Framing of Civil Society Organizations’ Communication on Facebook: Examining the Impact on Audience Engagement in Norway

1.1 Dual-Process Framing

Framing involves highlighting certain aspects of an issue and downplaying others, in order to influence how people perceive the issue and the proposed solutions (Boräng et al. 2014). Since policy issues have multiple potential dimensions, framing – selecting and emphasizing particular aspects of an issue – is an important tool for policy actors, including civil society organizations (Baumgartner and Mahoney 2008; Daviter 2009; Entman 1993; Klüver et al. 2015). Frames can be thought of as ‘schemata of interpretation’ which help people organize what they perceive (Goffman 1974). According to Entman (1993), frames ‘highlight some bits of information about an item that is the subject of communication, thereby elevating them in salience’ (Entman 1993, p. 53). If scholars agree, following Daviter (2009), that frames are about ‘what actors perceive to be at stake in an issue thus depends on what facet or dimension dominates the actor’s perception at a given time’ (Daviter 2009, p. 1118), they have differentiated frames into several categories, including generic and issue specific frames (Boräng and Naurin 2015; De Bruycker 2017; Eising et al. 2015) as well as emphasis and equivalence frames – such as opportunities versus risks, gains (benefits) versus losses (costs), and positive consequences versus negative consequences (De Bruycker 2017). In any cases, frames are strategic tools that civil society organizations rely on in order to achieve their political goals, both when it comes to influencing policy outcomes (Boräng and Naurin 2015; Bunea and Ibenskas 2015; Klüver, Mahoney, and Opper 2015; Rasch 2018) and to influencing public opinion (De Bruycker and Beyers 2019; Druckman 2004; DÜR 2019; Tresch and Fischer 2014).

If traditional framing studies have focused on the cognitive dimension of framing, social-psychologists have increasingly advanced dual-process theories, which describe how affective and cognitive systems, termed “hot” and “cool” (Bargh 1994), work together or sometimes conflict. Automatic processes, which have been part of brain function for millions of years, are fast, effortless, and efficient, closely linked to perception and causing cognitive and behavioral changes without conscious reflection. In contrast, controlled processes, a more recent human development due to language and the capacity for private reflection, occur slowly and with conscious awareness, and are more distinct from perception. Bargh’s research shows that many morally relevant behaviors can be triggered by automatic processes in response to subtle environmental cues. From such a viewpoint, moral reasoning is a rapid intuitive process (Haidt 2008) in which moral intuitions can be triggered by different issues and may impact on attitudes and behaviors (Graham et al. 2013). As a consequence, the frames emphasized in communicating issues related to climate change can be seen as influencing attitudes and behaviors through the triggering of moral intuitions (automatic processes) and through reflected rational arguments (controlled processes). In order to take into account for the duality of processes, topical framing is differentiated from moral framing in the rest of this contribution.

1.2 Topical Framing and Engagement with Climate Change on Social Media

Framing has been shown to influence public perceptions and attitudes towards climate change in multiple studies (Badullovich et al. 2020). According to a recent survey experiment in several countries reported by Dasandi et al. (2022), different frames significantly impact public support for climate policies connected to climate change. Overall, the results of the experiment revealed that a positive frame, with health and environmental themes, on a global and immediate scale, led to heightened public support for climate policies in the majority of the observed countries excluding India. The effects of these frames on the public’s perception of the scientific consensus on climate change were evaluated in a separate national experiment in the US (Linden et al. 2018). Furthermore, experiments examining public reactions towards different climate change frames revealed a varied response depending on the segment of the audience (Maibach et al. 2010). However, studies found that the effects of framing are contingent on political ideologies and partisanship (Singh and Swanson 2017; Weber 2016).

A recent review of the literature related to framing in climate change communication (Guenther et al. 2023), summarizing previous reviews of literature on communication framing (Borah 2011; D’Angelo et al. 2019; Matthes 2009) as well as on climate change media communication (Schäfer and Schlichting 2014), emphasizes the fact that most studies focus on text-based frames, are primarily based in sociological tradition of framing (Gamson and Modigliani 1989; Goffman 1974) – as opposed to the psychological tradition of framing (Scheufele and Iyengar 2014) – and are more concerned with thematic frames than generic frames. Guenther et al. (2023) advance the idea of “frame locations” – distinguishing four locations: the frames of communicators and journalists, frame content, audience frames, and framing effects – and propose to use “framing” as a bridging concept for organizing the literature related to climate change framing. Guenther et al. (2023) found that in frame content studies, thematic frames dominated over generic frames and that the two most commonly identified generic frames were “Climate action” frame (including hope, action, mitigation, adaptation, solution, efficiency, progress, innovation, and development frames) and the “Harmful impacts” frame (including ecological, meteorological, apocalypse, disaster, risk, threat, and consequences, effects frames).

