What is civil society and who represents civil society at the IGF? An analysis of civil society typologies in Internet Governance

Conference paper submitted for the GigaNet Annual Symposium 2020  Authors Nadia Tjahja (a,b), Trisha Meyer (a, b) and Jamal Shahin (a, b, c) a) United Nations University, Centre for Comparative Regional Integration Studies, Potterierei 72,   8000 Bruges, Belgium b) Vrije Universiteit Brussel, Institute for European Studies, Pleinlaan 2, 1050 Brussels, Belgium c) University of Amsterdam, Amsterdam School for Regional, Transnational and European Studies,     Kloveniersburgwal 48, Amsterdam 1012 CX, The Netherlands  Corresponding author Nadia Tjahja, ​nadia.tjahja@vub.be    Work in progress  Abstract


Introduction
For 15 years, the UN-mandated Internet Governance Forum (IGF) has brought different stakeholder groups together to engage in debate and discussion on public policy issues relevant to Internet Governance. The IGF has been a champion of the multistakeholder method, which brings together governments, business actors and civil society representatives. Much of the literature analysing the IGF (and other multistakeholder institutions) has tended to focus on the design of the multistakeholder models, or has provided critique on the potentials of these models to realise decisions, particularly focusing on the efficacy of the actors to achieve their goals. Yet, curiously enough, little research has been done on the composition of the actors engaged in these processes, and in particular, the heterogeneous group bundled together under the title 'civil society'.
We assert that one of the key reasons for this gap in the literature arises from the contested notion of civil society, one of the key actors in the multi stakeholder environment. Civil society engagement in the IGF is crucial, and yet there has been little research carried out on understanding who has actually taken part in IGF meetings under this moniker (see also Epstein and Nonnecke 2016) . In this article, we identify the conceptual gap between the elusiveness of the term, and the actual representation of civil society in the IGFs from 2006-2019. This paper aims to provide two contributions to discussions on norm entrepreneurship in internet governance. First, it aims to contribute to the growing body of literature on multistakeholderism, notably by furthering debates on the role of civil society in institutions of global governance (in this instance, the IGF). It aims to address criticism of how civil society is "deployed to legitimate and challenge the discourse and practice of global governance" (Amoore and Langley 2004, p. 89) . Second, we aim to contribute to the debates on the future of the IGF(+), by digging into the civil society stakeholder group to try to understand who is actually present in the discussions. This second objective is timely. Our contribution follows on from the publication of the High Level Panel on Digital Cooperation's report "The Age of Digital Interdependence" (United Nations Secretary-General's High-Level Panel on Digital Cooperation 2019) and the subsequent open consultation on developing updated mechanisms for global digital cooperation, including the potential design for an IGF+. There is a need to understand how the involved institutions view the role of civil society in this public policy space, and we attempt to provide some depth to discussions around civil society engagement at the IGF by analysing who is present. Our paper shows that using publicly available data on participation in IGF meetings, we can draw some tentative conclusions on who these stakeholders are.

