Tweeting in the ‘Expert Bubble’: Social Media Engagement of German Private Environmental Foundations

This paper explores Twitter adoption and social media engagement of private German environmental foundations. The study follows Lovejoy and Saxton’s (2012) approach to the hierarchy of social media engagement. It demonstrates the domination of an information provision role on Twitter and the almost equal relevance of action mobilizations and community building posts. At the same time, the study supplements the existing typology with the additional dimension of communication partners addressed in each type of tweet. Finally, using data from interviews, the study interprets and explains the role of social media tweets and patterns of engagement with different groups. In-depth analysis of tweets and interviews with the foundation representatives confirmed a limited use of social media as a means for dialogue and community development. Simultaneously, analysis shows that tweets posted by foundations predominantly address a professional community of other civil society actors, experts and politicians, creating an online expert bubble. Interviews confirmed that such online connections mirror offline cooperation networks that are perceived to be more important for successful communication and project development by the investigated organizations.


Introduction
The rapid rise and development of social media platforms have dramatically changed communication practices, providing new arenas for public expression, dialogue and mobilizations. Research on social media and non-profit organizations has thrived in recent years. While earlier scholarship optimistically viewed social media platforms as an equalising medium and stressed opportunities for stakeholder engagement, recent studies have taken a more critical stand. The substantial bulk of research shows that social media is employed by non-profit organizations predominantly for the purposes of information provision and rarely used to establish dialogue and deepen engagement between different actors.
Studies that investigate social media posts often focus on one of two broad research avenues: they either investigate content of the social media posts, or look at engagement potential and target audience . One of the most widespread approaches to the investigation of content relies on a three-fold hierarchy of social media posts developed by . Authors suggested that posts can serve three major purposes: information provision, community building or action mobilizations. The majority of studies based on this framework agreed that an information provision role dominated in social media posts of non-profit organizations, while a substantially smaller share of messages addressed community building or motivated actions (Park et al., 2016;Svensson et al., 2015;Zhou & Pan, 2016). Some studies presented contradicting evidence showing that certain non-profit organizations use social media more for community building and less for information sharing purposes (Auger, 2013;Campbell & Lambright, 2020). Given these mixed results, the present study attempts to & Galina Selivanova gselivanova@uni-bonn.de develop this promising theoretical approach and deepen our knowledge of the roles of social media posts. Studies that investigate communication audiences of non-profit organizations usually focus on strategies of community building and selection of the target audiences (Anagnostopoulos et al., 2017;. Previous scholarship suggests that organizations seek broader support and use social media to engage with donors (Maxwell & Carboni, 2016;Waters et al., 2009), and develop relationships with relevant stakeholders (Curtis et al., 2010). However, authors have also criticised non-profit organizations for not being able to promote and endorse their partners and collaborators online (Soboleva et al., 2017;Svensson et al., 2015). Relying on the social capital approach, (Saxton & Guo, 2020) convincingly argued that the benefits from social media engagement are tightly coupled with the quality and size of the social media audience. Importantly, literature that investigates the role of social media posts rarely overlaps with studies of target audience addressed through social media. The present study contributes to both streams of literature, focusing on the role of the contentaudience nexus. This article thus addresses three broad research questions: (1) How and for which purposes do German environmental foundations use Twitter? (2) Who are the main communication partners addressed in these tweets? (3) How does Twitter engagement fit in the general communication patterns of the selected environmental foundations?
On the empirical level, this research extends our knowledge, focusing on the understudied cases of nonprofit organizations-namely German private environmental foundations. 1 Previous research has predominantly addressed social media engagement of large US-based nonprofit organizations, while small community-based private foundations have received considerably less scholarly attention. In turn, smaller grassroots organizations are expected to be more interactive in their social media use . Furthermore, charitable foundations represent a special case of non-profit organizations with a higher level of financial autonomy and flexibility in terms of resource allocation (Goeke & Moser, 2021). While having attracted relatively little scholarly attention, private foundations present an interesting research puzzle that can illuminate interesting features of online communication.