Additionally, research suggests that the way in which an issue is framed can have a significant impact on how social media users engage with and share news. Wahlström et al. (2013) have found that climate activists use “global justice” or “system change” in their framing while Grotenbreg and Van Buuren (2017) have shown that the use of diagnostic, prognostic, or motivational frames in government policy depends on policy goals. Vu et al. (2020) have examined the framing of climate change by global climate nonprofits on Facebook, analyzing content from 289 NGOs in 18 countries. They found that, among the three protest frames (diagnostic, prognostic, or motivational frames), the diagnostic protest frame was the most popular, and that messages tended to focus on the present impacts of climate change rather than efficacy. Action was the most frequently discussed aspect of climate change, while efficacy was the least common. More generally, news sharing on social media appears to be influenced by generic frames (Valenzuela et al. 2017) such as conflict frames (stories that pit one group or idea against another), economic frames (related to users’ personal financial well-being), human interest frames (related to experiences and emotions of individuals), and morality frames (highlighting issues of fairness, justice, and rights).

1.3 Moral Foundations Framing and Support for Climate Change Policies

Social media platforms do not merely serve as communication mediums, but they have become influential platforms that can stimulate moral emotions and play a determining role in shaping users’ moral beliefs, as well as the diffusion of moral emotions in digitalized networks (Brady et al. 2020; Crockett 2017; Huskey et al. 2018). Frames are often most effective when they appeal to morally relevant intuitions that are strongly held by an individual i.e. when they appeal to individuals’ moral foundations (Feinberg and Willer 2012; Haidt 2001). Moral foundations theory (Graham et al. 2013) is a psychological theory that proposes that human moral reasoning and judgement are based on a small set of innate moral foundations or intuitions. Moral Foundations Theory argues that human moral judgment is mainly intuitive and non-rational, and it can be divided into distinct categories of moral intuition. Each category processes different kinds of moral stimuli and helps maintain elements of human social interaction. The concept of “moral foundation” refers, thus, to a set of psychological principles or building blocks that underlie people’s moral judgments and ethical beliefs. These foundations are considered to be innate and universal, shaping individuals’ moral reasoning and guiding their perceptions of right and wrong across different cultures and societies. These foundations or moral intuitions include five core domains: care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, and sanctity/degradation. According to the theory, individuals differ in the degree to which they rely on each foundation in their moral reasoning. The theory has been used to explain variations in moral judgment and political ideology across cultures and individuals (Jonathan Haidt and Graham 2007).

These moral foundations are rooted in biological evolutionary benefits and universal across societies in spite of differences in their cultural expression. Care/Harm, exemplified by maternal care, is a value common in mammals with a biopsychological basis that is similar across humans and societies. Fairness/Cheating is linked to reciprocal altruism. Fairness involves temporary sacrifice for another’s benefit, with an overarching reciprocal advantage. Sensitivity to fairness is universal and evident early in human development. Loyalty/Betrayal relates to group belonging and allegiance towards a group, and is often expressed by in-group cohesion and conflict with out-groups. Authority/Subversion is akin to different forms of dominance and social hierarchy. Attitudes towards authority tend to differ across cultures and political ideologies. Sanctity/Degradation is related to the immune system and physiological disgust. This foundation also influences moral reactions towards certain social groups, such as immigrants and sexual minorities.

Drawing on cross-cultural research and advancements in intuitive processing and moral emotion studies (Haidt 2001), Moral Foundations Theory (MFT) was built on Shweder’s work, explaining psychological factors in moral disagreement, political values, and cross-cultural differences in moral judgment. MFT’s moral foundations of care/harm and fairness/cheating, based on (Shweder et al. 1997) of autonomy, focus on individuals as recipients of moral actions. They respond to individuals’ suffering, rights, and welfare. Care/harm addresses compassion and harm reduction, reacting to violence or suffering. Fairness/cheating deals with justice and equality, responding to perceived injustice or imbalances. The other three foundations, termed binding foundations, strengthen social cohesion and protect the collective, focusing on individuals as agents of moral acts. Derived from Shweder et al.’s ethic of community, the loyalty/betrayal and authority/subversion foundations emphasize group unity and hierarchical order. Loyalty/betrayal responds to group loyalty threats, while authority/subversion reacts to challenges against hierarchy and tradition. The sanctity/degradation foundation, akin to Shweder et al.’s concept of divinity, safeguards the in-group’s purity and is triggered by actions violating its essence. Graham et al. (2013) suggested additional foundations like liberty/oppression, but these have received limited empirical study. MFT posits that these universal moral foundations, influenced by culture, ecology, and individual differences, produce a wide array of moral judgments, values, and practices, suggesting that all moral judgments stem from these innate, evolved intuitions.

