate governance.Studies have also insufficiently considered how social media is shaped by policymakers (Box 1), or how these policymakers view, and have their own opinions shaped by Twitter.

| THE EFFECTS OF SOCIAL MEDIA ON CLIMATE GOVERNANCE
While previous scholarship usually makes assumptions about the mechanisms through which Twitter debates may influence climate governance, these mechanisms are seldom explicitly studied.There are three main mechanisms described in this existing literature: norm diffusion, opinion leadership, and public (including citizen and elite) opinion formation.In this section, we suggest ways in which research about these three mechanisms could be linked to research about climate governance to foster a more coherent research agenda on the role of Twitter debates in climate governance.
We focus this discussion on Twitter data because much social media research on climate change draws on these data.To be sure, climate-related debates between state and non-state actors do not take place online only.Instead, these debates might also take place at climate conferences, (inter-)governmental meetings, and other global venues and social

BOX 1 Twitter communication by climate change policymakers
Twitter communication of climate change policymakers can, via public attitudes and behavior, have feedback effects on climate governance.For instance, international policymakers are becoming increasingly adept at using this platform (Ecker-Ehrhardt, 2020) and Twitter debates are shaped by the outputs of international organizations such as the Intergovernmental Panel on Climate Change (IPCC) (Pearce et al., 2014).However, we know little about the ways in which climate change policymakers engage on social media and what effects this engagement has.Research into global governance has pointed to the significance of individual policymakers' norm entrepreneurship as a factor in shaping governance-especially in global climate governance, where actors are typically confronted with epistemic ambiguity and knowledge uncertainty about global-scale climate risks (Lesnikowski et al., 2019).How policymakers use social media to engage with the issue of climate change, and how their opinions are being shaped by social media, are therefore important to know.settings more broadly.Moreover, Twitter debates might not be representative of public debates at large, as discussed above.
The Twitter platform consists of a website where users, on their profiles, can make posts called "tweets" and also share and comment on other people's posts.A tweet is a text-based message that is limited to a maximum of 240 characters.Tweets usually constitute the unit of analysis when conducting Twitter data analysis.After retrieving tweets from Twitter, researchers need to prepare or pre-process this data for analysis.The analysis of the data can be either descriptive or explanatory (cf.Sloan & Quan-Haase, 2022).In this section, we will draw on the latest methodological advancements in natural language processing methods currently being used to analyze large sets or corpora of texts, to exemplify how to link social media and climate governance research.

| Norm diffusion
Previous literature on climate policy issues such as green-washing, sustainability, and corporate social responsibility underlines that the content of norms and the dynamics of norm diffusion are important to understand (e.g., Kim & Lyon, 2011).However, those norms that have not made it onto the global agenda are rarely studied (Legro, 1997).Using Twitter data, it would be possible for scholars to trace the evolution of the most salient climate norms among non-state actors and detect their effects-even though they might not shape the political agenda (cf., Jang & Hart, 2015;Kirilenko & Stepchenkova, 2014).
To this end, Twitter data can be analyzed descriptively.This then enables researchers to detect major topics in the public debate about climate change, isolate climate-related norms, and map patterns across global actors, political issues, and over time.For example, pattern detection (e.g., by using dictionary-based approaches to text classification or topic modeling) and frequency analysis (e.g., by means of visualizations such as frequency plots) can be useful to identify how state and non-state actors across levels of government engage in Twitter debates about climate change.Dictionary-based approaches involve researchers forming lists of keywords-so-called "dictionaries"-related to a specific topic.These word lists are then checked against a text corpus to capture the extent to which the topic appears in the corpus.Topic modeling is another method that can be used to identify climate norms.This is an unsupervised method of natural language processing that is designed to extract abstract themes-so-called "latent topics"-from a given corpus (cf.Lesnikowski et al., 2019).
These methods have proven their potential to provide good insights into social media communication.Cody et al. (2015) made use of a dictionary approach in their aforementioned study.Kirilenko and Stepchenkova (2014) combined a dictionary-based approach with frequency analysis methods to identify the countries in which climate change debates are most prominent.Topic modeling can provide an effective method for the study of norm diffusion.It provides a useful way to discover those norms in the public debate which researchers might not have known about a priori.Climate change is a complex topic, and topic modeling enables researchers to draw out specific themes that appear in a text corpus without needing to inform the model about these themes beforehand (Lesnikowski et al., 2019).This can be done in an automated (or "unsupervised") way that does not require that researchers inform the statistical model with prior information or specify details such as the exact themes expected.Topic modeling is thus a useful way to identify climate-related themes in debates (Dahal et al., 2019;Falkenberg et al., 2022).
Taken together, such analyses of norm diffusion could be usefully tied to the insights about public opinion and nonstate engagement with climate change we discussed earlier in this article, enhancing understanding of how interactions on Twitter, through norm diffusion, shape climate governance.