The present study concerns relatively small, private environmental foundations based in Germany and investigates their patterns of social media adoption and online engagement. This research contributes to our understanding of the role of social media and the digital public sphere, illustrating the emergence of online communication bubbles and connections between online and offline realms. Content analysis of tweets confirmed domination of the information provision role of Twitter accounts and highlighted the higher frequency of tweets that engage other users, rather than serve the needs of foundations. Furthermore, findings suggest that foundations tend to communicate in the 'expert bubble', i.e., engage online with their professional communities. Twitter is rarely used to expand these communities, but helps foundations to reinforce and strengthen existing connections.
The study focuses on the level of social media posts and supplements this analysis with the organizations' own views on the role of social media. The goal of the study is addressed through multi-method research design that combines an in-depth description of Twitter use, based on the full sample of tweets; categorization of representative sample of tweets based on  typology and interviews with organizational representatives. The rest of the essay is organised as follows: the next section reviews previous scholarship, focusing on how nonprofit organizations, particularly charitable foundations, utilise social media. It also discusses literature that investigates stakeholders addressed by non-profit organizations online. In the following section, data and methods of the study are described. Next, results are presented, starting from the description of the typology of Twitter posts of the selected foundations focusing on groups that were addressed in each type of posts and concluding with the explanation of social media use. The final section summarises the findings and suggests further research directions.

Literature Review and Theoretical Underpinnings
The rapid rise and growth of social networking sites have triggered a new research agenda on the civil society use of Web 2.0 tools because they were expected to have an equalising potential of a low-cost communication medium, beneficial for less resourceful organizations (Coombs, 1998;Taylor et al., 2001). Following Kaplan and Haenlein (2010), social media is defined in this study as a ''group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of User Generated Content'' (Kaplan & Haenlein, 2010). Social networks such as Twitter allow not only individuals and groups to effectively broadcast to 'the world' (Palmer, 2014) or consume content, but to actively engage in content co-production and communication (Bonsón & Ratkai, 2013;Ellison & Boyd, 2013).
Later studies convincingly argued that organizations were not successful in the promotion of online conversations. Investigations of non-profits with different agendas, like sport (Anagnostopoulos et al., 2017) or healthcare (Ure et al., 2019), came to similar conclusions: social media is used to provide information and seek instrumental support, and to a lesser extent for interaction with the audience (also in . In this sense, Twitter is not different from other social media platforms: Bellucci and Manetti (2017, 897) claimed that only a handful of large US-based non-profits use Facebook 'to collect feedback and interact with stakeholders'.

Role of Social Media Posts
Lovejoy and  proposed an alternative approach towards organization's use of social media. Their study went beyond information sharing and dialogue dichotomy and suggested that posts should aim not only for starting an online conversation, but inspire activities among users. Authors proposed a hierarchy of social media engagement and developed a typology of roles that online posts serve. At the bottom of this hierarchy situated the most basic ''information provision'' category that requires the least level of engagement. Posts that are aimed at ''community building'' require more engagement, while ''action mobilizations'' posts were considered the most advanced form of communication. They concluded that the majority of tweets submitted by the largest US non-profits simply shared information about certain issues (56.8%). The second largest category was aimed at community building (25.8%) i.e., attempted to engage and strengthen connections with various stakeholders. The third and smallest category (15.6%) included tweets aimed at action mobilizations when users ''start doing something about it'' (Lovejoy & Saxton, 2012, 350). Similarly, Guo and Saxton (2014, 70) suggested that this threefold hierarchy corresponds to the stages of what they call ''mobilizations driven relationship building'' with specific groups of constituents. Their findings confirm that the first stage of information provision corresponds to reaching out to new audiences and logically constitutes the majority of social media posts; at the second stage organizations are trying to ''keep the flame alive'' and engage their communities; the third group of action-related tweets inspires stepping up to action and is usually aimed at the smallest group of devoted supporters.