Literature review
With this literature review, we intend to show that a large majority of literature in the field focuses on the effectiveness and legitimacy of civil society in political life, and that a conceptual gap exists between this literature focusing on the outputs of civil society and the composition of civil society (CS). The intention of this paper is not to provide a theoretical overview of all notions of this highly contested concept , although we 1 shall dedicate some space to describing some of the major tensions in the study of CS. We are primarily interested in the practical 'construction' of CS in international institutions. Through an analysis of the various definitions used by the World Economic Forum (WEF), the European Economic and Social Committee (EESC), the African Union (AU) and others, we hope to give some essence to the conceptualisation of civil society, which has been called "elusive" and "indeterminate" (Mouffe 2010) , or "ambiguous" by others (Amoore and Langley 2004) . We shall use this as a baseline for our reflection on the nature of civil society participation in the IGFs from 2006 -2019.
Before jumping into an elaboration of the major definitions that exist in these transnational institutions, we show how the confusion around the concept of CS has led to various different interpretations of their composition, functioning and role.
Civil society is considered to exist in the space between markets, public authorities, and private lives, or "the space in a society where collective citizen action takes place" (Malena and Finn Heinrich 2007) . It has often been linked to historical-cultural conceptions of politics and political interaction, and is essentially, in mainstream conceptualisations, seen as a space to discuss the emerging 'gap' in modern democracies between markets and states, CS, through the democratic tradition of 'stakeholder engagement,' has emerged to provide an opportunity to legitimise policy decisions made in neoliberal governance environments (Keane 2003) . In this sense, the integration of CS into policymaking processes has been seen as a way to ensure the legitimacy of those decisions, although this comes with a number of conditions (Foley and Edwards 1996) . CS is thus a conceptually and politically contested subject, even at the national level.
When thinking about CS beyond national borders, definitions become even more vague. First, CS is seen as a highly culturally-specific term that has been 'transferred' from Western perceptions of democracy onto the global scene (Kopecky and Mudde 2003) . Second, the use of the term 'civil society' within global institutional settings implies that some sort of space for collective action exists at the global level: i.e. the notion of CS engagement, even at the international or global level, is implicitly expected to generate effective and legitimate results or outcomes that feed into international policymaking processes. Although questions concerning the accountability of global civil society abound (Scholte 2004b;Erman 2010) , global civil society is seen as specifically serving the purpose of remedying the democratic deficits in global governance (Steffek and Ferretti 2009b) . Clearly, civil society as a domestic concept and civil society as a global one are not necessarily the same thing, but this begs the question what global norms are required to ensure CS can play a productive role in international multistakeholder fora (for an example from a different field, which highlights the confusion about conflicting definitions of CS, see Potjomkina et al. in press; see also Steffek 2013) .
Whilst efforts have been made to understand the conceptual and instrumental role (the effectiveness and legitimacy) of CS in international fora, the literature is largely silent on the actual constitution of CS in these spaces. This has resulted in a very broad brush being used to paint CS participation, which leads to conceptual and practical difficulties in understanding why civil society organisations (CSOs) participate in international fora. Thus, for the purposes of our research, the framing of Global Civil Society participation needs to take into consideration three different aspects: 1) the ambition to address global conceptions of civil society, rather than solely focusing on 'Western' type organisations; 2) the concreteness of the criteria for engagement in international fora as a member of 'Civil Society' (i.e. what does CS do there), and; 3) the 'place' that CS is given in the production of outputs whether they be policies, declarations, etc. In this paper, we focus on the first two aspects.
We do not assess the effectiveness, authority or legitimacy of CS, but focus on its composition, which we argue is a necessary step that needs to be carried out in order to understand how to engage effectively and legitimately with CS in questions of global governance. Through such a discussion, we contribute to a more 2 practical understanding of how CS participation is framed and deployed in IG. As a consequence of the reflections in this paper, we aim to contribute to discussions on how CS can engage more effectively in international and global governance.
Bearing these conceptual frames in mind, and given the lack of consensus in literature about the way in which CS is framed, it is apposite to engage in a discussion of how international organisations define CS and its role in their governance processes. The following paragraphs describe the definitions ascribed to CS by major transnational organisations. These definitions help us understand the organisational approach that dominates, allowing and enabling us to create a typology of CSOs.
From the collection of definitions that have been put together in Table 1 , we see a wide range of structures, norms, and roles emphasized for CS. Although the different definitions of CS treat it as one homogenous group, almost all definitions examined seem to recognise that this group comprises a large number of different actors that are diverse in structure, nature, and objective. Despite the diversity in definitions, there are also some key common attributes that can be extracted from the different frames.
First of all, civil society has a role in 'community representation'. It provides a collective representation of "objectives, constituencies or themes" (as per the WSIS definition) which go beyond the 'general interest' ascribed to public authorities. Second, civil society organisations do not represent their 'own' interests, but those of their community; they play the role of intermediary actor. Third, CS has no determinate organisational frame: CSOs can be formal or informally organised, with different legal and organisational constructions possible. Fourth, CS has no specific ties to geographic regions, but can represent local or global collections of interests. Finally, all definitions we examined focus on the necessity for CSOs to work as 'validation' for the multi stakeholder processes in organisations and formalised institutions, and thereby legitimise the institutional setting, by enabling for accountability through the checks and balances given by the presence of CSOs. 3 The working definitions provided above reveal the broad definitions that are used for CS in transnational and global governance institutions. The broadness of these definitions corresponds to much of the critique shown above that problematises our understanding of what CSOs actually are and subsequently how they can contribute to global policymaking.

Methodology
With our analysis, we seek to fill the gap on understanding what types of CSO participate in IGFs. We analysed the stakeholders that have been active at the Internet Governance Forum over the last 14 years with the aim of designing (and subsequently testing) a typology. This helps us address our research question: "Who represents Civil Society at the IGF?" . For practical reasons, we chose to focus on the on-site participation lists from 2006 until 2019. This information is publicly available, and can be found in the respective conference year at the bottom of the page of the IGF website homepage ("Documents") . Not 7 every year in our period had remote participation available, hence the decision to focus on on-site participation.
We outline our analysis in the three steps below in detail, as giving an understanding of how we conducted the analysis is crucial to validating the results. We also modestly propose that the IG stakeholder framework and the CS purpose-focused typology, with their accompanying definitions of stakeholders, could usefully serve to further both academic and policy debates on CS representation.
Step 1: Identifying civil society in the data set In recent years, the IGF has given registration options for four stakeholder groups: Government (by invitation only), Private Sector, Technical Community and Civil Society. However, this has not consistently been the case throughout the history of IGF participation. Throughout the years there have been instances in which multiple stakeholder groups were brought together under the label "Other", or separately listed (such as Academia and Press/Media). An overview of how CS was combined with other stakeholder groups over the years can be found below in Table 2 . To understand who represents civil society at the IGF, we had to exclude the other stakeholder groups that were included in the same category at registration and identify civil society in the data set. In line with the review of institutional definitions posted above in the literature review section (and based on (Meyer 2017) , we determined there are five distinct groups of actors that play a role in IGF meetings.