| Opinion leadership
In research on social media and climate change, a central topic for study is how and under what conditions non-state actors use social media to influence climate-related debates and policy (Pearce et al., 2014;Vu et al., 2020).One direction future research could take would be to investigate the interactions occurring on Twitter between actors within and across different social and policy communities (see also Hodges & Stocking, 2016).Twitter allows for a range of different kinds of interactions between users.They can relate to one another (by using the @ symbol which makes a post target another specific account), join in with a discussion about a particular topic (by using the # symbol which creates tags to group posts together), and engage in a general discussion of topics (by simply mentioning a word, such as "climate," in a tweet).Two additional functions ("likes" and "shares") allow researchers to capture interactions between users that are centered on specific tweets.
To use such data to analyze opinion leadership, network analysis could be particularly helpful.Network analysis refers to a set of techniques that uncover social relations between actors, or structures linking actors together, on the basis of the recurrence of links (or connections) between these actors.Network analysis can be used to establish who acts as opinion leaders, and who are followers on Twitter, identifying effects of public awareness of and opinion on climate policy issues (Tindall et al., 2021).For example, Vu et al. (2020) used network analysis to describe when NGOs from the Global North assume a leadership role in their networks, setting the political agenda regarding climate change.Pearce et al. (2014) used network analysis to examine Twitter debates surrounding a report of the IPCC, identifying a number of discussion communities that frequently interacted with each other in these debates.
Moreover, network analysis can be a tool to identify the potential risks of climate change discourses on Twitter.Falkenberg et al. (2022) have documented an increased polarization around climate change on the social media platform around the COP meetings, and have emphasized that this polarization may risk political deadlock if it fuels antagonism to climate action (see also Williams et al., 2015).An additional risk is the spread of misinformation and disinformation through opinion leaders (e.g., Coan et al., 2021;Hopke, 2021), which is a particular challenge for policymakers.
In sum, studying opinion leadership could be useful for understanding how public opinion and non-state climate action on Twitter shape climate governance.Opinion leadership might play an important role in shaping policy-maker perceptions of public debates.

| Public opinion formation
Twitter data can also be employed in the study of the formation of opinions about climate change and climate policy among citizens and elites.It can be used to examine how political communication or endorsement by individuals influences the opinions of citizens regarding climate change and climate policy (e.g., Park, 2020;Kirilenko & Stepchenkova, 2014).Future research about public opinion could build on such analyses using Twitter data, examining whether framings by specific actors of particular climate risks, such as climate-induced displacement, can make social media content spread and influence citizen and elite opinion (cf.Vu et al., 2021).
Such a research agenda could also provide evidence of the potential benefits and risks that specific framings of climate change, through influences on the broader public, can imply for climate governance (Lewandowsky et al., 2019;van der Linden et al., 2017).Specifically, sentiment analysis provides a promising method.Sentiment analysis is a text analysis method that is used to systematically identify the sentiment of texts from a corpus.In climate governance research, sentiment analysis has been used to detect positive and negative sentiments in public discussions about climate change (Cody et al., 2015;Dahal et al., 2019;Jost et al., 2019).In political communication studies, it has been shown that sentiment drives attention and information dissemination on Twitter and Facebook (Kraft et al., 2020).However, the question of how positive and negative sentiments in the climate change regime are shaped by interactions on Twitter remains unaddressed.
In conclusion, the study of citizen and elite opinion on climate change and climate policy is a promising avenue for future research in systematic investigations of how engagement in debates on Twitter impacts climate governance.This research avenue holds the promise of innovative theorizing and new empirical knowledge on the questions raised in the introduction to this article.

| CONCLUSION
This Focus Article has discussed non-state action and public opinion formation on Twitter about climate change and its governance.Twitter is an important communication tool and arena for political discussions about climate change, and Twitter data is often used in the study of non-state climate action and public opinion on social media.However, these studies are not typically framed in terms of their contributions to the climate governance literature, which implies that they are not widely received in this climate governance literature.
Thus, there is tremendous potential to develop a better understanding of how actors in the climate governance complex engage on Twitter, interact with the greater public, and are themselves shaped by such interactions.This greater understanding of climate debates on Twitter can inform climate governance research and advance theory on how social media, through norm diffusion, opinion leadership, and citizen and elite opinion formation, impacts climate governance.While this article has discussed some of the ways in which future research can achieve this by building on Twitter data, future research could usefully broaden the focus to other social media, such as Youtube or Weibo.Our suggestions may apply to other social media data in the sense of short texts, but platforms differ in their usage, user profiles, and data types, implying that this requires further study.
While climate governance research can gain much from integrating social media research, it is important to note that social media research stems from a research tradition that relies on different epistemologies, theories, and methodologies.Advancing on theories of climate governance typically requires theoretical reasoning ex-ante.In contrast, social media research techniques originate from inductive research traditions and need to be creatively adapted in the study of climate governance (cf.Lesnikowski et al., 2019).
All told, a better understanding of how climate governance and public debates on Twitter are related can also be practically useful.It can help actors identify leverage points that can be used to improve the domestic and international governance of climate change.Developing a better understanding of how social media relates to climate governance can contribute greatly to the research field and play a key part in improving this governance in the future.