The existence of hierarchy among the social media posts was confirmed by other research that investigated organizations with different agendas and situated across different political contexts Holmberg & Hellsten, 2016;Huang et al., 2016). These studies agreed that the most numerous category of posts was usually associated with information provision, followed by community building, while action-related posts appeared least frequently. In turn, Auger (2013, 374) concluded that among social media posts submitted by advocacy groups with different agendas, ''community function was most prevalent (55%), followed by calls to action (23%) and information about the organization (11%)' ' and Svensson (2015) demonstrated the prevalence of community-building posts with higher levels of two-way communication.
The only study that utilised a hierarchy of posts to investigate a sample of German organizations also classified 52.9% of posts as information provision, 26.82% as community building and 20.27% as action (Bürger, 2015). He demonstrated that subcodes related to events acknowledgements and events promotion appeared the most frequently in the sample. The author concluded that social media play different roles for foundations compared to other non-profit organizations: because foundations provide grants, rather than collect donations, they are less likely to ask for donations. Campbell and Lambright (2020) showed that US-based non-profit human service organizations used more action-related posts because they have higher levels of resource dependence and use social media to request support and actions from stakeholders. Following this line of argumentation, the opposite can be expected from charitable foundations that, as mentioned above, have lower resource dependence on external actors.
Despite some discrepancy between various studies, the hierarchy of social media posts proposed by  proved to be useful in bringing deeper understanding of online engagement patterns. For example, Saxton and Waters concluded that users ''liked'' dialogic messages and some mobilising messages more often, but were more likely to share one-way informational messages. Guidry et al. (2014) showed that information messages can produce higher levels of engagement, but calls to action and community building posts generated more direct responses from the audience. The present research suggests extending the application of  hierarchy to include another important aspect, namely communication partners addressed in social media posts.

Audiences of Social Media Posts
Scholars have argued that social networking sites could be particularly useful for civil society actors because they provide opportunities for low-cost communication with various actors (Waters et al., 2009) and allow the creation of virtual community spaces where like-minded people can communicate (Wertime and Fenwick, 2008) or ask for support (Bellucci & Manetti, 2017). Holmberg and Hellsten (2016) pointed out that non-profit organizations are more likely to target other users directly in social media than for-profit organizations. Saxton and Guo (2020) convincingly argued that benefits from social media engagement of any organization are tightly coupled with the quality and size of the social media audience of this organization. Furthermore, Ihm (2019) demonstrated that organisational activities generate various relationships with different stakeholders and various types of ties that produce different outcomes for online network structures of nonprofit organisations. Hence, it is important to investigate not only what organizations post on social media, but who is targeted in these messages.
Only very few studies have investigated communication partners and audiences addressed in the social media posts. Greenberg and MacAulay (2009) concluded that Canadian environmental advocacy groups, instead of simply announcing events or asking for donations, use social media to develop creativity and establish networks of collaborators. Briones et al. (2011) showed that the primary publics of the Red Cross social media communication included volunteers, the media, and younger users. Similarly, community sport foundations promoted their sports participation and education programs via Twitter because this platform has relevance for the younger population (Anagnostopoulos et al., 2017).
Previous research has also investigated strategies of community building and engagement tactics with specific audiences on social media platforms . Authors suggested that Twitter was used by USbased non-profits to inform and educate the general public on topics related to their missions. They concluded that public education and grassroots lobbying work better on Twitter because it is a tool for mass communication and tweets go out to everyone. However, Soboleva et al. (2017) demonstrated a relatively low and often indirect promotion of partners by a large US-based charitable organization and lack of strategic use of Twitter by these organizations for strengthening. Researchers often focus on how organizations seek support and engagement with their donors via social media platforms (Maxwell & Carboni, 2016;Waters et al., 2009) or develop relationships with broader groups of stakeholders (Curtis et al., 2010). For example, Zhou and Pan (2017) determined that Chinese grassroots organisations were more likely to communicate with volunteers via social media platforms due to a higher dependence on external resources compared to government organised non-profit organizations and corporate foundations.