Figure 1 visualises this Internet Governance Stakeholder
Framework, focusing in particular on their interactions. As civil society is the subject of our analysis, we placed it at the centre of the framework. Figure  1 overlaps the four different stakeholder groups and acknowledges an often forgotten fifth stakeholder group -the End User. The inclusion of the end user category ensures that we are able to make a distinction between CSOs, which represent 'collective action', and individuals, who are also directly or indirectly an end user of the Internet (Bertola 2004, pp. 3-5) . Every individual regardless of stakeholder group is an user of the Internet and therefore this framework exists within the idea that we are all Internet users and that the Internet is created for all of us.

Figure 1: Internet Governance Stakeholder Framework
Next, we provide our definitions of the IG Stakeholder categories, to demonstrate how we sought to distinguish civil society in Internet Governance from other stakeholders.

End User
First, we based our definition of the End User on RFC 8890 and defined End User as follows: "End Users are [non-technical]  This definition clarifies that the ecosystem consists of individuals from a variety of different backgrounds, without specific, and perhaps even contrary opinions, who are directly and indirectly impacted by the Internet.

Technical Community
Second, the Technical Community stakeholder group is, according to our typology, strictly limited to the group of organisations that focus specifically on managing and developing the technical specifications of the infrastructure of the Internet. Whilst the line is sometimes blurred between the technical community and CS, we constructed our typology to make a clear separation between these two actors (see below, where we address organisations that crossover different actors). We make a distinction between the actors that play a role in governing the internet's infrastructure and the actors that play a role in governance on the internet. In this sense, we eliminated a number of actors who positioned themselves as CSOs, whilst having the remit of a technical community member.

Government
Government, or state actors, as stakeholders at the IGF have often been invited participants. However, for the purposes of our coding exercise, many actors whom we would categorise as having a public and political role identified themselves as civil society. "Government" in the IGF sense has also been closely related to national level ministers and actors (we note this has varied over the years), and therefore did not always include parliamentarians or local authorities. We have removed these from our analysis, as -according to our understanding of CS -these actors do not specifically represent CS, but rather act as public representatives. Similarly, with this rationale in mind, we have allocated International Governmental Organisations (IGOs) to this group as well, despite the fact that IGO representation at the IGF has traditionally been listed as separate (see Table 2 ).

Private Sector
The private sector actor group is, according to our typology, limited to the group of organisations that have a profit-driven rationale. These actors are distinguished from CS by their representation of their corporate (self-)interests.

Intersecting stakeholders
Many actors in the IG ecosystem actually position themselves at the interaction of two or more of these stakeholder groups. For the purposes of this paper, decisions were taken to ensure that the CSO group consisted of groups that had CS representation as their sole (or dominant) purpose. In most cases where intersections were apparent, the authors were able to identify the dominant purpose due to their participation in the IGFs. In order to ensure that these intersections were correctly coded, the coders regularly consulted with each other, and ensured for intercoder reliability.
Examples of intersecting actors include IGOs that promote citizen-driven projects (e.g. the Council of Europe's No Hate Speech Movement ), Corporate Social Responsibility initiatives that aim to provide Step 2: Who is present in the participant data? Listing self-identification and geography To understand who is present in the participant data, we looked at each civil society organisation listed and checked the about page on their website (or on a trusted third-party website, such as LinkedIn) to note how the organisations self-identified themselves. We developed a keyword mechanism to label their purpose or stated goals. We also noted the country in which their headquarters are established.
Step 3: Civil Society Purpose-Focused Typology Our literature review established that definitions of civil society differ, and do not provide a consistent overview of the types of actors. To identify patterns of purpose, which we identify as our contribution to generating a typology of CSOs, we worked deductively on the 4,675 individual data entries identified in step 1, based on the keywords we used to categorise their self-identification. Initially, we clustered the way in which organisations identified themselves and found four overarching themes: Coordination entities, End User (Group) Representation, Knowledge/Capacity Building and Problem-Driven Advocacy (see Annex B). These themes are described in detail in the following section.