MFT posits that people have different moral intuitions that are triggered by different issues, and these moral intuitions can influence their attitudes and behaviors (Graham et al. 2013). Moral foundational framing has been shown to impact on charitable donations (Hoover et al. 2018), on attitudes and behaviors towards refugees (Mobayed and Sanders 2022), and on information virality in social movement protest (Wang and Liu 2021). Concerning climate change, Adger et al. (2017) have shown that people consistently use moral reasoning to explain their views on climate adaptation. Li and Su (2018) reviewed 10 experimental studies that looked at the effects of message framing on public engagement with climate change, using 26 comparison pairs. The study found that message framing generally had a positive effect on individuals’ engagement with climate change and its two subcategories – behavioral intentions and support for climate policy. It found that message frames emphasizing the environmental, economic, and moral dimensions of climate change had a small-to-medium impact on individuals’ engagement with climate change. However, message frames around public health implications or geographical identity had little effect on individuals’ engagement with the issue. The moral foundations of care and fairness have been found to be positively associated with support for individual-level actions to address climate change, such as reducing energy consumption, recycling, and using public transportation (Dickinson et al. 2016; Milfont et al. 2019). People who scored higher on measures of the moral foundation of authority and respect for tradition were less likely to support individual-level actions to address climate change (Milfont, Davies, and Wilson 2019). In the U.S., Feinberg and Willer (2012) found that liberals tend to view the environment in moral terms, while conservatives do not, and that this difference partially explains the relation between political ideology and environmental attitudes. The study also found that contemporary environmental discourse is based largely on moral concerns related to harm and care, which are more deeply held by liberals than by conservatives. However, reframing pro-environmental rhetoric in terms of purity, a moral value that resonates primarily with conservatives, largely eliminated the difference between liberals’ and conservatives’ environmental attitudes. The hypothesis according to which political polarization on environmental conservation issues may be caused by the chronic use of ideological and moral framing in persuasive messages that appeal more to liberals and egalitarians has been investigated by (Wolsko et al. 2016). Their study conducted a series of three experiments to examine the effect of moral framing on pro-environmental messaging on liberals and conservatives’ conservation intentions, climate change attitudes, and donations to an environmental organization. The results showed that while liberals did not differ across conditions, conservatives shifted substantially in a pro-environmental direction after exposure to a moral frame that portrayed protecting the natural environment as a matter of obeying authority, defending the purity of nature, and demonstrating one’s patriotism to the United States, underscoring the potential importance of moral re-framing in climate change communication.

1.4 The Norwegian Case

Norwegian civil society boasts a robust level of citizen engagement, evident in the widespread memberships and active volunteers within voluntary organizations. These organizations have historically fostered inclusivity, playing a pivotal role in promoting social integration and ensuring both social and political equality. Typically, these civil society organizations have adopted a hierarchical structure, encompassing local, regional, and national chapters, mirroring the popular movements model. Unlike societies where the state and civil sectors might be at odds, Norway thrives on a close-knit and cooperative relationship between the two. This harmonious bond underscores the nation’s state-friendly ethos.

The voluntary sector in Norway stands out for its impressive organizational density, housing a diverse array of organizations. In fact, nearly every interest or activity, be it political, economic, or cultural, finds representation within a dedicated club, association, or organization (Enjolras and Strømsnes 2018). However, over recent decades, the landscape of Norwegian civil society and its interplay with the state has evolved significantly. Key transformations include the emergence of novel citizen groups, a trend towards more specialized and decentralized organizational structures, heightened bureaucratization and professionalization, a transition from corporatism to lobbying, and the establishment of a renewed contractual relationship between the state and civil society (Enjolras and Strømsnes 2018).