Although some scholars have modified and extended  typology, to the best of our knowledge, there were no attempts to supplement coding schema with the information on target groups for different types of posts. This study aims to fill this gap and complement research on hierarchy of social media with the investigation of communication partners that are targeted by charitable foundations in tweets. The study suggests that the role of social media can be better explained by looking at audiences addressed in tweets from different categories. Furthermore, a broader look at online and offline cooperation priorities set by the selected foundations can give further insights of how social media is utilised by foundations. The study is guided by the following research questions: (1) How and for which purposes do German environmental foundations use Twitter? (2) Who are the main communication partners addressed in these tweets?
(3) How do patterns of Twitter engagement fit in the general communication patterns of the selected environmental foundations?

Methodology
The study followed several steps: firstly, descriptive statistics of Twitter accounts for each foundation and the content of their tweets were analysed. Secondly, qualitative semi-structured interviews were conducted with foundation representatives to assist interpreting the results of the first step. The multi-method design proposed in this study allowed for the triangulation of different data types that facilitated not only provision of in-depth description, but also explanation of social media engagement patterns. The sample for previous studies often included non-profit organizations with diverse agendas which can at least partly explain differences in social media use. The present research builds on the alternative approach and includes in the sample only organizations with an environmental agenda, eliminating it as a factor that accounts for differences in social media role. Furthermore, this study extends existing empirical knowledge looking at the understudied case of charitable organizations based in Germany.
Twitter data collection was performed with the NodeXL Pro software that provides abundant opportunities for collection of social media platforms information. This study focuses on the content of tweets, therefore all tweets submitted by each foundation were collected based on their unique IDs. Importantly, Twitter API limits data acquisition to the 3200 most recent tweets from each account, therefore some previous studies could investigate only the most recent tweets submitted by organizations. In turn, foundations selected for this study submitted less than 3200 tweets in total, making it possible to collect the full sample of their posts created between 2009 and 2020.
The full sample of tweets available for content analysis contained 8073 entries, 239 of which contained only URLs, Voluntas (2023) 34:540-551 543 and were deleted from the database. The remaining 7835 tweets were sampled randomly, controlling for the share of tweets produced annually by each foundation. The sample contained 2041 tweets or 26% of the database and was coded by three trained coders using MAXQDA20 software. Coding schema for content analysis was borrowed from  with three main types of tweets, and subcategories for the community building and action tweets. Following the original study by Lovejoy and Saxton and some later publications (e.g., Zhou & Pan, 2016) coders were instructed to identify the main idea of each tweet and classify posts based on the dominant communication roles. Coder training was based on several rounds of trial coding and discussions that brought agreement between coders to the level of 85%. When messages could be assigned to more than one classification, members of the research group discussed and resolved disagreements together. If one post served double or triple purposes and it was impossible to distinguish the main role, posts were classified according to the highest level of communication hierarchy (Action-Community Building-Information). The original coding schema was supplemented by the code for communication partners where applicable. ''Communication partners'' sub codes covered all major groups that foundations addressed in their tweets: NGOs, experts, business, political actors, media and the general audience.
The second stage of data collection involved semistructured interviews with the representatives of charitable foundations. Interviews were conducted in an online and offline format in September and October 2020. Purposive interview sampling targeted representatives of organizations who were responsible for media communication and public relations strategies. In total, interviews were conducted with thirteen foundation employees (see Table 1 in the appendix).

Typology of Tweets Posted by German Environmental Foundations
Similar to the previous research results, this study concludes that information provision tweets were the most numerous, with the total sample constituting 48.8% of all tweets. The second largest category in the aggregated sample was action-related (33.02%), followed by 17.9% of community building tweets. Results based on the aggregated data significantly differs from previous studies based on Lovejoy and Saxton's typology that usually classified action-related tweets as the least numerous category. Analysis of organization-level use of Twitter revealed the source of this serious bias: 67.5% of anstiftung foundation tweets were action-related tweets and only 18.2% shared information. Tweets submitted by other organizations followed a general pattern described in the previous literature that deployed this classification: on average 55.5% of their tweets played an information-provision role, while actionrelated tweets and community tweets comprised almost equal groups of 22.4% and 21.9% of the sample, respectively.