End User (Group) Representation
Building on the "End User" stakeholder group from the Internet Governance Stakeholder Framework, we included the representation of defined groups of internet users. This group replaces categories that have traditionally been labelled separately as social groups, women's associations, youth organisations, indigenous peoples, minorities, etc. We also included organisations that represent communities who do not have access to the internet or do not have a direct connection in this group. Through our deductive categorisation of the CS, we identified organisations that represented people who may not know they are a stakeholder group.
However, we draw two specific distinctions that serve to provide nuance to our categorisation: • The End User (Group) Representation type focuses on CSOs that represent interests that serve the wider community. Therefore, organisations that represent solely their own self-interest have been excluded from this category. For example, industry associations who seek to improve standards of work in their sector. In essence, when coding, we asked questions such as: does the organisation focus on the rights of that group to the benefit of the wider community in general? Or are they focused on promoting their interest within their sector? • Organisations that seek to represent end user (groups) are distinct from organisations who are guided by a certain philosophy or by a certain group of individuals. For example, organisations that are guided by a philosophy do not per se represent end users. They may use their philosophy to run their organisation, for example, they may be a feminist group that organised coding classes, but that does not per se mean that they represent women. Similarly, organisations that are led by a certain group of individuals do not automatically represent end users. If an organisation is youth-led, they may consist of members who are youth, but do not necessarily represent youth participation; for example, a youth-led media literacy class.
We created these distinctions because traditionally end users have been grouped together with organisations that claim to represent specific groups of end users. Our typology distinguishes from those who represent the rights and values of end user groups with those who have other purpose-driven goals.

Knowledge/Capacity Building
This type focuses on organisations that are building (educating) and exchanging (sharing) information in communities to create a more informed society. This brings together educational institutions, research organisations but also press/media organisations which create or share information.
Within this type, we have three sub-categories: Academia, Press/Media and Generic. We have left these as distinct sub-categories due to the role that the categories Academia and Press/Media have played in the Internet Governance field as across the years, they have had their separate registration category available when participants could register their attendance (see also Adama Samassékou 2003 for a discussion on the position of Press/Media) . Therefore to assess their participation, they have been marked additionally. The generic sub-category indicted all other organisations that are not academia or press/media, but fit within this type.

Problem-Driven Advocacy
This type includes problem-oriented organisations which advocate, lobby and/or provide services to resolve issues that reflect a need of the community. They are often solutions-oriented, have specific end goals and may provide services. Unlike IGO frameworks, this type does not include specific topics such as human rights, copyright or poverty. Every year new themes and topics arise in the field of Internet Governance as the sector continuously innovates and evolves. This results in the creation of new types of activism. Therefore we have decided to focus on organisations who seek to solve problems they see in their communities as a category in our typology.

Coordination Entities
The self-identified list indicated that there are organisations that manage people or organisations that provide services to organisations to coordinate. This type are organisations that coordinate actors or organisations (such as volunteer-management organisations) that act to serve the community. It also includes organisations that provide resources and funding for other organisations to engage in activities that serve the community (such as funds and foundations). An organisation or an actor can be deemed a coordination entity only if the activities it is coordinating serve the community and not business interests.

Limitations
We acquired the data based on individual participants' self-identification and therefore relied on their honest submission. We did not verify whether they truly represent or are employed at that organisation, nor did we check the status of their affiliation. We were also required to exclude participants from the data when they fell under one of the following categories: "not traceable" when a person and/or organisation cannot be found, "no info" when not enough information was available to make an assessment, "page unavailable" when the website was offline, the domain was sold, was geoblocked or was under construction, and "security risk" when malicious content was detected. While we tried to the best of our capabilities to find out more information about the organisations through third party websites, this was not always available. This means that our sample size was reduced but not significantly so (see 5.1 below).
Another limitation for our data analysis was language of the organisation. As the IGF is a truly international forum, attracting international and local organisations alike, organisations will have websites often written in their national or local language only. For these websites we relied on Google Translate or Chrome's integrated website translation plug-in to provide us with the translations. There were also individuals who self-identified with translated institutional names, rather than registering for the IGF in the original language of their organisation. This led to some being untraceable; we tried to overcome this by searching for the person and the affiliation, which broadened the scope of finding the respective organisation. Furthermore, due to translated institutional names, but also the cultural background of the organisation, we can not assign self-identified types by titles. For example, 'association' in certain countries implies collections of organisations, in other countries it refers to a type of organisation; similarly, 'foundations' are synonymous with finance, but in some countries foundations are indicated as not for profits.
A third limitation concerning CSO participation in the IGF concerns the availability of data. The data analysed provides lists of all individuals who did not object to having their name published on the IGF website.
Efforts were made to overcome questions of intercoder reliability: although coding was conducted by two coders, the framework and typology was extensively discussed by the authors and cross-checking was carried out. Additionally, insecurities were marked in red for collective decision-making at a later stage.