Furthermore, according to Grendstad et al. (2006), the environmental movement in Norway functions distinctively compared to other countries. Firstly, Norway’s political and social framework is unique in its inclusivity; opposing interests are not sidelined but are often embraced by the government. The second anomaly is ideologically driven, emphasizing the protection of humans within nature. This perspective is shaped by Norway’s emphasis on social hardship and self-sufficiency in local living, influenced by its vast geography and low urbanization rate. Lastly, Norway, as an oil-producing country, faces a complex dilemma. There’s an inherent tension between its climate policies, which aim to reduce greenhouse gas emissions, and its economy, given that a significant portion of its public budget is derived from oil-related revenues. In this context, civil society organizations (CSOs) in Norway are actively advocating for a reduction in oil production to align with environmental goals and combat climate change. Yet, and despite an active civil society, a recent comparative polls^[1] indicates that 24 % of people in Norway think it is false that climate change is mainly caused by human activities.

These characteristics position civil society organizations as key players in Norway’s climate governance. Consequently, they are anticipated to leverage their social media communication as an instrument for normative governance.

1.5 Research Questions and Hypotheses

The research by Dickinson et al. (2016) found distinct differences in climate change attitudes based on political ideology and moral foundations. Liberals, who placed a high value on care and fairness, were found to be more willing to personally mitigate climate change, while conservatives valued sanctity more and were also positively inclined to take action, albeit less so than liberals. Interestingly, in-group loyalty and respect for authority – moral foundations typically associated with conservative ideologies did not show a significant association with climate change action. Therefore, we propose:

Furthermore, as the above literature suggests, different moral foundations may provoke different types of engagement. To explore the relationships between the five moral foundations and three types of public engagement on social media (affective, cognitive, and social spreading engagement), we advance the following research question:

Additionally, as the literature on framing indicates, different framings of the climate challenge by CSOs in their communication on Facebook are likely to influence the reception and levels of engagement with their messages. Topical framing has been shown to have a positive effect on individuals’ engagement with climate change (Li and Su 2018). More specifically, frames that emphasize the environmental, economic, and moral dimensions of climate change have an impact on individuals’ engagement with climate change while frames emphasizing public health implications or geographical identity do not influence engagement with climate change communication. Yet, we do not know a-priori which topical frames characterize CSOs communication about climate change on Facebook and consequently formulate the following research question:

Finally, different cognitive frames may also combine with moral foundations in promoting various types of social media engagement. Indeed, following dual-process theories, affective and cognitive systems, or “hot” and “cool” cognitions are expected to work together in producing engagement with communication on social media. While moral foundation framing operates through intuition – i.e. “hot” cognition – topical framing is perceived through “cool” cognition. However, “hot” and “cool” cognitions can be expected to provoke a stronger reaction when they converge i.e. when moral foundations and topics convey messages that are closely related. To explore these possible interactions between these two types of cognition and their influence on engagement – the ways moral foundations interact with climate change topical frames in predicting different rates of affective, cognitive, and social engagement with CSOs communication on social media – we propose the following research hypothesis:

2.1 Sampling and Data Collection

The data were gathered starting with a sample of Norwegian national civil society organizations that have responded to a nationwide survey in 2019 (Arnessen and Sivesind 2020). A national organization was defined as a voluntary organization where the activities and services have a national scope and can be used regardless of where one lives in the country. The survey focused on the whole population of nationwide organizations and not a sample of it. The population was established on the basis of information obtained from public registers. In all, 958 national organizations among a population of 3,875 organizations answered the survey, yielding a response rate of 25 %. Facebook posts data for all those organizations that have answered the survey and have a Facebook Public Page (among the 958 organizations having completed the survey, 660 had a Facebook page). Facebook was chosen because it is the most pervasive social media platform among Norwegian users. According to IPSOS (2021), 83 % of the Norwegian population has a Facebook profile (compared to 27 % for Twitter) and 69 % use the platform daily (compared to 9 % for Twitter). Additionally, nearly 70 % of voluntary organizations with a national scope use Facebook pages. Access to the Facebook pages and the crawling of their public content was enabled via Facebook CrowdTangle. The crawled CrowdTangle dataset included all the posts published by the 660 organizations having a Facebook Page and containing the term “climate change”, with a total of 2,884 posts posted by voluntary organizations with a Facebook page during the period January 1, 2018 to January, 1, 2023. Manual examination of the posts has confirmed that the posts included in the data set are related to climate change.

2.2 Analytical Approaches

To answer the research questions, several computational methods applied to a text corpus will be harnessed and combined. These includes topic modeling, semantic network analysis, automatic text processing based on a Moral Foundations dictionary, and text classification using Large Language Models.

2.3 Measurements

2.3.2 Independent Variables

2.3.2.1 Moral Foundations

The five moral foundation variables – care/harm, fairness/cheating, loyalty/betrayal, authority/subversion, and sanctity/degradation – have been computed using the extended Moral Foundation Dictionary (eMFD) method (Hopp et al. 2021), and retaining for further analysis sentiment score.