At the next step of analysis, subcategories of community-building and action-mobilizations Tweets were addressed. The most prominent action-related subcategories aggregated across all foundations except anstiftung foundation included tweets that contained some lobbying or advocacy efforts (5.31%), tweets that encouraged readers to join a website or vote for an organization (4.93%) and those tweets that promoted events (4.55%). In turn, Twitter was rarely used to encourage material support for foundations: subcategories that communicated needs of foundations were significantly less visible (donation appeals-1.03%; calls for volunteers-1.5%; and learning how to help-1.96%).
In the community-building category, two sub-subcategories appeared to be the most prominent: Twitter was used to acknowledge current and local events (7.4%) and to express gratitude and recognition of other users (6.2%). At the same time, in line with findings in previous studies reviewed above, tweets were rarely used as a two-way communication medium: on average only 2.65% tweets of all foundations including anstiftung foundation solicited response and only 1.67% were classified as responses to other users.
Although differences in subcategories on the level of individual organizations are relatively low, some foundations stood out as using Twitter for specific purposes. For example, anstiftung foundation used Twitter to promote Repair Café events that comprised 67.22% of all its tweets. The lobbying /advocacy subcategory was particularly relevant for Stiftung trias (21.6%) and covered 4.1% of the total sample.
To summarise, particularly prominent were posts that were addressed externally i.e. those that involved other organizations, partners or discussed events and projects. Those posts that addressed needs of organizations, i.e., directed internally, were used less by the foundations. This could be explained by the modus operandi of private charitable foundations that are less dependent on external resources than other types of non-profit organizations.

Audience of Twitter Accounts of German Environmental Foundations
The present study focuses not only on the hierarchy of tweets submitted by the selected environmental foundations, but also investigates the audience that is being addressed in these posts. In total, 1055 mentions of various actors were detected in the sample-a number which does not correspond to the number of tweets, because multiple actors were sometimes mentioned in one tweet. Importantly, foundations mentioned other users with the Twitter @user tool in 62% of tweets, while in the other 38% of tweets, actors were referenced by their name or title of organizations (i.e. indirectly). For example, Deutsche Wildtiere Stiftung tweets about cooperation with the Naturstiftung David (both included in this study): We have recently set up the first Batcorders in MV as part of our pug bat monitoring together with partners such as the David Nature Foundation. We give bats a voice!. 2 Further analysis clearly finds that foundations' communication is skewed towards other civil society actors, such as non-profit organizations, IGOs, think tanks, and other foundations (28.4% of posts with mentions), followed by experts (20.3%), political actors (19.2%), media (11.5%), business (10.9%) and the general public (9.2%). Two-dimensional analysis showed that up to 30.7% of tweets with information provision function, 29.9% actionrelated and 24.2% communication-related tweets, mentioned other civil society actors. The second most important communication partners were experts and professionals, who work on topics related to the area of foundations' interest: around 20% of information and community-related and 22.8% of action mobilizations tweets mentioned experts, professionals or university staff. Foundations actively promote participation in their partners' offline events, endorse their projects and acknowledge other non-profit organizations or experts.
Political actors of different levels were the third largest group addressed by foundations. This category included inter-governmental institutions (e.g. EU bodies), national and local political actors such as political parties, and single politicians. In the action-related subcategories, tweets classified as ''lobbying and advocacy'' (36.7%) and response-solicitation (31.3%), targeted political actors. The share of tweets directed at media outlets and journalists was unexpectedly low, with more information-provision tweets. Frequently they mentioned interviews conducted with foundation representatives by a media outlet discussing a past event and did not address mass media to promote themselves. Similarly, private sector companies were not mentioned often, and were more visible in community-building. The general public was addressed in 16.1% of community-building tweets, particularly those soliciting a response (43.8%), but appear rarely in information-provision or action-related categories. These findings suggest some attempts at community building and audience expansion, but do not show efforts to mobilise support or inspire any actions (Table 3).