Data
We consolidated the "Civil Society" entries of the on-site participant lists (see Table 2) into one list. For years when no distinct CS category was available, we used the "Other" group. In addition, for certain years, it was necessary to include the entries of the Technical & Academic Communities and Press/Media. In total, this list consists of 14,079 participants who registered to the IGF as CS, Other, Technical Community/Academia and Press/Media. From these 14,079 participants, we narrowed down our field of analysis to organisations, as our analysis in this paper focuses on CSO representation at the IGFs. We identified and removed individual or independent participants and duplicate entries of individuals from the same organisation. After a thorough data clean, we were left with a data set of 4,632 unique organisations that had identified as civil society or other. Then we coded the 4,632 organisations according to our Internet Governance Stakeholder Framework (see Figure 1) to identify the CSOs (Step 1). Second, we identified how organisations self-identified themselves on their website or a relevant third-party website if their own page was unavailable (Step 2). From this list, we identified commonalities and patterns and clustered the organisations in a Civil Society Purpose-Focused Typology. Lastly, we used this typology to manually code the full data set to be used for our analysis of CSO representation at the IGF (Step 3). As part of the data cleaning process, we excluded 1,802 out of 4,632 organisations, or 38.9% of the dataset, leaving us with 2,830 unique CSOs for analysis.
In particular, we excluded Private Sector (621, 13.4%) and Technical Community (165, 3,6%) entries from the dataset, which was expected due to the "Other" stakeholder group used during certain years. 183 (4%) organisations were Government and 44 (0.9%) organisations were IGOs. We did not expect any government or IGO organisations to be present in this list at all, given that these participants could have registered as governmental representatives. However, as we found entries for military, police, local councils, we added the code "Government" and "IGO" to the Codebook retrospectively for exclusion from the dataset.
526 entries (11.4%) were "not traceable". We searched on the organisation name and cross referenced it with the individual participant that registered under that organisation, or with another participant who registered with the same organisation if no results appeared. Entries were considered not traceable, when there is no reference of the organisation online, and/or the organisation was not listed on the individual's CV. Also considered not traceable was when it was not clearly identifiable that an entry referred to a specific organisation. For example, this was the case when a person listed their name on the list twice (as both participant and organization). We also coded 48 (1%) entries as "no info". Where there was no website available, we relied on third party websites to provide information. However, in some instances, no information was available to successfully identify and thus encode the entry.
191 entries (4.1%) were coded as "page unavailable". This indicated that the website was offline, geoblocked, pages came up with server errors (404), site was unable to load, website was deleted, or the domain name had been transferred or sold. We also marked 24 (0.5%) entries as "security risk". These websites were flagged by security software as malicious websites and therefore we chose not to access them. As we cannot access or trace unavailable websites, this conveys to us that interest representation is ever-changing and continuously evolving in the online space, possibly reflecting the sustainability of particular initiatives. It also points to the need for digital archiving for posterity.
The presence of IGOs and governments in the initial list also indicates a point for reflection in the way in which CS is defined by the actors themselves; i.e. should there be an explicit distinction made between the public sector and CS to avoid misrepresentation or skewing of the IGF representation data.
Intersecting stakeholder groups

Figure 2: Intersecting Stakeholder Groups
After excluding non-civil society stakeholders from the CSO list, the vast majority of remaining organisations (90.8%) were categorised as civil society only. A little less than 10% of the organisations in the unique list combined civil society with another stakeholder function. Many entries marked as civil society and government were IGF-related (e.g. national IGF). Although the IGF is a multistakeholder event, we coded it CS + gov as the IGF is embedded in a UN body and would not be possible without host countries' logistical support. ISOC in our coding can be found at the intersection of civil society and the technical community. For ISOC, it is worth bearing in mind that these entries are mentioned in the CSO list, because individuals self-identified as civil society. When other ISOC participants registered as being part of the technical community, they were not part of the dataset. Further we found that through our coding an interesting intersection emerged between civil society and private sector. Trade unions, bar associations, federations of journalists and corporate social responsibility initiatives, in our view, represent hybrid organisations. They are neither purely civil society, nor private sector. Portraying these intersections is helpful to understand the fluidity between stakeholder groups. As a final note, we would have had many more entries marked at the intersections if we had mapped the entries according to their funding streams (private/public/state). Mapping financial flows at this scale, however, would have required a level of analysis that we could not achieve at this stage.

Figure 3: Civil Society Representation per Country at IGF (Map)
Over the years, CSOs from 155 countries have attended the IGF. Remarkably, a mere six countries make up 1,113 of the 2,830 -or 39% of CSOs present at the IGF, while for 27 countries, only one civil society organisation has ever been able to attend. Perhaps most striking of all, in our opinion, is that it takes the sum of 95 (of 155 -or 61.3%) countries to reach the amount of civil society representation from the United States alone (294 CSOs).
We expected high representation from large countries that have hosted the IGF (Brazil 2007, 2015, Germany 2019, India 2008, France 2018). Yet, despite the fact that the IGF has not been held in the United States or the United Kingdom, these countries have contributed a substantially high number of CSO participants. Of the top 15 geographical entries portrayed in Figure 4, 10 have been host countries. Besides the United States and the United Kingdom, Belgium and the Netherlands also appear in the top 15 without having been an IGF host (at 14th and 15th place). For Belgium, this is due to labelling CSO location by headquarter, which meant that many European initiatives received a Belgian label. For the Netherlands, no distinctive patterns were found. However, the Dutch government has been a strong supporter and faithful donor to the IGF over the years, and we speculate that this may raise the level of awareness among civil society, along with the influence of the historically-active technical community in the Netherlands. To briefly touch upon the remaining three IGF host countries that do not appear in the top 15 list of countries represented by CSOs, Egypt (2009, 33, 1%) and Lithuania (2010, 31, 1%) do not fall far behind. Surprisingly, the IGF 2006 host country Greece has only had one civil society organisation present over the entire period.