2.3.2.2 Frames Analysis: Topic Modeling

Topic modeling enables analyzing the textual content of the data to obtain an overview of the thematic structure of the updates on social media posted by CSOs. A forty-topic model emerged as the optimal solution to describe the topical structure of the Facebook posts by CSOs. The results of the analysis consist of a list of most frequent words for each of the 40 topics. The topics were interpreted and labelled manually after careful examination of the key-terms constituting each topics (see the Supplementary Information for details on the topic model). Each document (Facebook post) is a mixture of topics. The information regarding this mixture is described by the theta matrix produced by the Latent Dirichlet Allocation (LDA) implemented in R, in which columns represent the various topics and the rows represent the documents. Following the method proposed by Walter and Ophir (2019), the theta matrix – giving the probability of a topic in a given post [P(topic-k|document-d] – has been extracted from the topic model and the cosine similarity between topics based on the document they share has been computed.

2.3.2.3 Frames Analysis: Network of Topics

The pairwise cosine similarity between topics has been then used to create the semantic network, in which topics served as nodes and similarities between them as edges providing information about the extent to which topics are related. The semantic network offers a representation of the estimated topic model and constitutes the basis for the next step of the analysis. The next step consists in applying a community detection algorithm to the network in order to identify communities of topics that reveal common frames (Figure 3). Several community algorithms – Louvain, Walktrap, Spinglass, Greedy, and Eigen – were applied and the result of the Spinglass algorithm retained for further analysis and visualization in Figure 3. Based on this semantic network linking topics into “communities”, three communities or frames have been manually labelled: “climate politics” (green), “climate global crisis” (orange), and “climate mobilization” (violet) (Figure 3).

Figure 1:

Moral foundations probabilities computed with eMFD.

Figure 2:

Moral foundation sentiments computed with eMFD.

Figure 3:

Topic network drawn from CSOs Facebook posts. Nodes represent topics; edges represent co-occurrence of topics in documents (calculated using cosine similarity); size represents the prominence of each topic; color represents community membership. Network is weighted, fully connected and undirected.

Combining the results of the computation of moral sentiments and communication frames, with the dependent variables yields the variables to be used in the regression analyses, which descriptive statistics are displayed in Table 1. In addition, the variable “Followers” is used as control variable for regressing on the dependent variables. Our data source provided the number of followers of each Norwegian CSO’s Facebook account at the post’s publication time. We log-transformed the variable due to its skewness. Further, Whether a social media message contains richer media forms, such as photos, videos, or URLs, has been found to be associated with the level of public engagement (Liu, Xu, and Tsai 2020). To control for the effect of media richness, we adopted their measure by coding each post into one of the following three categories from the lowest to the highest level of media richness: (1) with hyperlinks (n = 1,684, 67.79 %); (2) with photos or other visual content (n = 602, 145, 24.23 %); and (3) with videos (e.g. YouTube videos, native videos uploaded directly to Facebook) or live streaming (n = 198, 7.97 %).

Table 1:

Summary statistics.

Variable	N	Mean	Std. Dev.	Min	Pctl. 25	Pctl. 75	Max
Total interactions	2,457	77.88	127.799	0	9	87	971
Comments	2,484	9.015	34.79	0	0	4	668
Shares	2,484	7.694	27.532	0	0	6	596
Affective positive	2,484	68.39	160.612	0	8	66	3,132
Affective negative	2,484	8.043	50.123	0	0	1	1,570
Care	2,484	0.648	1.069	0	0.094	0.807	25.331
Fairness	2,484	0.638	1.028	0	0.107	0.79	20.284
Loyalty	2,484	0.644	1.422	0	0.092	0.758	43.971
Authority	2,484	0.66	1.075	0	0.095	0.803	15.158
Sanctity	2,484	0.649	0.938	0	0.105	0.825	9.164
Autonomy	2,484	0.64	0.89	0	0.18	0.81	23
Community	2,484	0.65	1.1	0	0.17	0.8	30
Divinity	2,484	0.65	0.94	0	0.1	0.83	9.2
Frame-1 climate politics	2,484	0.202	0.049	0.058	0.172	0.224	0.619
Frame 2 climate global crisis	2,484	0.326	0.061	0.129	0.292	0.351	0.705
Frame-3 climate mobilization Followers Media richness 1 2 3	2,484 1,583 2,484 1,684 602 198	0.373 39,481.96	0.058 30,927	0.17 96	0.34 7,508	0.405 69,255	0.781 125,961