Analysis of the representative sample of tweets confirmed the domination of the information-provision role of Twitter; with one prominent exception, action-related and community building posts appeared with equal frequency in tweets of the environmental foundations. More detailed analysis of subcategories unearthed the importance of posts directed at other organizations and individuals, rather than mobilising help and support for the foundations themselves.
The most frequent tweets mentioned other non-profit organizations or professionals who work on similar topics, invited politicians to dialogue or addressed them with lobbying and advocacy efforts. This section concludes that Twitter is used for the endorsement of the professional community and as an indication of relevant cooperation ties, although levels of dialogue and two-way communication remain low. In turn, it is visible that communication and engagement strategies in such online ''expert bubbles'' differ from communication with donors or the general public. Foundations engage through social media with actors who belong to the same field and professional cycle, and do not use common Twitter tools like @mentions and @replies to attract attention, create dialogue, or highlight and promote cooperation. Indeed, Twitter is a tool for mass communication and, by definition, such posts go out to everyone , hence targeting various users and specific audiences can be particularly challenging for organizations, especially those with less resources.
It can be further suggested that foundations' Twitter accounts are not used to engage new users or mobilise a broader circle of supporters because they mirror professional communities that already exist. Twitter communication relies on pre-existing connections and is used to strengthen ties, endorse and show gratitude to important partners.

Role of Tweets Explained
This section summarises explanations of Twitter engagement and selection of communication partners and target audiences by the foundations' representatives. The information is drawn from interviews conducted with representatives of private foundations. The main aim of these interviews was to go beyond a description of social media posts and explain visible engagement patterns.
In general, employees responsible for foundations' external communications confirmed that Twitter is ultimately a medium for information \ … [ It is more of a carousel of information, a take and give, than a communication platform (Stiftung trias). The overwhelming majority of foundation representatives stated that Twitter is used only as an additional platform where one can post something that cannot be posted on the website. Similarly, Succow Stiftung representative stressed the crucial information provision role that their account plays in the case of spreading their message: when it comes to facts, they need to be publicly corrected. Of course, there are threads and replies, but there are also [informational] messages.
Twitter is perceived as a give-and-take platform (Stiftung trias) used not only for information provision, but for the collection of information on projects of other stakeholders or news outlets: I also use social media as a news channel for myself and subscribe to relevant channels (SÖ L). Communication strategies rarely consider social media platforms in general, and Twitter in particular, as central communication tools. Interviews confirmed large diversity in communication strategies of the selected foundations which is conditioned by flexibility of resource allocation among charitable foundations. As hypothesised above, relative resource independence from external environments determines the communication strategies and target audiences addressed by foundations: It is important to mention the organizational form, because it makes our communication strategy very independent. We don't have to recruit members. And we have to attract donors to conduct our projects, but not to survive (Deutsche Wildtiere Stiftung). Foundations that seek external funding have to adjust their communication strategies according to the requirements of donors. For example, GNF has to comply with the requirements of donors and publish results of their work; Succow Stiftung created an additional Twitter account for one of its projects to ensure better information targeting: If you are more precisely thematically, you can reach special groups of interested parties better.
The majority of interviewees pointed at resource constraints that considerably limit all active discussions and engagement on Twitter: long Twitter-threads are beyond our capacity (Stiftung trias) or limit social media use in general: If we had more resources, we could certainly set up an Instagram account or be more active on Twitter. This is not happening at the moment because of resource reasons (Stiftung Pfadfinden). In some cases, foundations deliberately focus on offline communication due to the audiences they address: over 50% of our target group are older supporters \ … [ some of them we cannot reach via social media, instead we distribute emails, or letters by post with a manual signature. The latter works almost better than email. Our donations also show us that personal addressing is preferred by this generation. (Jäger Stiftung). A combination of organizational traits and limited resources of small private foundations forces them to focus their communication strategies on conventional one-way forms of communication and limit social media use.