Figure 4: Civil Society Representation per Country at IGF (Chart)
Further, because we chose to label the CSO location by headquarter, this does not necessarily reflect where the organisations are operational. To give some few examples, the Bahrain Center for Human Rights is based in Denmark; the civil society initiative Tavaana Tech is headquartered in the United States, but focuses on Iran; the Assembly of Pro-Democratic NGOs of Belarus is located in Lithuania. We also observed that some CSOs based in non-Western countries are funded or have been started by CSOs in Western countries (such as the Indonesian foundation Satu Dunia being itself founded on initiative of Hivos, headquartered in the Netherlands).
Finally, we marked "non applicable" (in 7th place) when either the CSO location could not be found through web search or the organisation has headquarters in multiple countries across the globe. This is for instance the case for Amnesty International, the Better Internet for Kids network, and many IGF and ISOC entries.

Civil Society Purpose-Focused Typology
Civil society types that "register" for the IGF Figure 5 provides insight into the purpose or role of civil society organisations that attend the IGF, which is based upon Step 3 of our data analysis. The vast majority of CSOs aim to build knowledge and/or capacity among their communities (76%), followed by advocacy to raise awareness and resolve societal problems (22.1%). 8.2% of CSOs represent end users directly and 14.6% take on a coordination role. Approximately 9 1/5th of the organisations (18.7%) combine different purposes, which can be found at the intersections of the venn diagram below. In the following sections we will elaborate on our findings within each of these purpose-focused subcategories.

9
The percentages mentioned in text combine all subsections of a civil society role (all parts of one circle in the venn diagram). We believe this is an accurate portrayal of the CSOs' purpose, but do note that this means the percentages add up to more than 100%. We invite the reader to consult Figure 5 for the disaggregated numbers for each subcategory.

Coordination
From the entire complete civil society dataset, we found that 412 (14.6%) out of 2,830 organisations had a role in coordinating people or organisations to provide services that have a community-focused goal. While there are organisations whose sole purpose is to provide coordination services, most organisations enjoy a dual type in which they coordinate, but combine with a problem-driven goal, the aim to share knowledge or build capacity, or represent certain vulnerable groups. These coordinating entities seek to foster relationships through collective action, likely indicative of a desire to participate and embody the multistakeholder model.
Within this subcategory, we identified funding bodies, such as funds, charities and foundations . Funds and 10 foundations, such as ISIF Asia and the Rockefeller Foundation, award grants to deserving projects that aim 10 'Foundations' do not immediately correlate to fund-awarding bodies. This depends mostly on cultural and linguistic background (see limitations section for more), and can be found across the entire typology spectrum.
to empower communities. Charities, such as the NSPCC, raise money to invest in expenses for volunteers, programmes and providing support in communities.
Other foundations, such as the Mozilla Foundation, supports fellows who are working on key Internet issues.
Additionally they connect open Internet leaders and rally citizens to the cause of an open Internet. Further, organisations, such as ICVolunteers, coordinate individuals to volunteer in organisations across the globe. Organisations like these provide coordination services for others to be able to provide community services.
We also found associations, federations, initiatives, coalitions and consortiums that collect people and organisations to collaborate towards common goals. To list some examples, the NGO Federation of Nepal's goal is to bring civil society in Nepal together to create a peaceful, democratic and just Nepal; and the association A New Governance gathers together over 200 organisations from 53 countries to address personal data issues.
Finally, the coordination subcategory features IG-unique organisations, such as the Internet Society's Local Chapters, the IGF Dynamic Coalitions, Regional IGFs, and National IGFs, which encompass the three forms of coordination mentioned in the previous paragraphs (funding, coordinating people, and collective action) by bringing IG people together for collective action through working groups and events, to provide funding for projects or travel and support to serve the IG community.