Interviewees acknowledged that Twitter communication clearly flows in the direction of politics and the professional publics (Facho¨ffentlichkeit) in the broadest sense, partly also the press. (Stiftung trias). Similarly, the Deutsche Wildtiere Stiftung representative explains that on Twitter you can reach many of the so-called decisionmakers, yes, journalists, politicians and so on, that is why they primarily use Twitter to lobby. Other organizations stress that, as non-profit and non-political actors, they are very careful in connecting with, retweeting or following business representatives or politicians: That might make a wrong impression if we follow specific political actors. Of course, they might be interesting for us, but it can also signal that I represent opinions, so I am careful (Stiftung David). Hence, even under conditions of resource constraints, some foundations act strategically on social media. Representatives of the anstiftung further described the selection of friends and followers as a conscious and mindful process: When I see that someone is following us, we also follow them. I don't do that in everyday life either, it happens in waves. I sit down and concentrate on who has subscribed to us recently or who has been following us recently. Then, I check their accounts and follow back or not. \... [ These can be organizations or individuals, but I look at their profiles and what they are doing, and what their latest news was. If that suits us as a foundation or I see that it is an individual who can be interested in our work, I follow them back. If I have a feeling that it's a commercial provider who just wants to create a network, I won't follow them back because we as a foundation don't do anything that has a commercial approach.
Not all foundations described a rigorous selection of their communication partners, but most of them stated that Twitter is used to connect with like-minded individuals and organizations with similar agenda who tweet in the vicinity of relevant topics (Stiftung für Bären). Similarly, an SÖ L representative explains that organic industry is relatively manageable, i.e. everyone has contacts with each other, I see that myself, too: many of our followers also know each other and I can meet them online or in other contexts. In this sense, interviews supported our previous findings: foundations select communication partners among experts and other civil society actors from the same field.
Finally, interviews have confirmed close connections between online and offline communication activities of the selected foundations. Twitter accounts of these organizations are linked to the partners familiar to the foundations through their offline work. At the same time, project partners are selected mostly offline because we all already know them (Stiftung für Bären). Similarly, an SÖ L representative explained that on Twitter he orients himself towards those actors that we also meet in daily work. It can be concluded that Twitter and other platforms are rarely used to expand communities, reach out to new audiences or search for new partners. Foundations selected for the study use Twitter to interact with the existing community and to enhance these connections. Foundation representatives often stressed that calls for support or mobilization function better through non-social media tools e.g. calling directly, arranging offline events, or engaging relevant target groups through direct non-digital invitations. These results echo suggestions made by Zhou and Pan (2017) who described the non-profit sector in Western democracies as comprised of ''digital migrants''. Unlike in China, the emergence of civil society in Germany did not co-occur with the emergence of digital technologies; they established their mode of operation long before social media development and now face difficulties in adopting new technologies.

Conclusion
The study applied  hierarchy of social media posts to examine use of Twitter by German charitable foundations. Firstly, the study provided a comprehensive review of social media use and role of Twitter posts; secondly, it focused on the communication partners of these foundations. Finally, it outlined explanations of Twitter engagement suggested by the foundations' representatives, putting social media use in a general context of communication strategies. The study contributes to our understanding of social media engagement of private foundations and provides additional empirical evidence to the broader research agenda on the role of social media and dynamics of the digital public.
At the first glance, these findings are consistent with previous scholarship: analysis of a representative sample of tweets showed the limited relevance of Twitter as a platform for two-way communication and dialogue and confirms the particular prominence of an information provision role, while action-related and community building tweets were less prominent in the sample. Foundations attempt to use Twitter as a platform for communication, but for a number of reasons these attempts are limited. Simultaneously, interviewees stressed that Twitter helped to acquire information on events and projects of their partners-hence it plays not only information provision, but also an information acquisition role. Further analysis of the 'Information-Community-Action' hierarchy of social media posts reflects types of online engagement that foundations offer to their audience without development of two-way communication. Following Bürger (2015), present analysis of tweet subcategories has indicated that they were more often aimed at the foundations' external milieu, rather than the satisfaction of their own needs. Low frequency of such tweets as 'calls for volunteers' and 'donation appeals' might be explained by the specific nature of private foundations that have relative resource independence compared to other types of non-profit organizations. At the same time, analysis demonstrated that tweets were used for advocacy efforts, events promotion, or expression of recognition and gratitude. This is a clear demonstration of attempts-probably not always successful or skilful-to address relevant communication partners. These results suggest that, despite all limitations, Twitter is used as a tool to connect with stakeholders, highlight important contacts and monitor their activities.