Figure 6: Coordination Entities
End User (Group) Representation

Figure 7: End User (Group) Representation
With the End User (Group) Representation subcategory, we analysed the entire spectrum (incl. intersections with other subcategories, or 8.2% of CSOs) to be able to understand which groups are represented in the community. The top three groups represent women (18%), youth (16.7%) and children (15.5%), which, combined, comprises 50.2% of our dataset in this category. However, this does not include the joint categories of women and children (3.4%) and girls (0.9%). Combined, they amount to 54.5% of our end user (group) representation dataset. This is significant in comparison to the other communities represented. We grouped the different professions together (15%); they are followed by persons with disabilities (6%), faith-based (3.4%) and minorities & indigenous peoples (3.4%). Other represented communities included LGBTQI (3.4%), Internet users (2.1%), specific geographic representation (1.7%) and consumers (1.7%).
The literature conveyed that traditionally organisations have included End User (Group) Representation by listing representation groups, such as women, youth, indigenous peoples, etc. Our data shows that representation includes new vulnerable groups, which would have normally fallen outside the traditional listed categories such as sex workers (Crested Crane Lighters) and veterans (Association Internationale des Soldats de la Paix). Now they have been grouped under "Profession", which also includes journalists, media workers, lawyers, pedagogues and librarians. This category is notable, while it can be remarked that professions are often represented by trade unions, there are unique types of users who have a profession but do not have an official body, and thus are not being represented in official institutions. Other unique groups include adult learners, domain name users, and victims and families of illegal immigration that may not have been considered as end user groups previously.
Delving deeper into the other major group categories, persons with disabilities (6%) includes representation of persons who are deaf, blind or visually impaired and have autism. Faith-based groups (3.4%) include christianity, buddhism and judaism. The minority and indigenous people includes the Dalit, indigenous people, Native Americans and the Roma (3.4%). Geographical representation (1.7%) refers to groups who have sought to represent their background as participants to structure, this included Hispanics, Africans and Bangladeshi. Although these groups are represented, it is clear that there is scope for more diversity in End User (Group) Representation to ensure that the IGF is open and accessible to all peoples from different backgrounds regardless whether that is cultural, geographical or based on their daily needs.
Knowledge/Capacity Building 1,800 (63.6%) of the 2,830 civil society organisations we analysed have the sole aim to build knowledge and capacity in their communities. If we combine this purpose with others, this increases to 2,149 (75.9%) of 2,830 of civil society organisations attending the IGF over the years seek to improve their communities through knowledge and capacity building efforts. It is worth reminding the reader that this is by far the largest subcategory that arose from the dataset.
Of these 1,800 civil society organisations, 999 (55.5%) represent academia and 396 (22%) have a press/media affiliation. If we compare academia to the total CSO representation, they comprise 35.3% of all civil society at the IGF, and press/media accounts for 14%. If we include press/media also present in other subcategories (an additional 22 organisations, e.g. journalist associations and media workers unions included in end-user group representation), the press/media figure increases to 14.8%, meaning that academia and press/media represent half (50.1%) of the civil society organisations present at the IGF.
It should be noted that our analysis of academia (universities, colleges, departments, centers, research institutes, etc.) did not distinguish between academic staff and students attending the IGF, as we only considered the institutional, not the individual level at this stage of the research. Arguably, students should be excluded from the dataset, if we consider that they attend the IGF representing their own end user interests rather than civil society interests. Additionally, as mentioned when describing the overall findings (5.1), we did not distinguish between public, private and state-owned (academic or press/media) institutions. This would have led to a normative debate under which conditions organisations represent civil society, compared to their owners/funders' interests. We did, however, exclude media industry groups, such as Time Warner and Bloomberg Industry Group, from the civil society dataset, marking them as private sector instead. Considering academia and press/media represent half of civil society organisations at the IGF, we thought it interesting to analyse where academia and press/media (in the knowledge/capacity building subcategory) hail from. We compared the top five geographical locations of these two stakeholders against the IGF host countries and the geographical locations of all civil society organisations (5.2) .  (67) Switzerland (63) Mexico (63) Brazil (135) United States (100) United Kingdom (73) Germany (56) India (52) France (49) Mexico (34) China (30) Switzerland (26) Canada (22) India (43) Germany (34) Azerbaijan (34) Turkey (32) United States (23) Indonesia (17) France (17) Lithuania (14) Kenya (12) Brazil (11) Academic institutions in Brazil, United States, United Kingdom, Germany and India are most represented at the IGF. Six of the top ten countries for academia hosted the IGF. Furthermore, the top six countries (United States, Brazil, Germany, United Kingdom, India, France) for all CSOs and academia are identical (as Table 4 shows the ranking differs, but all 6 countries are the same). This finding is to be expected (academia comprise 35.3% of all civil society at the IGF), it confirms that there is no noticeable discrepancy in overall and academic representation.
Turning to press/media, media organisations in India, Germany, Azerbaijan, Turkey, United States have attended the IGF most. Six of the top ten countries for press/media are the same as the top ten in the overall ranking, while nine of the top ten countries for media/press presence hosted the IGF. This latter finding is also unsurprising, as the IGF is likely to attract international press attention. Perhaps more striking is the minimal presence or lack of press/media at some IGFs: Greece (none), Egypt (2), Mexico (3) and Switzerland (1).
If we compare geographical location across all civil society, academia and press/media, we notice that three countries, namely the United States, Germany and India consistently feature in each top five.

Figure 8: Problem-Driven Advocacy issues
With Problem-Driven Advocacy, we sought to have a better understanding of the issues that the IGF community is addressing in their communities. Here we focused solely on the pure Problem-Driven Advocacy subcategory and did not include any of the mixed sections so that we could focus on single issues that did not overlap with the Coordination, End User (Group) Representation or Knowledge/Capacity Building subcategories.
The four biggest topics are Human Rights (11.6%), Safer Internet (3.8%), Sustainable Development (3.8%), and Democracy (2.3%), Digital Rights (2.3%), and Empowerment (2.3%). There is a strong focus on Human Rights, however within this type the organisations specified how human rights should be addressedwhether that was in communications, in digital or digital technology, on the Internet, in strategic litigation, to combat the threat of technology, to foster sustainability, to have in the digital environment or for women. Also other types emphasised their use: with "Access" we focused on getting access to specific areas or topics so it includes access to the internet, to abortion, rading through technology and technology in itself. With ICT, there was ICT4D, for Peace, regulation and human rights.
We noticed that there were several themes in which the Problem-Driven Advocacy issues could be grouped together. Political issues, such as autonomy, extremism and polarisation, and the Millenium Development Goals were raised. In addition to campaigns such as free public wifi, anti neo-nazism and EU accession. Other societal issues focused on community building and development, child protection, digital rights, civil responsibility and sustainability.
Among the topics are issues that are directly related to Internet Governance, such as data protection, digital rights, spam and copyright, but also issues that raise standards in our society, such as the Millennium Development Goals, Sustainable Development Goals and Human Rights.
An interesting pattern that we identified was the divide between civil society organisations that seek to 'empower' versus 'protect'. 'Empowerment' organisations focus on achieving particular goals for the Internet, such as the Free Software Foundation Europe (FSFE), which lobbies for software freedom. They believe that because software is ingrained in our daily lives, access and transparency are imperative. Organisations like the FSFE take an active stance in IG discussions and target policymakers to achieve particular goals or standards. On the other hand, 'protection' organisations address topics by which society is being (negatively) influenced. Organisations, such as the Family Online Safety Institute, aim to protect children online and focus on good digital parenting to ensure online safety. Other organisations that advocate for change seek to prevent harm from crimes, such as combatting online sex abuse (International Association of Internet Hotlines (INHOPE)).

Discussion and conclusion
This paper set out to identify what types of Civil Society Organisations participate in Internet Governance Forums. It has shown that a number of diverse actors are present, but also shows that these are geographically and thematically dominated by a handful of key actor types.
In order to assess the diversity of CSO participation in the IGFs from 2006-2019, we constructed a framework (theInternet Governance Stakeholder framework) that distinguished between different types of actor, and showed the intersections between them. It also revealed to us that many individuals were self-identifying as civil society but may indeed be representatives of other stakeholder types, such as governmental bodies, or even the technical community. In this way, we are able to show a more representative view of CSO participation in IGFs, which provides nuance to the discussions on civil society participation in this multistakeholder exercise.
Similarly, the typology, which has been developed as a reflection on the stated purposes of the participating CSOs reveals that there is a wide range of desired objectives and goals to be achieved by those who attend the IGFs. This is, in itself, a key finding that should, however, not be surprising to those who have participated in IGFs. The fact, however, that the data analysis shows that problem-driven advocacy groups, coordination bodies, knowledge/capacity building organisations and end-user groups are represented as one homogenous body of civil society may lead to confusion over the possibilities for concrete and measurable outputs to emerge from an IGF+.
We assert that this is related to the theoretical and institutional challenges to providing working definitions for civil society participation in multistakeholder exercises, as described briefly above in our literature review. Our paper has therefore described a typology of CSOs, defined along lines of purpose, rather than legal status or organisational capacity. This may facilitate the process of 'effective' engagement in future iterations of the IGF(+).

Opportunities for future research
This exploratory study has shown that there is a hidden wealth available in the statistics of the IGF that can inform our policymaking capacity in terms of topics, representation but also providing access and ensuring outreach to specific communities. This paper gave us the opportunity to explore the available data that is available in the IGF participants list. We used Microsoft Excel to manually code the 14,097 entries to understand the data thoroughly enough in order to design a codebook for automated categorising, which would be needed to design a full database. A subsequent research activity would be to analyse who is (registered civil society) at the IGF over time, which would allow us to examine the diversity of CSOs over different IGFs.
Additionally, although we have focused on CSO participation in this paper, the data we have collected also provide us with the opportunity to focus on individuals. This will allow us to i) provide a more granular analysis on the size of each representation, and ii) examine how individuals move across organisations over time.

Annex A. Internet Governance Stakeholder Framework Codebook Code Diagram
This codebook informs our decision-making in assigning our codes to organisations. Mostly coding was deductive because we anticipated finding codes in the data based on the Internet Governance Stakeholder Framework. However, the codes Government and IGO were inductive. We added the code Government during step 1 when we noticed organisations were included in civil society such as the military, police, local councils, etc. We grouped them and included them under Government. We added the code IGO during step 1 when we identified the presence of IGOs in the dataset. Identifying the stakeholders was important to our dataset because we required the exclusion of organisations which were not identified as Civil Society .

End User
Definition : The code End User refers to individuals or non-technical users whose activities the IG ecosystem is designed to support.