Content analysis demonstrated that the majority of communication partners addressed via Twitter belonged to the professional networks of the selected foundations. Most frequently, tweets mentioned other non-profit organizations, experts and politicians, while media, private sector and the general audience were addressed less frequently. Although foundations are more flexible in defining their communication strategies, they select means to reach relevant audiences and put emphasis on specific communication means, often leaving other social media tools behind. Nonetheless, our study confirms that social media communication partners were selected strategically by the majority of organizations to demonstrate common interests, follow relevant stakeholders or emphasise the non-political stand of some foundations. Interviews confirmed that online connections often mirrored offline contacts and partnerships of foundations. Social media in general, and Twitter communication in particular, relies on these already existing connections and is used to strengthen ties, endorse and show gratitude to their offline partners. The study concludes that instead of expanding community, Twitter reinforced offline professional communities and created an online ''expert bubble''.
The study furthermore indicated specific traits of communication and engagement patterns in such ''expert bubbles''. Unlike previous scholarship that relied on a quantitative approach analysing factors relevant for social media use (e.g. Bürger, 2015;Campbell & Lambright, 2020;Huang et al., 2016;Zhou & Pan, 2017), the present study used qualitative data to unearth underlying explanations of social media engagement. For example, analysis suggests that community building tweets are less relevant for foundations because Twitter communication is aimed at sharing information with already known partners. To a certain extent, the study shares criticism of earlier scholarship (e.g. Soboleva et al., 2017) and suggests that foundations do not employ Twitter tools to enhance communication with stakeholders and make clearer demonstration of relevant connections.
Overall, findings presented in this essay contribute to a broad research agenda on the digital public sphere. These results tap into the ongoing debate on online flows of information, selective exposure and the development of communication bubbles (Pariser, 2011), contributing to the discussion on the emergence of what we call ''expert bubbles''. Such communication bubbles are characterised through the exclusion of certain voices, by processes of social selection and community formation (Nguyen, 2020). On the one hand, these findings share a ''pessimistic'' view on the digital public sphere, warning against the fragmentation and development of homogenous closed communities (Schäfer 2015). On the other hand, our study has demonstrated a close connection between online and offline public spheres in the case of private charitable foundations. Previous scholarship produced mixed evidence investigating connections between online and offline activities, ranging from a lack of connections to strong interdependence (see overview in Greijdanus et al., 2020;Koltsova & Selivanova, 2019;Zhuravskaya et al., 2020). The present study confirmed interrelation between offline and online activities, demonstrating that, in some cases, offline communications spill over into the online realm and produce online ''expert bubbles''.
Knowing that charitable foundations are relatively flexible in defining their agenda and projects, the study concludes that changes in their offline activities and engagement with different actors can also expand their online communities and potentially mitigate the ''bubble'' effect. This cautiously optimistic interpretation stretches beyond the results of the present study and calls for the further investigation of the dynamics of online and offline communication and the evolution of ''expert bubbles''. Undoubtedly, while results presented in the paper have limited generalisability due to their focus on environmental foundations based in Germany, future research can use these findings as a starting point and incorporate organisations with different agendas located in different social and political contexts. This study illustrated mirroring offline connection in the digital public sphere among ''digital migrants'' in a democratic political context. Future research on the engagement patterns and stakeholders involved in online communication can use these results to compare the development of the digital public sphere in democratic and non-democratic societies.

Appendix
See Tables 1, 2 and 3. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons. org/licenses/by/4.0/.